| { |
| "paper_id": "J77-1004", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T02:49:38.165706Z" |
| }, |
| "title": "A JUNCTION GRAMMAR APPROACH", |
| "authors": [ |
| { |
| "first": "Alan", |
| "middle": [ |
| "K" |
| ], |
| "last": "Melby", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Eldon", |
| "middle": [ |
| "G" |
| ], |
| "last": "Strong", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Ronald", |
| "middle": [], |
| "last": "Lytle", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Millett", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Fredericka", |
| "middle": [], |
| "last": "Bell-Berti", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Katherine", |
| "middle": [ |
| "S" |
| ], |
| "last": "Harris", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Some", |
| "middle": [], |
| "last": "Acoustic", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "", |
| "pdf_parse": { |
| "paper_id": "J77-1004", |
| "_pdf_hash": "", |
| "abstract": [], |
| "body_text": [ |
| { |
| "text": "t h e i r g e n e r a l approaches.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The r e s e a r c h d e s c r i b e d i n t h i q paper uses J u n c t i o n Gramnar as a t h e o r e t i c a l b a s e , and L i n e a r P r e d i c t o r C o e f f i c i e n t (LPC) methods as aq a n a l y s i s -s y n t h e s i s t e c h n i q u e . M o t i v a t i o n s f o r t h e s e d e c i s i o n s a r e presentee S & c t i o n I b e g i n s with a n e x p l a n a t i o n of some s e n t e n c e s which a r e b e i n g s t u d i e d . 'For example, t h e r e i s l i k e l y a stress on \"study\" i n t h e s e n t e n c e he boys who s t u d y g e t good g r a d e s , \" i f t h e c o n t e x t i s \"but t h e boys who d o n ' t g e t bad grades.\" The message r e a l i z a t i o n forms one s t r u c t u r e a s a w h o l e . I t s c o n s t i t u e n t s -a c o u s t i c r e a l i z a t i o n , higher l e v e l prosody, and s y n t a x -s e m a n t i c s -i n t e r a c t ~i t h each other very c l o s e l y ; a d e c i s i o n niade a t any l e v e l d e r i v e s immediately from t h e o b t a i n e d r e s u l t a t t h e l e v e l above, and a f f e r t s 2 decisgon a t t h e level below.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The remainder of t h i s s e c t i o n c o n s i s t s of a survey of some of t h e c u r r e n t work i n t h i s a r e a i n t h e USA ( a t MIT, B e l l Labs, and Stanford U n i v e r s i t y ) , i n Germany, and i n the USSR. Then the s e c t i o n w i l l conclude w i t h an i n t r o d u c t i o n t o the p r e s e n t r e s e a r c h .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "A t MIT, Allen (1976) i s working on p i t c h control as an element i n h i s o v e r a l l p l a n t o produce a system capable of producing s y n t h e t i c sp,eech from u n r e s t r i c t e d E n g l i s h text. H e p o i n t s o q t t h a t although a s y n t a c t i c and semantic a n a l y s i s i s needed, n b e x i s t i n g automatic algorithm can provide that a n a l y s i s reliably f o r e n t i r e sentences of u n r e s t r i c t e d t e x t . So h e h a s e l e c t e d t o do a l o c a l a n a l y s i s of t h e s e n t e n c e f i r s t and then t i e t o g e t h e r t h e local a n a l y s e s i n t o a s e n t e n c e l e v e l a n a l y s i s i f p o s s i b l e . The analyzer i s thus designed so t h a t i f a t some point complete s e n t e n c e analysis i s blocked, the p a r t i a l a n a l y s e s a r e s t i l l u s e f u l i n g e n e r a t i n g t h e p i t , &~v e s t i g a t e such q u e s t i o n s as when and why elements of a v e r b s t r i n g a r e s t r e s s e d . For example, he no'tes t h a t t h e s e n t e n c e \"A farmer was e a t i n g t h e carrot1' w i l l r e c e i v e emphasie on \"eating\" if'it i s i n r e s p o n s e t o a questJon about what the farmer i s doing. A l l e n currectky n o t e s t h a t :", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "A . MIT", |
| "sec_num": null |
| }, |
| { |
| "text": "The discovery and c o o r d i n a t i o n of all these effects 1 s a l a r g e and c o n t i n u i n g e f f o r t , and i t i s c l e a r t h a t ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "A . MIT", |
| "sec_num": null |
| }, |
| { |
| "text": "A t S t a n f o r d U n i v e r s i t y , t h e r e i s a research p r o j e c t on g e n e r a a v e pl-osodics ir the Jsstitute f o r Mathematical Studies i n the Social Sciences (IMSSS). Researchers on t h i s p r o j e c t a r e developing a system which, u l t i m a t e l y , i s intended to d o s y n t h e s i s in real t i m e for use in computer-assisted i n s t r u c t i o n a t IMSSS ( L e~i n e , 1976). Their technique i s t o compile a lexicon of words i n LPC code (Atal and Eanauer, 1971) and t h e n , when a given s e n t e n c e i s to be s y n t h e s i z e d , c o n c a t e n a t e t h e code f o r each word, a d j u q t i n g d u z a t i~n s and p i t c h c o n t o u r s as needed.", |
| "cite_spans": [ |
| { |
| "start": 482, |
| "end": 506, |
| "text": "(Atal and Eanauer, 1971)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stanford University", |
| "sec_num": null |
| }, |
| { |
| "text": "Whife O l i v e throws away t h e o r i g i n a l p i t c h contour of each word, t h e IMSSS approach i s to adjust t h e original contour of the ward and t h e n f u r t h e r smooth t h e contour s o t h a t each word will n o t sound s e n t e n c e f i n a l ,", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stanford University", |
| "sec_num": null |
| }, |
| { |
| "text": "The IMSSS group u s e s t h e ideas of Leben (1976) , who r e l a t e s h g l i s h prosody t o tone languages i n t h a t he views both tone languages and English as having a suprasegmental melody which i s combined with t h e segmental phonolbgical elements. The IMSSS group (Levine, 1976: 3 ) defines melody as a sequence of \"auto segmental tones (autonomous from t-he phonological segments) s e l e c t e d from t h e t o n a l r e p e r t o i r e of t h e I n Germany Isacenko and Schadlich ( 1 9 7 0 ) , performed a n i n t e r e s ting s e r i e s of experiments on t h e p e r c e p t i o n o f German i n t o n a t i o n . In answer t o q u e s t i o n (4) above (concerning t h e c h o i c e of an a n a l y s i s s y n t h e s i s t e c h n i q u e ) , w e have chosen t o work i n i t i a l l y w i t h an LPC synthesis technique (as did O l i v e a t Ball Labs and Levine a t S t a n f o r d ) because an LPC s o f t w a r e package w a s a l r e a d y a v a i l a b l e a t BYU.", |
| "cite_spans": [ |
| { |
| "start": 39, |
| "end": 51, |
| "text": "Leben (1976)", |
| "ref_id": null |
| }, |
| { |
| "start": 277, |
| "end": 295, |
| "text": "(Levine, 1976: 3 )", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stanford University", |
| "sec_num": null |
| }, |
| { |
| "text": "But long range p l a n s i n c l u d e t h e use of an a r t i c u l a t o r y f u n c t i o n a l model (Flanagan, 1975).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stanford University", |
| "sec_num": null |
| }, |
| { |
| "text": "We now t u r n our attemioa to certain l i n g u i s t i c phenomena which w e c o n s i d e r e s p e c i a l l y interesttng. First, we w i l l i l l u s t r a t e the phenomena with sample sentences which will be d i s c u s s e d i n Nevertheless, you a r e s u r e t h a t i t was e i t h e r John or Mary, I n t h i s c o n t e x t , you would p u t stress on \"John\" and on \"Mary\" and a f a l l i n g pitch a t t h e end of t h e sentencze. Then you would expect a r e p l y of \"John\" o r \"Mary.\" expanding t h e s e n t e n c e t o \"The boys who study g e t good g r a d e s but t h e o t h e r s do n o t .\" I f \"study\" i s stressed, \"others\" i s i n t e r p r e t e d as \"boys\", namely the boys who do n o t study. I f , however, \"boys\" is stressed, \"others\" may no l o n g e r be i n t e r p r e t e d as \"boys,\" b u t it can b e i n t e r p r e t e d as \"girlst' o r '\"men who study\" o r some other group of woilld b e used, f o r example, in t h e s e n t e n c e \"John, our mailman, i s going t o retire m March,\" t o show t h a t t'John,'t and \"our mailman\" a r e d e f i n i n g t h e same person independently ( c f * the t r a d i t i o n a l restrictive n o n -r e s t r i c t i v e d i s t i n c t i o n ) .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "In the above examples, we considered full s u b j u n c t i o n s , (e.g.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "\"John, our mailman\") but the same s p e c i a l i z a t i o n s apply t o interjunctfons, Remainder. The second t y p e of s p e c i a l i z a t i o n mentioned i n r e l e v a n t to t h e e n t i r e sentence. The A-tree system of r e p r e s e n t a t i o n i s v e r y . \"John drove t o the store\" Versions lble (2) I n an i n t e r j u n c t i o n , any r i g h t i n t e r j u n c t i o n causes a stress on t h e primary operand, and a left hyphen subjunction causes a stress on the V3 o f the subor- I n the c a s e of the sentence a t hand, t h e i m p l f c i t frame 11 m o d i f i e r , being a r i g h t i n t e r j u n c t i o n , causes the primary operand, t h a t i s , t h e element t o which t h e Frame I1 f e a t u r e i s a p p l i e d , t o be s t r e s s e d . Thus, we have chosen t o work w i t h a n e n t i r e s e n t e n c e as a u n i t . E s s e n t i a l l y , (1) The topmost node of t h e J-tree defines a phrase.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "EQUATION", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [ |
| { |
| "start": 0, |
| "end": 8, |
| "text": "EQUATION", |
| "ref_id": "EQREF", |
| "raw_str": "A ( d i d ) PV + I A v N &or N come John Mary A come Ei .&. B D i d or John Mary (stress)", |
| "eq_num": "(stress" |
| } |
| ], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "(2) If the ptedicate consists of more than a single vexb and a single object, the verb and object w i l l be made into a phrase which w i l l then be joined t o the subject.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "(3) The cantents of each subordinate tree of the J-tree (which i s e forest of trees), i s phrased under the dominating tree. The c r i t e r i a f o r a s s i g n i n g [+ stress] t o a node are:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "(1) A Frame XI f e a t u r e i n the J-tree,", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "(2) A l e f t o r r i g h t hyphen sub j u n c t i o n ( i n d i c a t i n g remainder),", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "(3) The operands of an \"OR\" i n t e r r o g a t i v e . ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I* THEORY", |
| "sec_num": null |
| }, |
| { |
| "text": "To conclude this chapter w e present some graphs of p i t c h contours for the sentence \"The boys who study g e t good grades.\" Figure 1 7 shows a natural, a rule-generated and a manual p i t c h contour for sentence 3b", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 128, |
| "end": 138, |
| "text": "Figure 1 7", |
| "ref_id": "FIGREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "D. Sample P i t c h Contours", |
| "sec_num": null |
| }, |
| { |
| "text": "(\"The boys who study g e t good grades\"). Figure 18 shows a natural and a rule generated pitch contour for sentence 3c (\"The boys who study g e t good grades\"). Note that these two contours ar imposed on the same s e t of LPC analysis parameters t o produce the two readings. ", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 42, |
| "end": 51, |
| "text": "Figure 18", |
| "ref_id": "FIGREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "D. Sample P i t c h Contours", |
| "sec_num": null |
| }, |
| { |
| "text": "I have one. -E have one.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "20.", |
| "sec_num": null |
| }, |
| { |
| "text": "The c a t that the dog chased g o t away.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "4 .", |
| "sec_num": "2" |
| }, |
| { |
| "text": "John buys r i c e ? 18. They are eating app-s.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "26.", |
| "sec_num": null |
| }, |
| { |
| "text": "They are e a t i n g a p g l e a .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "19.", |
| "sec_num": null |
| }, |
| { |
| "text": "22.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "19.", |
| "sec_num": null |
| }, |
| { |
| "text": "-1 have one. ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "19.", |
| "sec_num": null |
| }, |
| { |
| "text": "The cat t h a t the dog chased Table 1 gives t h e r e s u l t s of t h e first p a r t , where sentences were r a t e d on a s c a l e from 1 ( m e c h a n i~a l ) t o 5 ( n a t u r a l ) . h'atural p i t c h contours r e c e i v e d the h i g h e s t s c o r e as expected, followed by manual contours based on the n a t u r a l contour, rule-generated c o n t a u r s and monotone 11 contours\" i n t h a t order. ", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 31, |
| "end": 38, |
| "text": "Table 1", |
| "ref_id": "TABREF4" |
| } |
| ], |
| "eq_spans": [], |
| "section": "25.", |
| "sec_num": null |
| }, |
| { |
| "text": "Some of the s e n t e n c e s were n o t w e l l t r a n s m i t t e d by the above d e f i n i t i o n . A c o n s i d e r a t i o n of t h e s e i n d i c a t e s t h e kinds of problems that arose. For example, since the f i r s t word of any normal d e c l a r a t i v e s e n t e n c e r e c e i v e s some extra s t r e s s , t R e listeners had difficulty d i st i n g u i s h i n g \"John drove t o t h e s t o r e \" from \"John drove t o the store. APPENQIX C", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "C. Transmission Problems", |
| "sec_num": null |
| }, |
| { |
| "text": "The p i t c h contoux g e n e r a t i (3) a. The boys who study get good grade&.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "COMPUTER IMPLEMl3NTATION", |
| "sec_num": null |
| }, |
| { |
| "text": "b. The boys who study g e t good grades.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "COMPUTER IMPLEMl3NTATION", |
| "sec_num": null |
| }, |
| { |
| "text": "c. The boys who study get good grades. ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "COMPUTER IMPLEMl3NTATION", |
| "sec_num": null |
| }, |
| { |
| "text": "A. Below are two l i s t s of t h e same 34 sentences. You w i l l hear t h e f i r s t l i s t w i t h a $ second pause after each sentence. Just l i s t e n and don't write anything. Then 10 secqnds l a t e r , you w i l l hear t h e second l i s t with a 3 second pause after each sentence. This time, during t h e pauses, r a t e each sentence by w r i t i n g down a number a f t e r , 1. I h a v e o n e .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "I. NATURALNESS OF INTONATION", |
| "sec_num": null |
| }, |
| { |
| "text": "T h e cat t h a t t h e d o g c h a s e d g o t a w a 11 2 l t m e a n s \"b\" ''3\" m e a n s I t I1 C , l r 4 \" m e a n s \"d\" . The quqstion, as h e s t a t e s i t , i s :", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 16, |
| "end": 53, |
| "text": "t t h e d o g c h a s e d g o t a w a", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "2,", |
| "sec_num": null |
| }, |
| { |
| "text": "Given t h a t t h e r e is a d i s c r e t e l i n g u i s t i c i n p u t to the mechanism of speech production a t some s t a t e , and given t h a t t h e mechanism t h a t transmits t h i s input i Incoming information, on the o t h e r hand, i s decoded to o b t a i n t h e pragmatic J-tree which stimulated i t , and then each junction i n the tree is executed by a semantic processor, r e s u l t i n g i n a The p o i n t again, of\" course, i s t h a t while articulatofsy structure and semantic s t r u c t u r e are symbalically r e l a t e d , they axe n o t the same and should not be confused o r intermingled. The effects of the terminal consonant on the midpoint of t h e s t r e s s e d vowel are not as l a r g e as t h o s e of t h e i n i t i a l consonant. In other wordb, t h e carryover effect of t h e first consonant on t h e s t r e~s e d vowel i s larger than t h e a n t i c i p a t o r y effect on t h e second. (2) The syllables are diagrammed using the a p p r o p r i a t e adjunction type.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "2,", |
| "sec_num": null |
| }, |
| { |
| "text": "(3) An interjunction i s constructed using syllable-final and s y l l a b l ei n i t i a l c o n s t i t u e n t s , (The label node i s given as C s i n c e bseems t o exert assimilative force over m.) - ( 4 ) The l a b e l node of the sub juxlction attaches t o t h e more heavilystressed s y l l a b l e . 5 ", |
| "cite_spans": [ |
| { |
| "start": 205, |
| "end": 210, |
| "text": "( 4 )", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "2,", |
| "sec_num": null |
| }, |
| { |
| "text": "t h i s appendix, the reader can determdne from the codes which v e r s i o n w a s used for each question.)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "2 ( R , N ) 23(Hl,H4) 2 4 ( N , R )", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Synthesis of Speech from U n r e s t r i c t e d ~e x t", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Allen", |
| "suffix": "" |
| } |
| ], |
| "year": 1976, |
| "venue": "Proc. IEEE", |
| "volume": "64", |
| "issue": "", |
| "pages": "433--442", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Allen, J. (1976) \"Synthesis of Speech from U n r e s t r i c t e d ~e x t , \" Proc. IEEE, Vol. 64, No.4, pp. 433-442, A p r i l 1976", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Speech analysis and s y n t h e s i s by l i n e a r p r e d i c t i o n of the speech wave", |
| "authors": [ |
| { |
| "first": "B", |
| "middle": [ |
| "S" |
| ], |
| "last": "Atal", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [ |
| "L" |
| ], |
| "last": "Hanauer", |
| "suffix": "" |
| } |
| ], |
| "year": 1971, |
| "venue": "J. Acoust. Soc. Amer", |
| "volume": "50", |
| "issue": "2", |
| "pages": "637--655", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Atal, B.S. and S . L . Hanauer (1971) \"Speech analysis and s y n t h e s i s by l i n e a r p r e d i c t i o n of the speech wave,\" J. Acoust. Soc. Amer., Vol. 50, N0.2, pp. 637-655", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Linguistic Units and Speech Production, t 1 an invited paper presented at the 85th meeting of the", |
| "authors": [ |
| { |
| "first": "", |
| "middle": [], |
| "last": "Peter F A Maondilage", |
| "suffix": "" |
| } |
| ], |
| "year": 1973, |
| "venue": "", |
| "volume": "13", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Peter F a MaoNdilage, \"Linguistic Units and Speech Production, t 1 an invited paper presented at the 85th meeting of the Acoustical Society of America, Boston, Massachusetts, A p y i l 13, 1973, 3", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "Towards a UnifiedqPhonet!c ~h e o r", |
| "authors": [ |
| { |
| "first": "Philip", |
| "middle": [], |
| "last": "Lieberman", |
| "suffix": "" |
| } |
| ], |
| "year": 1970, |
| "venue": "Linguistic Inquiry", |
| "volume": "3", |
| "issue": "", |
| "pages": "307--322", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Philip Lieberman, \"Towards a UnifiedqPhonet!c ~h e o r y , \" Linguistic Inquiry, Vol. I, No. ' 3 (July, 1970), 307-322.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Chain as an Analysis-By-Synthesis Model", |
| "authors": [ |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Farid", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Onn", |
| "suffix": "" |
| } |
| ], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Farid M. Onn, ' '~~e e c h Chain as an Analysis-By-Synthesis Model;", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "The initial exposition of junction the~ry appears in Eldon G", |
| "authors": [], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "The initial exposition of junction the~ry appears in Eldon G.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "~tructural Derivation in Russian", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Lytle", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Grammar O F Subordinate", |
| "suffix": "" |
| } |
| ], |
| "year": null, |
| "venue": "19'73. Additional article$ on junction grammar In -B W Linguistics Symposium Proceedings", |
| "volume": "1", |
| "issue": "", |
| "pages": "1971--1976", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Lytle, A Grammar o f Subordinate Structures in English, The Hague: Mouton and Co., 1 9 7 4 , and also in Eldon G. Lytley \"~tructural Derivation in Russian,\" unpublished Pb .D. dissertation, University of Tllinois (Champaign-Urbana), 19'73. Additional article$ on junction grammar In -B W Linguistics Symposium Proceedings, 1971-1976.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "The Case for Case", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Fillmore", |
| "suffix": "" |
| } |
| ], |
| "year": 1968, |
| "venue": "Universals. in Linguistic Theory", |
| "volume": "", |
| "issue": "", |
| "pages": "1--89", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "b~harles J. Fillmore, \"The Case for Case, \" Universals. in Linguistic Theory, ed. Emrnon Bach and Robert T. Harms (Holt Rinehart, 1968), pp. 1-89.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "The tprm phoneme is here used in reference to an articulation unit", |
| "authors": [], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "The tprm phoneme is here used in reference to an articulation unit, not an a c o u~t i c unit.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "On t h e o t h e r h a d , i f t h e c o n t e x t i s \"but t h e g i r l s who s t u d y g e t poor g r a d e s , \" then t h e r e i s probably stress on \"boys.\" The v a r i o u s r e a d i n g s of \" t h e boys who s t u d y . . . \" and o t h e r s e n t e n c e s a r e e x p l a i n e d w i t h i n t h e 3ut1ct;Son Grammar framework. An overview i s g i v e n of a system f o r g e n e r a t i n g p i t c h c o n t o u r s f o r a s e n t e n c e f r o m a J u n c t i o n Grammar s e m a n t i c o -s y n t a c t i c r e p r e s e n t a t i o n . S e c t i o n I a l s o i n -l u d e s a d e s c r i p t i o n of an e x t e n s i o n of J u n c t i o n Grammar whi& d e f i n e s a n o b j e c t c a l l e d a n a r t i c u l a t i o n t r e e , correspondtng t o each j u n c t i o n tree. A j u n c t i o n t r e e c o n t a i n s semanticos y n t a c t i c i n f o r m a t i o n b u t no l e x i c a l i n f o r m a t i o n . An a r t i c u l a t i o n t r e e c o n t a i n s segmerital i n f o r m a t i o n about e a c h l e x i c a l item and s u p r a s e g m e n t a l o r prosod lc informatiofi combining t h e l e x i c a l items i h t o p r o s o d i c u n i t s . Semantic d i s t i n c t i o n s i n j u n c t i o n trees a r e recoded as d i s t i n c t i o n s i n t h e p r o s o d i c s t r u c t u r e of a r t i c u l a t i o n t r e e s and t h e n a r t i c u l a t i o n Vrees a r e used t o g e n e r a t e p i t c h c o n t o u r s . J u n c t i o n t r e e s and a r t i c u l a t i o n trees are i n c l u d e d as f i g u r e s f o r s e v e r a l s e n t e n c e s . Sectqon I1 d e s c r i b e s -h o w p i t c h c o n t o u r s a r e g e n e r a t e d , i n c l u d i n g t h e r e c o d i n g of j u n c t i o n t r e e s a s a r t i c u l a t i o n t r e e s , t h e assignment of ~n i t i a l and f i n a l p i t c h l e v e l s and p i t c h a t n u c l e a r s y l l a b l e s , apd h o w t h e g e n e r a t e d c o n t o u r s a r e combined w i t h a n a l y s i s p a r a m e t e r s and s y n t h es i z e d i n t o speech. It s h o u l d b e noted that t h e j u n c t l o n t r e e s are e n t e r e d manually r a t h e r than by a u t o m a t i c a n a l y s i s , i n t h e c u r~e n t implementation. The t e * t i n c l u d e s s e v e r a l g r a p h s of n a t u r a l p l t c h c o n t o u r s as w e l l g~ c o n t o u r s g e n e r a t e d by t h e computer s y s t e m . The p i t c h c o n t o u r system produces a s y n t h e s i s o u t p u t foL e a c h r e a d i n g of a s e n t e n c e . T h i r t y -f i v e s e n t e n c e s , some w i t h n a t u r a l , some with hand-drawn, and some w i t h m a c h h e -g e n e r a t e d p i t c h contours were e v a l u a t e d f o r n a t u r a l n e s s and \" i n t e l l i g i b i l i t y \" of i n t o n a t i o n i n four t y p e s of tests. R e s u l t s of t e s t i n g s e v e r a l s u b j e c t s showed t h a t t h e g e n e r a t e d p i t c h c o n t o u r s were judged n e a r l y as n a t u r a l as hwnan-produced c o n t o u r s , and e x c e p t f o r some s p e c i f i c problems i n v o l v i n g d u r a t i o n , t h e g e n e r a t e d c o n t o u r s were i n t e l l i g i b l e i n t h e s e n s e o f c a u s i n g t h e l i s t e n e r t o p e r c e i v e t h e i n t e n d e d r e a d i n g of t h e s e n t e n c e . The t e x t l n c l u d e s a q u a n t i t a t i v e summary of t h e r e s u l t s of the e v a l u a t i o n . For t h e c o r p u s of s e n t e n c e s t r e a t e d so f a r , J u n c t i o n Grammar", |
| "num": null |
| }, |
| "FIGREF1": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "computer based t e x t s y n t h e s i s systems r e q u i r e a means f o r generatips s e n t e n c e -l e v e l p i t c h c o n t o u r s . These c o n t o u r s mvst have a c e r t a i n d e g r e e of \"human fidelity\" if t h e s y n t h e t i c speech i s t o sound natural, t h a t i s , n o t t o o machine-like. The p i t c h c o n t o u r s i n c u r r e n t l y o p e r a t i o n a l t e x t s y n t h e s i s systeqs are s t i l l n o t p e r f e c t l y n a t u r a lsounding and thus c o m p u t e r g e n e r a t i o n of p i t c h c o n t o u r s i s a t o p i c of c u r r e n t i n t e l ' e s t . This i n t e r e , s t i s shown, f o r example, by A l l e n a s he d i s c u s s e s pause a n d d u r a t i o n i n t e x t s y n t h e s i s and t h e n g o e s~o n t o say: I f temporal c o n t r o l presents great problems i n t h e d e s c r i p t i o n of s p e e c h , t h e n t h e problems of fundamental f r e q u e n c y $fO), o r p i t c h c o n t r o l , are a t l e a s t a s d i f f i c u l t , Once a g a i n , problems a r i s e due t o t h e f a c t t h a t t h e \u00a30 is c o r r e l a t e d with many f a c t o r s , i n c l u d i n g vowel tongue h e i g h t , p r e v i o u s consonant, breath g r o u p c o n t o u r , s y n t a c t i c and s e m a n t i c c o n t e n t of words, whether a sentence i s a questfon, intonation e f f e c t s , and word boundary g l o t t a l i z a t i o n . ( A l l e n , 1976: 440) Given t h e need f o r further r e s e a r c h i n p i t c h c o n t r o l , a q u e s t i o n remains of how t o approach t h e problem. The a u t h o r s f e e l i t i s i m p o r t a n t t o work within a l i n g u i s t i c model t h a t i n t e r r e l a t e s s e m a n t i c and p h o n e t i c phenomena. L a t e r on i n A l l e n ' s a r t i c l e h e makes t h e f o l l o w i n g s t a t e m e n t (which c o i n c i d e s w i t h o u r p h i l o s o p h y ) : The c u r r e n t use of s o p h i s t i c a t e d means f o r p i t c h r e c o r d i n g , coupled wf t h increased i n t e r a c t i o n between l i n g u i s t i c s and s p e e c h r e s e s r c h e r s , s h o u l d , however, l e a d t o s i g n i f i c a n t l y improved pitch control programs which are based on sound l i n g u i s t i c a l l y m o t i v a t e d t h e o r y . ( A l l e n , 1976: 4 4 The need f o r i n t e r a t i o n between l i n g u i s t i c s and speech r e s e a r c h i s f u r t h e r explained b y U m e d a (1976: 450):", |
| "num": null |
| }, |
| "FIGREF2": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "c h contour and o t h e r p r o s o d i c c o n t r o l s such a s d u r a t i o n and pause. In response t o toe need f o r a t h e o r e t i c a l framework f o r r e l a t i n g a t e x t and i t s p i t c h contour, A l l e n i s u s i n g t h e i d e a s of H a l l l d a y (1970) ( e . g . d i s c o u r s e focus) to", |
| "num": null |
| }, |
| "FIGREF3": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "s u b s t a n t i a l setnantic and d i s c o u r s e -l e v e l knowledge i s needad t o correctly predict p r o s o d i c p a r a m e t e r s . \" ( A l l e n , 1976: 441) B. B e l l Labs S e v e r a l w o r k e r s a t Bell Labs have a t t a c k e d the problem of : o n t r o l l i n g pitch in speech s y n t h e s i s , Olive (1975) d e s c r i b e s a system \u20acor g e n e r a g i v g p i t c h contcrulrs f s r t h e s e n t e n c e t y p e \" a r t i c l e -s u b j e c tv e r b -a r t i c l e -o b j e c t \" w i t h a n o p t i o q a l a d j e c t i v e on t h e s u b j e c t o r o b j e c t . TTi s method for: g e n e r a t i n g t h e p i t c h c o n t o u r w a s t o r e c o r d s e v e r a l sentences of the specified t y p e u s i n g random words and t o a v e r a g e t h e natural pi'tch c o n t o u r s t o o b t a i q p r o t o t y p e c o n t o u r s . Then t h e c o n t o u r far each word was approximated b y a f o u r t h ~r d e r polynomial t o \" f a c i l i t a t e l i n e a r stretching a n d compressfon o f the f u n d a m e n t a l f r e q u e n c y c o n t o u r . 11 O l i 3 e r e p o r t s t h a t by u s h g t h i s p i t c h c o n t o u r g e n e r a t i o n s y s t e m , <n c~n j u n c f i o n w i t h a w o r d c o n c a t e n a t i o n schesrie i n which the words a r e s t o r e d i n l i n e a r p r e d i c t o r c o e f f i c i e n t (LPc) code, the s y n c h e s i z e d s e ntences w e r e of high quallty Umeda, at B e l l Labs, i s a l s o concefned w i t h p i t c h c o n t o u r s , 11 8 s s e r t i n g that Among a c o u s t i c companents, p i t c h ( t h e fundamental frequency of t h e v o i c e ) shows the a~s t direct r e l a t i o n t o higher l e v e l prosody, s t r e s s and b o u n d a r i e s \" (Umeda, 197'6: 4 4 8 ) . Umeda's a l g o r i t h m f o r c o n t r o l l i n g p r o s o d i c paldiueters i s based on n s y n t a c t i c a n a l y s i s of t h e input t e x t . The analyzer f i t s e a c h clause i n t o a $emplate c o n s i s t i n g of t h e f o l l o w i n g o p t i o a a l s l o t s : s e n t e n c e m o d i f i e r , s u b j e c t , verb, o b j e c t or complement, r a i l m o d i f i e r , and p u n c t u a t i o n mark. A poPnt where t h e above o r d e r of t e m p l a t e elements i s v i o l a t e d i s marked as a boundary, and bvundaries a r e later u s e d t o a s~i g n pauses and intonation (Umeda, 1975).", |
| "num": null |
| }, |
| "FIGREF5": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "These tones are treated t h e o r e t i c a l l y a s d i s c r e t e fundamental frequency levels, b u t t h e n they are realized p h o n e t i c a l l y as c o n t i n u~u s -contour&. I n o r d e r t b a s s i g n tones t o key s y l l a b l e s , a program analyzes t h e sentefice t o be s y n t h e s t z e d using a s i m p l e p h r a s e s t r u c t u r e grammar which brackets phrase^, clauses and o t h e r complex c o n s t i t u e n t s , and i n d i c a t e s boundarierj between maj o r c o n s t i t u e n t s . D. Germany Complementary t o p i t c h contour g e n e r a t i o n , i s t h e s t u d y of t h e p e r c e p t i o n of p i t c h c o n t o u r s .", |
| "num": null |
| }, |
| "FIGREF6": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "a t u r a l s e n t e n c e s i l l u s t r a t i n g d i f f e r e n t i n t o n a t i o n p a t t e r n s were r e c o rded and monotonised a t v a r i o u s fundamental f r e q u e n c i e s ( e . g . 150 H e r t z and 178.6 H e r t z ) . Then t h e t a p e s of t h e monotone v e r s i o n s were c u t and s p l i c e d a t v a r i o u s p o i n t s . The s p l i c e d t a p e s t h u s had an a r t i f i c i a l l y s i m p l i f i e d i n t o n a t i o n o f e x a c t l y two t o n e l e v e l s . The team found t h a t t h e y could change t h e way l i s t e n e r s p e r c e i v e d c e r t a i n ambiguous s e n t e n c e s by changing o n l y t h e p o i n t s a t which t o n e s w i t c h e s o c c u r r e d .E. USSR I n t h e USSR, fiaavel et! a l . (1976) have a l s o performed some experiments i n m a n i p~l a t i n g p i t c h c o n t o u r s w h i l e l e a v i n g o t h e r parameters c o n s t a n t . They a r e i n t e r e s t e d i n f i n d i n g ways t o \"decrease t h e amount of in\u00a3 ormation n e c e s s a r y f o r t h e d e s c r i p t i o n of p i t c h c u r v e s w i t h o u t d i s t o r t i n g t h e parameters i n t e r p r e t e d by man as p r o s o d i c c h a r a c t e r i s t i c s of a s e n t e n c e . \" They base t h i s s e a r c h on the assumption t h a t man h a s o n l y a limited s h o r t term memory a v a i l a b l e f o r s t o r i n g the p i t c h c o n t o u r and s o makes d e c i s i o n s concerning the prosody of a s e n t e n c e by extracting p r o s o d i c f e a t u r e s which c o n t a i n c o n s i d e r a b l y less i n f o r m a t i o n than t h a t needed t o r e c o n s t r u c t e x a c t l y t h e same p i t c h c o n t o u r . They conclude from t h e s e experiments t h a t d e c i s i o n s such as d e c l a r a t i v e v e r s u s i n t e r r o g a t i v e are based on t h e p o s i t i o n of t h e rise o r f a l l i n p i t c h and n o t on the d i f f e r e n c e i n p i t c h from h i g h t o low. They a l s o conclude that i n d e t e r m i n i n g e m p h a s i s , the p o s i t i o n of t h e p e a k value of t h e second derivative of t h e p i t c h contour i s very significant. F. Brigham Young University (BYU) The research i n p i t c h c o n t o u r g e n e r a t i o n t o be d e s c r i b e d i n t h i s p a p e r addresses b a s i c a l l y t h e same q u e s t i o n s as t h e various p r o j e c t s surveyed above: (1) What t h e o r e t i c a l base might one use t o r e p r e s e n t s y n t a c t i k and semantic information? ( 2 ) How does one c o n v e r t l i n g u i s t i c i n f o r m a t i o n , both et sentencel e v e l and d i s c o u r s e -l e v e l , t o t h e a l g o r i t h m i c c o n t r o l of p r o s o d i c parameters? (3) What a s p e c t s of the p i t c h contour (e.g. 1st and 2nd d e r i v a t i v e s , t r a n s i t i o n s relative t o key syllables, and a c t u a l frequenty) are s i g n i f i c a n t i n causing i n t o n a t i o n and emphasis o p t i o n s t o be perceived? What s y n t h e s i s technique s h o u l d be used t o i n c o r p o r a t e t h e p r o s o d i c controls i n t o a working system (e.g, LPC s y n t h e s i s , formant s y n t h e s i s , o r a r t i c u l a t o r y s y n t h e s i s ) ?We have chosen to u s e Junction Grammar (JG) as a theoretical framework within which t o look f b L answers t o questions(1) and( 2 )absve. Junction Grammar r e f e r s t o a l i n g u i s t i c model formulated by L y t l e ( 1 9 7 4 ) -Subsequently, Junction Thesry has been used to formulate a new theory of phonology in wfiich a s e m a n t i c o -s y n t a c t i c r e p r e s e n t a t i o n (called a j u n c t i o n -t r e e ) i s recoded as a general a r t i c u l a t o r y represen-t a t i o n (called an articulation-tree) (Lytle, 1976). J u n c t i o n Grammar extended t o i n c l u d e J u n c t i o n Phonology was selected f o r u s e i n t h e BYU p r o j e c t because i t seems t o p r o v i d e some significant i n s i g h t s and a f l e x i b l e framework for our r e s e a r c h . It should be pointed o u t t h a t a t p r e s e n t t h e r e is no completely automatic a l g o r f t 5 m f o r o b t a i n i n g a d e t a i l e d and powerful r e p r e s e n t a t i o n of syntax-semantics from g e n e r a l E n g l i s h t e x t . For t h i s reason, o t h e r r e s e a r c h e r s (e.g., Allen a t MIT, Umeda a t B e l l Labs, and Levine a t Stanford) have chosen to use a simple representation which can be obtained a u t o m a t i c a l l y . The a u t h o r s ' r e s e a r c h , however, t a k e s advantage of a l a r g e r p r o j e c t ( L y t l e , 1975) which usesman-machine ---i n t e r a c t i o n t o o b t a i n a more powerful r e p r e s e n t a t i o n than can be obtained automatic a l l y . T h e r e f o r e , i t w a s decided t o u s e the f u l l power of J u n c t i o n Grammar r e p r e s -e n t a t i o n s i n hopes of a f u t u r e a u t o m a t i c a n a l y z e r r a t h e r than use some ' r e s t r i c t e d v e r s i o n of J u n c t i o n Grammar and be f o r c e d t o a d d t o i t piece by piece t o accoufit f o r more and more phenomena.", |
| "num": null |
| }, |
| "FIGREF7": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "T o gain i n s i g h t i n t o t o p i c (3) above (concerning which a s p e c t s of t h e p i t c h contour a r e s i g n i f i c a n t t~ p e r c e p t i o n ) , w e experimented w i t h manually s p e c i f i e d p i t c h c o n t o u r s .", |
| "num": null |
| }, |
| "FIGREF8": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "i n t u i t i v e terms and then i n terms of J u n c t i o n Grammar junction-trees ( J -t r e e s ) and articulation-trees (A-trees).'Rle s e c t i o n w i l l conclude with a block dLagram of what a fully developed J u n c t i o n Grammar t e x t synthests system would look l i k e and a block diagram of the system a s c u r r e n t l y iaplemented. A. I n t u i t i v e Presentation of Some Test Sentences v Consider the sefitence \"John d r o v e t o t h e s t o r e . This sentence can be read several d i f f e r e n t ways depending on t h e d i s c o u r s e context, shows f i v e p o s s i b l e readings and t h e i r c o n t e x t . Whatever system is used t o represent the l i n g u i s t i c s of this sentence, i t should be p o s s i b l e t o r e p r e s e n t each of these f o u r readings u n i q u e l y . S e n t e n c e P o s s i b l e c o n t e x t l a John drove t o the s t o r e .", |
| "num": null |
| }, |
| "FIGREF9": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "I f you receive as a reply s i m p l y \"yes\" t h e n t h e p e r s o n r e s p o n d i n g e i t h e r d i d n o t understand or is t r y i n g to b e f u n n y . ) On t h e o t h e r hand, s u p p o s e a whole crowd came to a p a r t y and you have a message which you must d e l t v e r to e i t h e r John o r Mary. In t h i s c o n t z x t , you may o r may not stress \"John\" and \"Mary\" b u t you would c e r t a i n l y end t h e s e n t e n c e w i t h a r i s i n g p i t c h . Then you would e x p e c t a ~e s / n o r e p l y , o r perhaps a y e s / n o w i t h a d d i t i o n a l v o l u n t e e r e d i n f o r m a t i o n s u c h as \"Yes, John is over t h e r e i n t h e corner . I 1 A g a i n , w e would like o u r system o f r e p r e s e n t a t i o n to h a n d l e this d i s t i n c t i o n . The two readings of \"Did John or Mary come?\" are summarized I n Figure 2. Sentence 2a D l d John or Mary come? (falling p i t c h a t end) 2b Did John or Mary come? P o s s i b l e Response John came, Yes, t h e y a r e both here.( r i s i n g p i t c h at end)", |
| "num": null |
| }, |
| "FIGREF10": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Did John or Mary come? F i n a l l y , consider t h e sentence \"The boys who s t u d y get good g r a d e s . \" Idhat difference i n meaning i s t h e r e 1n s t r e s s i n g \"study\" a s opposed t o stressiqg \"boys\"7 The d i f f e r e n c e can b e i l l u s t r a t e d b y", |
| "num": null |
| }, |
| "FIGREF11": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "s t u d e n t s i n c o n t r a s t w i t h boys. Once again, o u r system of representation needs t o handle this d i s t i n c t i o n , and handle i t i n a way c o n s t s t e n t with the t r e a t m e n t of o t h e r d i s t i n c t i o n s . Three r e a d i n g s o f t h i s sentetlce are summarized in Figure 3 .", |
| "num": null |
| }, |
| "FIGREF12": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Possible c o n t i n u a t i o n 3a The boys who squdy g e t good g r a d e s . . . as i s usually t h e c a s e . ( n e u t r a l ) 3b The boys who s t u d y g e t good g r a d e s , . . but t h e boys who spend a l l thei: time p l a y i n g b a s k e t b a l l get poor grades.", |
| "num": null |
| }, |
| "FIGREF13": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "The boys trho s t u d y get good g r a d e s . . . but f o r some r e a s o n the g i r l s (even the g i r l s who study) get poor grades.", |
| "num": null |
| }, |
| "FIGREF14": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "T h e boys whd study g e t good grades B. J u n c t i o n Grammar R e p r e s e n t a t i o n s o f the Same SentencesW e now d i s c u s s how J u n c t i o n Grammar represents t h e above d i s t i n c t i o n s i n its r e p r e s e n t a t i o n s . If the r e a d e r i s n o t as y e t familiar with J u n c t i o n Grammar, it might be a d v i s a b l e t o c o n s u l t Apperidix A b e f o r e roeading this s e c t i o n . A s i n d i c a t e d therein, some recent refinements of J u n c t i o n Grammar a r e not y e t a v a i l a b l e i n published form. W e therefore b r i e f l y discuss two of them here. One i s the s p e c L a l i z a t i o n s of subjunction i n J-trees, and t h e o t h e r i s t h e e x p l i c i t r e p r e s e n t a t i o n of m o d a l i z e r s . D i r e c t a n of Subjunction First c o n s i d e r t h e t h r e e major s p e c i a l i z a t i o n s of s u b j u n c t i o n shown i n F i g u r e 4 . S p e c i a l i z a t i o n s of DIRECTION: -r e s t r i c t i v e a s s o c i a t i o n induces a remainder induces no remainder F i g u r e 4 , S p e c i a l i z a t i o n s of Subjunction i n J-trees A r i g h t sub junc tion (* 0 ) of t e n signifies t h a t i n f o r m a t i o n i s t o b e e n t e r e d i n t o t h e h e a r e r ' s memory n e t . For example, when w e read t h e s e n t e n c e \"I saw a l o s t c h i l d w i t h a scraped knee t h i s morning, and 1 h e l p e d him f i n d his mother,\" w e enter (according t o J u n c t i o n t h e o r y ) i n t o o u r memory a s l o t f o r a c h i l d who w q s l o s t . The j u n c t i o n between \"a\" and \" c h i l d \" would b e N (\"a\") *-N (\"child\") , I f w e n e x t read t h e s e n t e n c e , \"The c h i l d had been crying f o r two h o u r s , the poor t h i n g , \" w e would recover the s l o t f o r the child and add to it the information that he had been c r y i n g . The j u n c t i o n between \"the\" and \" c h i l d \" i n this c a s e would b e N (\"the\") * * N (\"child\"). The t h i r d t y p e of s u b j u n c t i o n ( * * . )", |
| "num": null |
| }, |
| "FIGREF15": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "e.g. \"John, xho i s our mailman\"). I n a normal, restrictive m o d i f i c a t i o n , a l e f t subjunction i s used. For example i n , \"Please give T h e t h e yellow book on the second s h e l f , \" \"yellow'1 and \"book\" would b e joined a s follows (Pig. 5) . J-tree f o r \"yellow book\" For an explanation of the various nodes in this representation f o r a s i m p l e phrase see L y t l e (1975). I n t h e s e n t e n c e \"Of Tom, John and Rudolph, -John drove t o t h e s t o r e , ' ' t h e p r e p o s i t i o n a l p h r a s e \"of Tom, John and Rudolph\" does n o t r e s t r i c t the meaning of \"John\" i n the way \"yellow\" r e s t r i c t e d \"book\" i n the p r e v i o u s example. A c t u a l l y i n t h i s case, \"John\" restricts the scope of t h e p r e p o s i t i o n a l phrase. A s -a r e f l e c t i o n of this, the p r e p o s i t i o n a l phrase i s i n t w j o i n e d w i t h \"John\" using a s i g h t s u b j u n c t i o n as i l l u s t r a t e d i n Figure 6. u r e 6. R i g h t i n t e r j u n c t i o n W e c a l l t h i s an example of Frame I1 m o d i f i c q t i o n b e c a u s e t h e 1 I r i g h t s u b j u n c t i o n i s r e l a t i n g John\" t o a second frame of r e f e r e n c e ( i . e . Tom, John and Rudolph). On the o t h e r hand, \"yellow book'' i s a h a m e I m o d i f i c a t i o n b e c a u s e i t r e s t r i c t s \"book\" w i t h i n i t s own frame of r e f e r e n c e (i.e. i t d e t e r m i n e s whi-ch book w e a r e t a l k i n g a b o u t ) .", |
| "num": null |
| }, |
| "FIGREF16": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "i s an i n d i c a t i o n of remainder. The concept of remainder ( L y t l e , 1974) i s concerned w i t h whether all o r o n l y p a r t of a s e t i s r e f e r r e d t o . If one desires t o i n d i c a t e whether there i s a remainder i n a s u b j u n c t i o n , b e simply. r e p l a c e s t h e d o t w i t h e i t h e r a hyphen o r an e q u a l s s f g n .The Hyphen o p t i o n . For example, from t h e s e n t e n c e \"Please give f l m e t h e y e l l o w book on t h e second shelf, w e must assume t h a t t h e r e a r e books of some c o l o r o t h e r t h a n y e l l o w on t h e second s h e l f . These o t h e r c o l o r e d books are t h e remainder and we could diagram \"yellow book\" more s p e c i f i c a l l y t h a n before as f o l l o w s (F i g u r e 7Equals o p t i o n . One common c a s e of t h e e q u a l s o p t i b n i s f o r e x p l i c i t m o d a l i z e r s ( e . g . a r t i C l e s ) . For example, t h e phrase \"The c h i l d \" could b e diagrammed as follows ( F i g u r e 8), u r e 8 . E x p l i c i t modalizer .The i d e n t i t y of \"child\" i s r e t r i e v e d and placed i n t h e a r t i c l e \"the1', f i l l i n g i t e n t i r e l y and l e a v i n g no remainder. However, f o r our purposes, w e will l e a v e the modalizers i m p l i c i t and s i m p l y -u s e N (the) c a t . Thls b r i e f d i s c u s s i o n of s p e c i a l i z e d sub3unction and modalizers w i l l s u f f i c e f o r u s t o reexamine t h e t h r e e sample s e n t e n c e s p r e s e n t e d a t t h e beginning of the c h a p t e r , b u t t h i s time i n terms of J -t r e e s and A-trees.\"John drove t o t h e S t o r e . \" F i g u r e 9 shows t h e J-tree and A-tree f o r t h e n e u t r a l r e a d i n g of \"John drove t o t h e s t o r e \" ( s a n t e n c e l a ofF i g u r e 1). The J-tree (a semantico-syntactic r e p r e s e n t a t i o n ) i s c o n s i s t e n t w i t h t h e v e r s i o n of J u n c t i o n Grammar d e s c r i b e d by L y t l e (1975). The Atree ( a phonological r e p r e s e n t a t i o n ) i s e o n s i s t e n t w i t h J u n c t i o n Phonology ( L y t l e , 1976), except that the i n t e r n a l s t r u c t u r e of t h e V3 nodes i s n o t shown. This A-tree s p e c i f i e s t h a t t h e s e n t e n c e i s t o be pronounced i n t w o u n i t s \"John\" and \"drove t o t h e s t o r e 1 ' , and \"drove t o t h e s t o r e \" i s f u r t h e r d i v i d e d i n t o ' \"drove\" and \"to t h e s t o r e . I' The sub j u n c t i o n s numbered 1 and 2 i n d i c a t e the r e l a t i o n s between t h e s u b -p h~a s e s . I n a n a r t i c u l a t i o n tree, a l e f t s u b j u n c t i o n between H c o n s t i t u e n t s i n d i c a t e s t h a t t h e r i g h \"John drove t o t h e s t o r e \" V e r s i o n l a operand is p r~s o d i c a l l y subordinate t o the l e f t operand. As for the p i t c h confour, a l e f t subjunction causes a dbwnward p i t c h s h i f t . S i m i l a r l y , a r i g h t subjunction causes an upward shift. The extra subjunction a t the top of the A-tree i s available f o r a d d i n g prosodic feature specifications", |
| "num": null |
| }, |
| "FIGREF17": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "i b l e and a different A~t r e e could be used if i t were d e c i d e d to group t h e e l e m e n t s of t h e sentence d i f f e r e n t l y . At t h e bottom of Figure 9 is a simplified version of the A-tree, which i s used throughout the r e s t of this p a p e r t o make t h e trees easier to r e a d . But i t should be noted that t h e computer implementation uses the trees in t h e i r f u l l form. Having d e s c r i b e d t h e J-tree and A-tree for the neutral form of \"John drove t o t h e s t o r e , \" w e now consider how t h e trees d i f f e r for the four other versions shown i n Figure 1. I n v e r s i o n s b, c and d we stress \"John,\" \"drove\" and \"to the s t o r e t t r e s p e c t i v e l y . T h i s stregs is the r e f l e c t t o n of an implicit frame I1 modifier i n t h e J-tree ( s e eFigure 10). For example, according to Junction theory, when t h e context is \"lhodrove t o the s t o r e ? \" , \"~o h n \" i s i m p l i c i t l y modified by a right i n t e r j u n ction which indicates t h a t John has been s e l e c t e d out of a set of p o s s i b i l i t i e s . A possible explicit frame I1 modifier would be: \"Of t h e persons who ~i g h t have gone t o t h e s t o r e , John drove t o the s t o -e . A t this point, it is worrh Ciscussing a very general r e l a t i o n s h i p that has been observed between J-trees and English p r o s o d i c stress (Figure 11) :(1) I n a full s~b j u n c t i o n , any t i m e a remainder i s induced (i.e. by *o r -*) i n an operand, the other operand receives a stress (e.g=two *boys).(Continued on page 2 3 . )", |
| "num": null |
| }, |
| "FIGREF18": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "d i n a t e part of t h e i n t e r j u n c t i o n t o which t h e t o p i c i s joined as an emclitic.", |
| "num": null |
| }, |
| "FIGREF19": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "J-trees and English prosodic stress", |
| "num": null |
| }, |
| "FIGREF20": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Thus we have accounted f o r t h e th'ree stressed v e r s i o n s of \"John drove t o the s t o r e . \" The i n t e r r o g a t i v e v e r s i o n ( v e r s i o n le of Figure 1 ) has a [+ verify] f e a t u r e on t h e top of t h e J-tree. That i s , t h e l i s t e n e r i s asking f o r v e r i f i c a t i o n of what w a s s a i d . This f e a t u r e i s retofded as a p r o s o d i c [+ v e r i f y ] f e a t u r e i n t h e A-tree. F i g u r e 1 2 shows t h e k-trees f o r these f i v e versions. Having covered this first example i n d e t a i l , l e t u s examine t h e two other sample sentences in a more abbreviated f a s h i o n . id John o r Mary Come?\" Figure 13 shows t h e J-tree and A-tree f o r each v e r s i o n of \"Did John o r Mary come?\". A s seen i n t h e s e f i g u r e s , t h e semantico-syntactic d i f f e r e n c e between t h e two v e r s i o n s i s where the i n t e r r o g a t i v e i s placed, on the whole sentence o r on the conjoined s u b j e c t . The prosodic d i f f e r e n c e i s that i n version 2a, \"John\" and 'Wary\" are stressed ( s t i m u l a t e d by the interrogation on t h e OR j u n c t i o n ) , w h i l eJohn -Figure 12. \"John drove t o the store\" Versions l bl e", |
| "num": null |
| }, |
| "FIGREF21": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "\"Did John ar Mary come?\" i n v e r s i o n Zb, the A-tree i s marked [unfinished] because of t h e [yesno i n t e r r o g a t i v e ] f e a t u r e on t h e J -t r e e . A \" f i n i s h e d \" v e r s i o n would be \"Did John o r Mary come o r not?\". \"The'boys who study.\" F i g u r e 1 4 shows J -t r e e s and At-trees f o r 1t t h e t h r e e v e r s i o n s of \"The boys who s t u d y g e t good grades. The J -t r e e s d i f f e r o n l y i n t h e t y p e of s u b j u n c t i o n #between \"boys\" and \"who\". In t h e A-tree, \"boys\" o r \"who study\" i s stressed according t o t h e type of s u b j u n c t i o n i n t h e J -t r e e , f o l l o w i n g the r u l e s t a t e d above. T h i s conc l u d e s our d i s c u s s i o n of how J u n c t i o n Grammar h a n d l e s t h e t h r e e samp,le s e n t e n c e s p r e s e n t e d a t the beginning of t h e s e c t i o n . C. Text S y n t h e s i s Yodel W e now c o n s i d e r a fully-developed JunctPon Grammar t e x t s y n t h e s i s system (Figure 15). This system i n c o r p o r a t e s t h e J u~c t i o n Grammar model of t r a n s l a t i o n s o t h a t t h e i n p u t t e x t might be i n Spanish and t h e o u t p u t i n English. I n t h i s f u l l system, J-trees a d j u s t e d ( t r a n s f e r e d ) f o r the t a r g e t language vould be needed as w e l l as f u l l y s p e c i f i e d A-trees. The A-trees would i n c l u d e t h e i n t e r n a l s t r u c t u r e of t h e V 3 nodes, and t h e information i n t h e A-tree would be converted i n t o parameters t h a t d r i v e a f u n c t i o n a l analog of t h e v o c a l cords and t r a c t . C l e a r l y , p u t t i n g t o g e t h e r such a system would b e a v e r y ambitious p r o j e c t . A r e s t r i c t e d v e r s i o n . A t p k e s e n t , we have implemented only a r e s t r f c t e d v e r s i o n of t h e f 6 l l system, i l l u s t r a t e d i n Figure 16. I n t h i s system we have i s o l a t e d t h e p i t c h contour from q t h e r c o n t r o l parameters.", |
| "num": null |
| }, |
| "FIGREF22": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "g i t * , ' A Figure 14. \"The boys who study get good grader.\" The boys .* who study b The boys *. who study u s t e d J -t r e e + I Junctian Grammar S y n t h e s i s I Junction Grammar Transfer 2 A-iree & g e n e r a l a r t i c u l a t o r y ) A fully-developed Junction Grammar T e x t S y n t h e s i s System The currently implemented system. t we LPC-analyze the spoken input sentence, enter a J-tree for t b sentence, recode the J-tree as an A-tree, generate a pitch contour from the A-tree, replace the natural p i t c h contour with the generated one, and PC-synthesize t o prqduce a spokkn output sentence. 11. METHOD The model described in Section I provides a representation f o r the semantico-syntactic information underlying prosodic contrasts and a very f l e x i b l e framework for representing phrasing and prosodic features at the general articulatory l e v e l . But we have not y e t s p e d f i e d how a J-tree i s recoded as an A-tree or how the p i t c h ,contour is actually obtained from the A-tree. This chapter will describe the computer algorithms that have been implemented td perform these two conversions.Of course, they should not be taken as any kind of f i m l statement concerning the task as they are under continuing development.A. Recoding a J-tree as an A-treeThe general form of t h e A-tree is obtained by traversing the J-tree according t o the language specific order stored i n the J-tree. A t each node the algorithm decides whether or not to declare a phrase, thus allowing nested phrases. The criteria for declaring a phrase are:", |
| "num": null |
| }, |
| "FIGREF23": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Each operand of a conjunction forms a phrase.The assignment of prosodic features t o the A-tree (f .e. [+ stress] , [+ u n~i n i s h e d phrase] , and [+ verify contour] ) i s fairly strsightf orward.", |
| "num": null |
| }, |
| "FIGREF24": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Tne d i r e c t i o n a l i t y of t h e s u b j u n c t i o n s between n -c o n s t i t u e n t si n t h e A-tree Is left except i n t h e following s i t u a t i o n s : There i s a r i g h t s u b j u n c t i o n between t h e A-tree phrases from a simple verb and i t s complex o b j e c t i n the J-tree,(2) I f a phrase i s marked [+ s t r e s s ] , t h e sub-phrases of the phrase are subordinated t o i t by a d j u s t i n g t h e d i r e e t i o n a l i t i e s of the sub j u n o t i o n s .B. Background of t h e A-tree t o P i t c h Contour AlgorithmWith t h i s overview of t h e J -t r e e t o A-tree conversion a l g o r i t h m ,we d e s c r i b e an algorithm t o o b t a i n a p i t c h contour from an A-tree. The evolutionary phases i n the development of t h i s algorithm were: Plots. W e p l o t t e d p i t c h ahd i n t e n s i t y a g a i n s t time f o r various r e a d i n g s of s e v e r a l sentences. Manual Contours. I n order t o determine which a s p e c t s of t h e p i t c h contour are essential t o natural-sounding synthesis, we programmed a system t o allow manual s p e c i f i c a t i o n of t h e p i teh contour with linear i n t e r p o l a t i o n between s p e c i f i e d p o i n t s and t o then p e n n i t l i s t e n i n g comparison of s y n t h e s i s o u t p u t s with n a t u r a l versus manual contours.F i r s t Algorithm. Based QXI these i n i t i a l experiments, we programmed a simple p i t c h contour a l g o r i t h m that imposed on each p h r a s e a contour s e l e c t e d from a f f x e d i n v e n t o r y of c o n t o u r s and algebraically added in a pitch \"bubble\" t o t h e syllable o f a p r o s o d i c a l l y s t r e s s e d V3. In this i n i t i a l system we were a b l e t o c r e a t e m u l t i p l e readings of s e n t e n c e s l i k e \"John drove 'to the store\" from a s i n g l e set of LPC analysis parameters, varying only the p i t c h contour. I n other words, w e concluded t h a t although t h e perceptual phenomenon c a l l e d prosodic o r suprasegmental stress i s well-known t o be based on several a c o u s t i c parameters, including p i t c h (i.e. fundamental frequency), i n t e n s i t y and d u r a t i o n , i n st l e a s t some c a s e s , changing o n l y t h e p i t c h c o n t o u r is s u f f i c i e n t t o cause a word t o b e perceived as s t r e s s e d or not s t r e s s e d . However, a f t e r considerable t h e o r e t i c a l d i s c u s s i o n s , we decided t o abandon the approach of u s i n g a fixed i n v e n t o r y of p r o t o t y p e c o n t o u r s and try a more dynamic approach, which we w i l l now d e s c r i b e . C. Current A-tree t o P i t c h Contour Algorithm Given an A-tree and an o p t i o n code t o i n d i c a t e i n i t i a l and f i n a l v a l u e s and bounds on p a r a m e t e r s , the a l g o r i t h m a s s i g n s an i n i t i a l and f i n a l p i t c h basea on the o p t i o n code. Then t h e A-tree i s t r a v e r s e d i n l e f t -r i g h t order. Upon e n c o u n t e r i n g e a c h V 3 , we a s s i g n a p i t c h t o t h e c o r e of its nuczear syllable as follows: The f i x s t -V 3 ' r e c e i v e s the i n i t i a l p i t c h of the sentence. (2) A left s u b j u n c t i o n c a u s e s a r a t i o decrement (about 0.90) t o the l a s t assigned pitch.(3) A right s u b j u n c t i o n causes a r a t i o increment (about 1,12) in r e l a t i o n t o the l a s t assYgned pifch.", |
| "num": null |
| }, |
| "FIGREF25": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "A conjunction causes no change t o d a t e , b u t f u r t h e r r e s e a r c h i b needed. An B -c o n s t i t u e n t domxpstlng m u l t i p l e V3's r e k e i v e s the average of t h e most r e c e n t l y a s s i g n e d p i t c h l e v e l and the highest p i t c h a s s i g n e d t o any of i t s operands. Then t h e contiaurs between n u c l e a r s y l l a b l e s a r e defined a s v a l l e y s whose depth i n c r e a s e s w i t h t h e d i s t a n c e i n time between t h e n u c l e a r s y l l a b l e s i t j o i n s . A f t e r t h e i n i t i a l contour i s d e f i n e d , twc. t y p e s of contour adjhstments a r e added: (1) ~djustments i n t h e p i t c h contbur caused by s t o p consonants. W e c a l l t h e s e s t o p d i s c o n t A n u i t i e s because when the speech,waveform becomes voiced a g a i n a f t e r a s t o p , t h e p i t c h i s s i g n i f i c a n t l y h i g h e r than when the s t o p began b u t soon s e t t l e s down t o a v a l u e which would b e p r e d i c t e d by smooth i n t e r p o l a t i o n ctf the p i t c h contour over the unvoiced segment. The p i t c h \"bubble\" a s s o c i a t e d w i t h a s t r e s s e d V 3 . Although t h e above a l g o r i t h m i s n o t complete, i t works reasonably w e l l and does have one a l r e a d y mentioned a s p e c t which w e r e p e a t h e r e f o r emphasis : T h e , p i t c h contour i s generated from t h e A-tree i n a completely dynamic manner. That i s , t h e r e i s no f i x e d inventory of p i t c h l e v e l s o r phrase contours. Each new p i t c h l e v e l i s assigned r e l a t i v e t o previous v a l u e s assigned and i n accordance w i t h preassigned a b s o l u t e p i t c h limits ( e g g . 60 Hz'and 200 Hz) and t h e o v e r a l l s t r u c t u r e of the A-tree. This means t h a t , although we have s o f a r r e s t r i c t e d o u r s e l v e s t o c a r e f u l l y spoken speech, t h i s system may have the f l e x i b i l i t y t o e v e n t u a l l y allow s y n t h e s i s of v a r y i n g speech rates, i.e. very slow and c a r e f u l o r v e r y fast and sloppy speech by a p p r o p r i a t e o p t i o n codes i n t h e 35 J-tree to A-tree algorithm and the A-tree to pitch contour algorithm.", |
| "num": null |
| }, |
| "FIGREF26": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "With unvoiced segments filled in f o r easier comparison with rule-generated contours", |
| "num": null |
| }, |
| "FIGREF27": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Natural contour for sentence 3b (\"The boys who study g e t good grades\") Natural Pf tch Contour Rule-generated P i t c h Contour Figure 17 b. Natural and rule-generated contours for sentence 3b Natural and Manual Contours f o r sentence 3b boys Natural Figure 18a. Natural and rule-generated contours for sentence 3c(\"The boys who study get good grades.\") u r e 18b. Natural and rule-generated contours for the sentence \"The c a t t h a t the dog chased got away.\"111. EVALUATION AFD DISCUSSIONW e produced a demonstration t a p e of LPC s y n t h e s i z e d speech using n a t u r a l , monotone, and rule-generated p i t c h contours.Figure 19shows the c o n t e n t s o f the tape. Various s u b j e c t s s a i d that although t h e s e n t e n c e s w i t h rule-generated p i t c h contours did n o t sound as n a t u r a l as t h e n a t u r a l v e r s i o n s , they could c l e a r l y p e r c e i v e t h e same d i s t i n c t i o n s i n the r u l e v e r s i o n s +is were made i n t h e n a t u r a l versfons. Thus w e e s t a b l i s h e d two c r i t e r i a of evaluation: n a t u r a l n e S s of i n t o n a t i o n , and \" i n t e l l i g i b i l i t y \" of i n t o n a t i o n , by which w e mean a human l i s t e n e r can c o r r e c t l y p e r c e i v e which r e a d i n g of a multiple-reading sentence as intended. A. Format of t h e T e s t I n o r d e r t o o b t a i n a q u a n t i t a t i v e e v a l u a t i o n of the system, we devised the following four part t e s t , which was presented t o 17 s u b j e c t s . The s e n t e n c e s i n t h e test c o n s i s t e d of 35 v e r s i o n s made from a dozen sets of T 2 C a n a l y s i s parameters by imposing v a r i o u s n a t u r a l , manual, monotone, and rule-generated p i t c h contours on them. I n t h e f i r s t p a r t listeners were asked t o r a t e r e a d i n g s of 34 s e n t e n c e s on a s c a l e from 1 t o 5, where I1 tr meant the i n t o n a t i o n sounded mechanical o r monotone, and 5 meant the i n t~n a t i o n sounded n a t u r a l . I n t h e secorid p a r t , l i s t e n e t s were presented with 24 s e n t e n c e pairs and asked t o indicate whether the first or second sentence s~u n d e d more n a t u r a l . The t h i r d and fourth p a r t s of t h e t e s t d e a l t w i t h i n t e l l i g i b i l i t y of intonation. I n both of theski parts, the s u b j e c t s heard a sentence and i n d i c a t e d which of s e v e r a l p e s s i b l e readings the i n t o n a t i o n bas-intended to convey. The only d i f f e r e n c e between t h e s e l a s t two parts was the method of d e s i g n a t i n g t h e d i f f e r e n t readings. I n the t h i r d p a r t , r h e NATURAL vs . GENERATED INTONAT I O N N a t u r a l I n t o n a t i o n Generated I n t o n a t i o n 1'. John drove to the store. 2. John drove to the store, (monotone)3. John drove to the s t o r e . John drove to the store.", |
| "num": null |
| }, |
| "FIGREF29": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "", |
| "num": null |
| }, |
| "FIGREF30": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "John buys r i c e ? Contents of Preliminary Test Tape readings were designated by underlining and using a perfad or question mark at the end. In the fourth part, t;he readings were d e s i g n a t e d by an i n d i c a t i o n of a t y p i c a l c o n t e x t f o r that reading. [Appendix D c w t a i n s a d d i t i o n a l d e t a i l s of the test and t h e r e s u l t s ) .", |
| "num": null |
| }, |
| "FIGREF31": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "p a i r e d t-test a p p l i e d t o the a v e r a g e scores for n a t u r a l and r u l e contours f o r each l i s t e n e r showed a s t a t i s t i c a l l y s i g n i f i c a n t o v e r a l l p r e f e r e n c e f o r n a t u r a l contours.I n part 2 , i n a balanced s u b s e t of 4 2 p a i r e d comparisons where n a t u r a l , manual and r u l e v e r s i o n s were p a i r e d i n a l l p o s s i b l e ways, the n a t u r a l contours received 87 v o t e s , the manual ones received 76 and the r u l e contours received 41. Several s u b j e c t s mentioned after t h e test t h z t t h e n a t u r a l , hand and rule versfons og the second sentence, he he cat that the dog cnased got may\") were i n d i s t i n g u i s h a b l e ib n a t u r a l n e s s of i n t o n a t i o n . Usin& a non-parametrfc sign test t e c h n i q u e , we p o s t u l a t e d t h a t i f t h e r e were a s i g n i f i c a n t p r e f e r e n c e for one pitch contour method over a n o t h e r , t h e l i s t e n e r s would b e c o n s i s t e n t i n their choice, r e g a r d l e s s of t h e order of p r e s e n t a t i o n . S p e c i f i c a l l y , i f f o u r o r fewer s u b j e c t s out of 17 changed t h e i r minds, w e can conclude a preference for a given p a i r and i t s r e v e r s e . Using t h i s c r i t e r i o n , w e found t h a t f o r t h e f i r s t s e n t e n c e , t h e n a t u r a l version was s i g n i f i c a n t l y preferred but f o r t h e second s e n t e n c e , t h e r e was no c l e a r p r e f e r e n c e f o r t h e n a t u r a l over the rule v e r s i o n .I n p a r t s 3 and 4 , we t a b t e d f o r \" i n t e l l i g i b i l i t y \" of i n t o n a t i o n by p r e s e n t i n g s e n t e n c e s and a s k i n g which of s e v e r a l p o s s i b l e r e a d i n g s was intended. W e e v a l u a t e d t h e r e s u l t s of t h i s part by p r e p a r i n g conf u s i o n m a t r i c e s . (Figure 20.) Each one d e a l s w i t h readiilgs of a s i n g l e s e n t e n c e , showing r e a d i n g t r a n s m i t t e d and p i t c h contour method (N=natural, R=rule) compared t o reading received by the l i s t e n e r s . All r e a d i n g s are l i s t e d i n Appendix D. A simple Chi-Square test shows t h a t f o r a given row of one of these confusion m a t r i c e s , 24 c o r r e c t v o t e s o u t of 33 o r 34 are s u f f i c i e n t t o show s i g p i f i c a n c e a t t h e .05 level. R e s u l t s f o r p a r t 4 were s i m i l a r .", |
| "num": null |
| }, |
| "FIGREF32": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "r iAnother problem s e n t e n c e was \"Did John o r Mary come?\" Although t h e two Confusion matrices for part 3 rule version6 were clearly distinguishable (one with falling and one with riaing terminal intonation), the listeners made many i n c o r r e c t choices. This may have bean due t o e i t h e r of the following two f a c t o r s : As with the other sentences, a l l t h e rule v e r s i o n s were based on a single s e t of a n a l y s i s parameters, and d u r a t i o n was held c o n s t a n t .I n this sentence, d u r a t i o n plays a g r e a t e r r o l e than i n o t h e r s , and t h i s may have influenced judgment. (2) There may have been some c~n f u s i o n about what the v e r s i o n s meant, and there may have been confusion w i t h ie p o s s i b l e third r e a d i n g i n which \"Johnt1 and \"~a r y \" a r e stressed and y e t the i n t o n a t i o n i s r i s i n g a t the end.D m Termination ProblemsAnother problem mentioned by s e v e r a l s u b j e c t s wzs t h a t the i n t o n a t i o n on some version8 ( r u l e and hahd versions only) was n a t u r a l up until t h e very end of t h e sentence. Re have determined t h a t this i s a problem i n s h a p i n g + t h e contour from t h e l a s t nuclear syllable t o t h e f i n a l p i t c h of t h e sentence, a s s i g n i n g an a p p r o p r i a t e fYna1 p i t c h , and determining the i n t e r a c t i o n between the p i t c h of the l a s t n u c l e a r s y l l a b l e and the sentence f i n a l p i t c h . Further research-is needed i n thls a r e a .C. DiscussionThis paper is the r e p o r t of an attempt to g e n e r a t e e p i t c h contours i n speech s3~1tbes3.s using J u n c t i o n Grammar as a t h e o r e t i c a l base. Since t h e v a r i o u s r e a d i n g s of each sentence were ma& by imposing different p i t c h contours on the same analysis parameters without changing durations, some versions were less than natural. However, this was t o be expected and we feel that it was even desirable in that it pointed out some specific cases in which durationedjustments are necessary.The evaluation a l s o pointedout t h e need f o r f u r t h e r research on the shaping of t h e contour from t h e l a s t V 3 t o the efid of t h e sentence. We also realize t h e need t o i n c o r p o r a t e some refinements i n t o the system i n order t o (1) make degrees of adjustment f o r f r i c a t i v e s and s t o p s , (2) improve the n a t u r a l n e s s of t h e contours between nuclear s y l l a b l e s , (3) make adjustments for t h e inherent pitch of $owel$ (Flanagan and Landgraf , 1968) . Based on the results of the e v a l u a t i o n t e s t , w e f e e l i t i s appropriate to continue use of the Junction Grammar framework and t o attempt t o develop a word concatenation v e r s i o n w i t h d u r a t i o n , pause and i n t e n s i t ? c a l c u l a t i o n s , t o attempt better shaping of the contour a f t e r the l a s t nudlear syllable, and t o examine many more sentence types in order t o f u r t h e r t e s t t h e adequacy of t h i s framework f o r dealing with t h e problem of generating prosodic c o n t r o l parameters i n speech s y n t h e s i s . ACKNOWLEDMENT S The author would like t o express deep a p p r e c i a t i o n t o co-author Eldon Lytle f o r many t h e o r e t i c a l discussions, t o W.J. Strong f o r s h a r i n g h i s acoustical expertise and LPC analysis-synthesis programs, and to Ronald Millet t f o r his excellent suggestions i n our innumerable d'iscussions during t h i s research and for doing the FORTRAN coding of the J-tree input, d i s p l a y mechanism, the conversion algorithms from J-tree to A-tree, and from A-tree to p i t c h contour. PITCH In t h i s paper p i t c h c o n t o u r is used to mean fundamental frequency contour PROSODICS There are w6xd-boundary e f f e c t , phrase-level stress contours, and clause-level phenomena which affect the waveform. These factors are referred t o as the suprasegmental or prosodic features of speech. t o code t o speech-waveform. v3 A s y l l a b l e . See L y t l e (1976) for a more precise definition.", |
| "num": null |
| }, |
| "FIGREF33": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "o n system d e s c r i b e d i n this paper has been implemented on a PDP-15 computer, equipped w i t h s variety of p e r i p h e r a l devices c o n f i g u t e d a s shown-in Figure 21. The VT-15 allow6 t h e user t o c a l l a package of s u b r o u t i n e s from FORT& t o p l o t p o i n t s o r d r a w lines o r c h a r a c t e r s . The system uges the DEC s u p p l i e d DOS-15 o p e r a t i n g s y s t e m . The PDP-15 i s equipped w i t h 32 K 1 8 -b i t words. This i s n o t enough memory f o r our m q i n p i t c h contour g e n e r a t i o n program s o w e use t h e DOS-15 CHAIN AND EXECUTE f a c i l i t y t o overlay programs t h a t need n o t be core r e s i a e n t s h u l t a n e o u s l y . A s i n d i c a t e d i n F i g u r e 21, t h e r e a r e two d i s k d r i v e s on t h e system. One i s a s t a n d a r d DOS-15 system pack f o r system programs and user f t l e s . The o t h e r d r i v e i s mainly f o r speech d a t a . Data on packs nounted bn this d r i v e i s accessed through s p e c i a l assembler s u b r o u t i n e s t h a t a r e not p a r t of the DOS-15 o p e r a t i n g system, This a l l o w s t h e user t o s t o r e data c o n t i g u o u s l y a t a higher transfer r a t e than p o s s i b l e u s i n g standard DOS-15 f i l e s . This i s especialA7 important i n t r a n s f e r r i n g large amounts of d a t a from t h e A/D t o d i s k o r from t h e d i s k t o t h e D/A i n r e a l t i m e . Thus t h e system can d e a l w i t h l o n g e r segments of speech t h a n can b e s t o r e d i n in-core b u f f e r s a t ohe time. In o r d e r t o d e s c r i b e the p i t c h contour system, w e w i l l d e s c r i b e t h e major d a t a f i l e s and o f f -l i n e s u p p o r t programs t h e system r e q u i r e s , For each s e n t e n c e to be processed, the system needs (1) an e n t r y i n a speech d 9 r e c t o r y f i l e (SPCDTR) which i n d i c a t e s t h e a d d r e s s a c~d l e n g t h on the speech data disk of t h e LPC a n a l y s i s parameters. (2) An i d e n t i f i c a t i o n file which specifies the word b o u n d a r i e s , e t c . and t h e file names of the J-tree f i l e s for the various r e a d i n g s of t h e s e n t e n c e . The J-tree c o n t a i n s kejls t o o b t a i n l e x i c a l i n f o r m a t i o n a b o u t each word from a m a s t e r l e x i c o n f i l e . (3) A J-tree f i l e f o r each r e a d i n g . I n o r d e r t o p r e p a r e a s e n t e n c e f o r p r o c e s s i n g , i t i s t a p e r e c o r d e d , t h e n digitized a t a lOKHZ sampling r a t e u s i n g a program c a l l e d DIGTXZ. Then i t i s LPC analyzed and o p t i o n a l l y examined Qn t h e g r a p h i c s d i s p l a y , u s i n g a program c a l l e d ANAPLT. The \"PLT\" a t the end of the name r e f e r s t o t h e f a c t t h a t this program w i l l a l s o produce a h a r d copy p l o t of t h e p i t c h contour if d e s i r e d . The p i t c h c o n t a u r g e n e r a t i o n program i s c a l l e d JTSPCH (\"J-Tree t o speech\"). When this program i s e x e c u t e d , i t p r e s e n t s a list of a v a i l a b l e s e n t e n c e s and asks t h e u s e r t o i n d i c a t e which read5ng t o u s e i n t h i s case. Then t h e program r e a d s the J -t r e e f i l e and c r e a t e s a J-tree i n p o s t f u r n o t a t i o n . T h e program then o p t i o n a l l y d i s p l a y s the J-tree on t h e g r a p h i c s u n i t , depending on t h e s t a t u s of the console sense s w i t c h e s . Then t h e J-tree is converted t o an A-tree, which a g a i n i s o p t i o n a l l y d i s p l a y e d . Then a p f t c h c o n t o u r i s g e n e r a t e d from the A-tree and d i s p l a y e d . F i n a l l y , t h e p i t c h c o n t o u r i s combined w i t h t h e LPC a n a l y s i s parameters r e t r i e v e d from disk ( g a i n f a c t o r , voiced/uniroiced decision and 1 2 linear p r e d i c t o r c o e f f i c i e n t s p e r 10 msec of speech waveform) and t h e c o n t a i n e d parameters are used t o s y n t h e s i z e a speech wayeform which i s s t o r e d on a temporary disk a r e a and repeatedly played through t h e D/A c o n v e r t e r t o a loudspeaker o r headphones for e v a l u a t i o n . If d e s i r e d , the u s e r can t h e n s a v e i t permanently on disk. Another p r o c e s s i n q o p t i o n i s t o create a manual p i t c h c o n t o u r i n s t e a d of gene r a t i n g i t from an A-tkee. The manual contour can b e catered e i t h e r by drawing i t on the graphics u n i t with the Mght pen or by entering a list of t i m e and pitch c o o r d i n a t e s on thsr teletype t o a subroutine that i n t e t p o l a t e a l i n e a r l y b e t w e e n them. Of course, the sentence can a l s o be synthesized using the natural pitch contour retrieved from the original a n a l y s i s data. After sav2ng s e v e r a l syntehsized sentences, one can listen t o a l i s t of sentences w i t h any dr s i r e d pause between them u s i n g a multiple 1 4 6 t e n i n g p r b g r h c a l l e d MULTIL. MULTIL can receivv i t s control input from either the t e l e t y p e or from a data file. This option allowed us t o create a c o n t r o l f i l e with the regular e d i t i n g f a c i l i t i e s of the operating system and then i n y t r u c t MULTIL t o read i t , creating the evaluation test t a p e i n one continuous recording s e s s i o n without any t%pe s p l i c i n g . APPENDIX D MORT3 DETAILS ON THE EXALUATION This appendix c o n t a i n s the following information: An e d i t e d v e r s i o n of the e v a l u a t i s n response form given t o the subjects and t h e n f o u r t a b l e s showing a l l responses. Note t h a t the p a r t s of the response form are numbered IA, I B , IIA and I I B . T h i s e d i t e d response form shows which v e r s i o n s were used throughout the t e s t b u t does n o t c o n t a i n c e r t a i n unnecessary d e t a i l s p r e s e n t i n t h e a c t u a l response form used. Each version i s i ' d e n t i f i e d by a code c o n s i s t i n g of a n w b e r (1-8), a l e t t e r (a-e) , a letter (N, R, M o r H) and p o s s i b l y another number (1-4). The f i r s t two c h a r a c t e r s i d e n t i f y the sentence and reading; as follows: (1) a. John drove t o t h e s t o r e . b. John drove t o the s t o r e . c. John drove t o t h e s t o r e . d. John drove t o t h e s t o r e . e. John drove t o t h e s t o r e ?(2) a. Did John o r Mary come? (fairing a t end)b. LlLd John or Mary come? ( r i s i n g a t end).", |
| "num": null |
| }, |
| "FIGREF34": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "a. John, 30e ahd Fred bliy riae. (7) a -The cat that the dog chased got away. Joha buys rib? The neqt character ideneif ies the naturc! of the p i t c h contour as follows: N = Natural R -Rule (ganerated by rule). M = Yonotone Cconstant fundamental frequency) H = Hapd (manually specpfied)If a number follows the R Lt %ad t c a t e s which hand made contour w a s used.", |
| "num": null |
| }, |
| "FIGREF35": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "i t , The rating s c a l e i s 1 t o 5. Remember t h a t the e v a l u a t i o n c r i t e r i o n i s i n t o n a t i o n only. So please do no9t l e t your judgements be in\u00a3 qugqced by c r a c k l e s o r pops o r h i s s e s . A rating a\u00a3 1 means t h e i n t o n a t i o n sounded mechanical o r unnatural, for example, monotone o r the way computers talk i n cartoons. A r a t i n g of 5 means the intonazion sounded natural, t h a t i s , you can imagine the sentence was produced by a human speaker speaking c a r e f u l l y . Please t r y t o dis'tribute your s c o r e s over the e n t i r e range from 1 t o 5 . Before you b e g i n , p l e a s e read over t h e entire test t o become familiar w i t h i t , because you w i l l have only a f e w seconds to respond t o each question. The test w i l l l a s t 17 minutes. (The f o l l o w i n g fouf pages are an edited, abbrevdated form of the rest of t h e response s h e e t s . The codes i n parentheses were n o t on t h e a c t u a l response sheets. By c o n s u l t i n g t h e key on the previous pages of I A .", |
| "num": null |
| }, |
| "FIGREF36": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "The c a t t h a t t h e d o gc h a s e d g o t a w a y .", |
| "num": null |
| }, |
| "FIGREF37": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "J o h n d r o v e t o t h e s t o r e . S E C O N D TIME THROUGH: R a t e e a c h s e n t e I have o n e . . . . . . . . . . . . . . .", |
| "num": null |
| }, |
| "FIGREF38": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "J o h n d r o v e t o t h e s t o r e . . . . . . . e natural n a t u r a l J J . . . . . . . .", |
| "num": null |
| }, |
| "FIGREF39": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Did John o r M a r y c o m e ? . J o B n or M a r y come?. . . . . . . (2aH1) (2aH2)", |
| "num": null |
| }, |
| "FIGREF40": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "s p o n s e t o : \"How d i d J o h n g e t t o t h e s t o r e ? \" d. In r e s p o n s e to: \" W h e r e d i d J o h n d r i v e ? \" e. To ask f o r v e r i f i c a t i o n of what w p o n s e m e a n s t h e s u b j e c t l e f t t h a t q u e s t i o n b l a n k . T a b l e D -2 R e s p o n s e s f o r p a r t IB. \"I\" m e a n s t h e s u b j e c t c h o s e t h e f i r s t e l e m e n t o f a p a i r ; \" 2 \" m e a n s t h e s e c o n d e subject chose version a .", |
| "num": null |
| }, |
| "FIGREF41": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "e n t a t i o n MacNeflage h a s pointed up t h e d i f f i c u l t y of m e d i a t i n g between a b s t r a c t u n i t a r y phonological r e p r e s e n t a t i o n s and t h e continuous nature of the dynamic speech chain, suggesting t h a t u n i t a r y phonological represerrtations are analogous t o a sequence of eggs conveyed t o t h e wringer of a washing machine, while t h e scrambled mess t h a t emerges f r o 9 t h e wringer i s what must a c t u a l l y be d e a l t with by t h o s e engaged i n computer a n a l p s i s and 1 s y n t h e s i s of voice.", |
| "num": null |
| }, |
| "FIGREF42": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "s incapable of d i s c r e t e u n i t s of o u t p u t , what i s t h e n a t u r e of t h e transforma i o n , a t t h e p e r i p h e r a l staget, of one form t o t h e o t h e r . 5 Lieberman l i k e w i s e n o t e s a r e l a t i v e n e g l e c t of t h e p h o n e t i c l e v e l of speech, concldding that a quantitative and expl$c$t phonetic t h e o r y has y e t t o be developed, and suggesting t h a t a s u c c e s s f u l attempt t a ' c o n s t r u c t such a theory should be s t r u c t u r e d i n terms of the a a a t o m f z , physiologic, and n e u r a l mechanisms of speech producrion and p e r c e p t i o n .Onn, similarly motivated by the n o t i o n that speech .ought t o be described in the c o n t e x t of the organic mechanisms responsible f o r it, supgests, t h a t : It may, be argued that an abstract r e p r e s e n t a t i o n may be regarded as i n s t r u c t i o n s for p a r t i c u l a r types of behavior of the kpeech-generating mechanism. When t h e s e i n s t r u c t i o n s are carried out, t h e v a r i o u s reactions occurring between a f f e r e n t p h y s i o l o g i c a l structures w i l l yield 4 quasicontinuous gesture i n whieh t h e discrete l n s t r u c t i o n s i n i t i a t i n g the gesture a r e no longer always observable as d i s t i n c t comporlents. F i n a l l y , the exe u t i o n of these i n s t r u c t i o n s produces t h e a c o u s t i c s i g n a l . E The p~irpose of t h e present paper is t o o u t l i n e briefly a new system of p h o n o l o~c a l d e s c r i p t i o n c u m e n t l y being used a s a basis f o r voice s y n t h e s i s a t BYU which attempts t o s a t i s f y the c r i t e r i a suggested by ITacNeilage, Lieberman, and Onn r e f e r e m e d above. The d e s c r i p t i v e system i n question i s based on t h e J u n c t i o n Gramar Model of language developed by myself and my colleagues over the past e i g h t years.5 It is a model specifically s t r u c t u r e d i n terms of speech-related organs, e i t h e r a s they are known oi hypothesized, An Overview of t h e Junction Grammar Model A fundamental t e n e t of j u n c t i o n theory i s t h a t l i n g u i s t i c d e s c r i p t i o n must involve n o t s h p l y m u l t i p l e stages of d e r i v a t i o n , but m u l t i p l e types of data and d a t a processing r e q u i r e d t o simulate t h e f u n c t i o n s of d i f f e r e n t body organs. (See Figure 1.) Thus, t h e semantic components of t h e grammar a r e designed t o gsocess d a t a structured f o r specific semantic tracts, a s i t were; the a r t i c u l a t o r y component i s designed t o process data s t r u c t u r e d for the vocal t r a c t , t h e audio component i s designed t o process d a t a s t f u c t u r e d f o r the auditory t r a c t , and s o on. Of course, such a model r e q u i r e s d i s t i n c t rule systems and procedures to o p e r a t e on t h e d i f f e r e n t data types i n the various tracts.", |
| "num": null |
| }, |
| "FIGREF43": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "r t h e r tenet of junction theory i s t h a t data types may n o t be intermingled. To dq s o would, f o t example, be tantamount to feeding i n s t r u c t i o n s f o r both the h e a r t 4nd diaphragm t o the diaphragm. Of course, semantic i n s t r u c t i o n s could not be executed by a vocal t r a c t , nor could a r t i c u l a t o r y i n s t r u c t i o n s be executed by a semantic t r a c t . This means, i n eff e o t , t h a t a \"deep s t~u c t u r e \" i s not transfdrmed ( i n the usual sense of the word) into a surface q t r u c t u r e , b u t r a t h e r t h a t semantic data must be usedt o stimulate a r t i c u l a t o r y i n s t r u c t i o n s , orthogrziphic i n s t r u c t i o n s , motor i n s t r u c t i o n s required t o produce g e s t u r e s , t o make one blush, e t c . Thus, i n JG semantic representations t h e r e a r e no l e x i c a l items, s i n c e these are considered t o be a r q i c u l a t o r y inS$xuctions. Similarly, t h e r e i s no semantic inf ormdtion i n phonological repyesentations, s i n c e these a r e a d i f f e r e n t data type. The various d a t a types a r e considered t o be symboli z a t i o n s of each other, not t r a n s T d m or d e r i v a t i o n s of each other. Data stimulation between the various t r a c t s o r components of the system i s accomplished by context s e n s i t i v e coding/decoding procedures, which are intended t o simulate the neural i n t e r f a c e s which coordinate the function bf body organs involved i n speech production. Jupction Grammar takes its name from Junction Rules (J-rules), (SeeFigure 2.) J-rules s t r u c t u r e d a t a t o be processed by t h e various components of t h e grammar. The e s s e n t i a l i n g r e d i e n t s of ev2ry.J-rule are two o r more operands, an operation specifying hdw the operands a r e t u be joined, and a labelling operation which assigns a category t o t h e operands taken as a u n i t . Thus, i n junction grammar not only do r u l e s f o r con-Junction require an operation symbol (visa the phrase structure r u l e S+S & S).but a l l J l u l e s , regardless of t h e i r s p e c i a l i z a t i o n , schematic of t h e model i n its present form i s given i n Figure 3. Basic semantic d a t a i s presumed t o r e s i d e i n t h e form of an information n e t . Drawing upon information i n t h e net, J-rules or g a d ze and s t r u e t u r e inf ormatlon pragmatically, i.e. f o r use i n specific utterances i n s p e c i f i c discourse environments. Fillmore's arguments f o r semantic case r e l a t e s p e c i f i c a l l y t o the need t o d i s t i n g u i s h between b a s i c semantic r e l a t i o n s and pragmatically motivated grammatical r e l a t i o n s . The semantic junction trees (J-trees) generated by J-rules then serve a? the b a s i s f o r coding up a r t i c u l a t o r y i n s t r u c t i o n s , i n s t r u c t i o n s t o t h e arm and hand f o r writing, o r motor i n s t r u c t i o n s of p m d r y types necessary t o produce body language.", |
| "num": null |
| }, |
| "FIGREF44": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "d d i t i o n s t o o r changes i n t h e information net. Junction trees occur in both semantic and a r t i c u l a t o r y data. However, the qpexands and operations are of a t o t a l l y different nature from type to type, since i n t h e semantic component they c o n s t i t u t e complexes of i n s t r u c t i o n s t o be executed by t h e semantic processor, while i n the a r t i c u l a t o r y component they c o n s t i t u t e complexes of i n s t r u c t i o n s t o be execueed by the vocal t r a c t . The operands of semantic trees a r e sememes, i.e. u n i t s which define l o c a t i o n s and s t a t e s i n t h e information net; tEe operands of a r t i c u l a t i o n trees are articulemes, i.e. u n i t s which r e l a t e t o l o c a t i o n s and s t a t e s of t h e vocal tract.", |
| "num": null |
| }, |
| "FIGREF45": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "and 5 are the semantic and a r t i c u l a t i o n trees, r e s p e c t i v e l y , f o r t h e u t t e r a g c e [~a y s a i y t ] . Notice, s p e c i f i c a l l y , that while Why d i d you are not immediate semantic c o n s t i t u e n t s , they are immediate Etrticulatory c o n s t i t u e n t s ;", |
| "num": null |
| }, |
| "FIGREF46": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "sememas. There is no lexical data in semantic trees.", |
| "num": null |
| }, |
| "FIGREF47": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "Figure 4 .", |
| "num": null |
| }, |
| "FIGREF48": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "B a s k Junction Types Junction theory p o s i t s three basic j u n c t i o n o p e r a t i o n s and numeroud subtypes depending upon the d a t a tvpe beinn described. (11 Adjuqcfion r e s u l t s i n the f~rmation of c e r t a i n nuclear u n i t s which serve as a s k e l e t o n t o whicL o t h e r elements may attach. I n semantic trees, p r e d i c a t e s and p r e d i c a t i o n s are formed v i a adjunction. I n a r t i c ul a t i o n t r e e s , semi-syllables and s y l l a b l e s a r e formed via ad junction. (2) Subjunction r e s u l t s i n overlapping c o n s t i t u e n t s of c o n t r a s t i n g rank, i.e. where one i s i n some sense s u b o r d i n a t e t o t h e o t h e r . I n semantic trees, modifiers i n a l l t h e i r variety are subjoiried. In a r t i c u l a t i o n trees, clustered consQnants ar,e subjoined, as w e l l as adjacent s y l l a b l e s having d i f f e r e n t degrees of gtress. Segmental s t r u c t u r e s are a l s o subjoLned t o prosodic c o n s t i -t u e n t~ t o account f o r t h e supra-segmental aspects of a r t i c u l a t i o n . (3' Conjunctipn results i n the format i o n of compounds c o n s i s t i n g of u n i t s of the same category and rank. I n semantic trees, compounds based.onand, -' or andbut are formed via conjunction. I n A-trees, conj u n c t j o n y i e l d s evenly spaced non-overlapping u n i t s having t h e same degree of stress. Now, i n the context of this rather general i n t r o d u c t i o n t o t h e s u b j e c t , l e t us consider dynamic phonological r e p r e s e n t a t i o n s corresponding t o the a r t f e u l a t o r y s t r u c t u r e of s y l l a b l e s , words, and phrases.The i p t a i t i v e articttllatory u n i t of which words consist i s the s y l l a b l e , which i s i n turn com;posed of phonemes. Generally speaking, s y l l a b l e s have as t h e i r nuclear component a coatinuous phoneme w l t h v o c a l i c p r o p e r t i e s . This nuclear phoneme may be d e l i m i t e d both initially and f i n a l l y by a phoneme having consonantal p r o p e r t i e s . Eence, w e observe s y l l a b l e s of the f o l l o w t n g s t r i n g types: D = d e l i m i t e r ; W =,rhucleus; 0 i s n u l l however, we invoke t h e concept of a n u l l d e l i m i t e r $, t h e n t h e s e f o u r s y l l a b l e p a t t e r n s can b e r e d u c e d t o a s i n g l e type, DWD, where D may be e i t h e r n u l l o r non-null. The use of t h e n u l l d e l i m i t e r $ is a c t u a l l y more than a s i m p l i f y i n g assumption, s i n c e i n many c a s e s non-null segmentals r e p l a c e $ i n the a r t i c u l a t i o n stream e i t h e r as f u l l geminates o r p a r t i a l s of neighboring d e l i m i t e r s . A r t i c u l a t o r y Adiunction A s noted above, junction theory attributes to adjunction those kernel configurations upon which all else i s b u i l t up. Since s y l l a b l e s are t h e i n t u i t i v e u n i t s from Qhich words and phrases a r e formed, w e a t t r i b u t e them t o adjunction. There are two basic s y l l a b l e t y p e s , corresponding t o whether t h e sy1labi.c nucleus is joined to t h e initial o r f f n a l d e l i m i t e r , The two cases a r e illustrated i n F i g u r e 6. NUCLEAR-INITIAL SYLLABLE MUCLEAR-FINAL SYLLABLE Figure 6. Two b a s i c s y l l a b l e t y p e s . Recent research provides useful criteria f o r deciding when t o ~' e e each type. Bell-Berti and Harris report t h a t :", |
| "num": null |
| }, |
| "FIGREF49": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "t h e purposes of t h i s d i s c u s s i o n , l e t u s assume that stressed s y l l a b l e s and s y l l a b l e s w i t h s t r o n g vowels are n u c l e a r -i n i q i a l and t h a t o t h e r s y l l a b l e s a r e n u c l e a r -f i n a l . It i s p o s s i b l e , of course, t o formulate j u n c t i o n r u l e s which are not binary, s o t h a t a t h i r d syllable type whose nucleus was equally joined t o both i n i t i a l ahd f i n a l d e l i m i t e r s could be used, W e a v o i d this foumal complication, however, until forced t o i n t r oduce i t by empirical c o n s i d e r a t i o n s . Notice t h a t the use of structure t o r e p r e s e n t s y l l a b l e s makes i t unnecessary t o u s e a f e a t u r e such as [+syllabic]. In comparing the use of this f e a t u r e t o that of the s t r u c t u r a l n o t a t i o n proposed, w e n o t e t h a t each appears t o make d i s t i n c t claims about t h e n o t i o n s y l l a b l e . S p e c i f i c a l l y , t h e feature asserts t h a t a vowel i s syllabic,whereas the tree claims that s p k c i f i c sequences of segmentals c o n s t i t u t e s y l l a b l e s whose nuclear element i s a p a r t i c u l a r segment. Node Labels Turning now t o the matter of node labels, w e observe t h a t ih p r a c t i r e it is d e s i r a b l e to further s u b c a t e g o r i z e D and W in terms of more s p e c i f i c a r t i c u l a t i o n c l a s s e s . W e t h e r e f o r e define D t o i n c l u d e obstruent consonants (C) , liquids (L) , g l i d e s (G) , and n u l l ) For W, vowels (V) and liquids (L) a r e i n d i c a t e d , and perhape i n some cases even continuant o b s t r u e n t s , assuming t h a t expressions s u c h as v o c a t i v e \"pssst\" are t o be analyzed as s y l l a b l e s also. We note parenthetically that g l i d e s (G) are s u s p e c t , since they appear t o be f u n c t i o n a l v a r i a n t s of vowels, i,e, vowels f u n c t i o n i n g d e l i m i t i v e l y . This, however, i s not a problem, s i n c e the use of J-rules t o r e p r e s e n t a r t i c u l a t o r y structures makes i t just a s f e a s i b l e t o consonantalize a vowel by r u l e a s i t i s $n t h e semantic component t o nominalize a verb by qule. In short, the ae of j u n c t i o n trees t o r e p r e s e n t a r t i c u l a t o r y s t r u c t u r e b r i n g s a g r e a t deal of d e s c r i p t i v e power t o b e a r , should we need i t . Thus w e supplant D and W w i t h more d e s c r i p t i v e l y s p e d f i c node l a b e l s and (append t o them some element of t h e i r r e s p e c t i v e v o c a b u l a r i e s as t e r m i n a lu n i t s , as i l l u s t r a t e d byFigure 7.", |
| "num": null |
| }, |
| "FIGREF50": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "The s i g n i f i c a n c e of V 2 and V3 as non-terminal labels i s t h a t of semi-syllable and syllable, r e s p e c t i v e l y . Bear i n mind t h a t t h e o p e r a t i o n symbols appearing between operands a r e r e p r e s e n t a t i v e of t h e artcculatorv junctions ( t r a n s i t i o n s ) between them. Hence non-terminal nodes symbolize a r t i c u l a t o r y sequences c o n s i s t i n g of t h e phonemes they dominate plus t h etransitions necessary to account for continuous movement from one d i s t i n c t i v e vocal tract state to the next. This s i g n i f i e s , in effect, that glven a junction instruction of the form X O Y = 2, there exists a transition T = O ( X ,~) , such that XW i s a continuous articulatory sequence Z cons i s t i n g of the d i s t i n c t i v e units X and Y mediated by transitional T. This aspect of the fornulation is advanced as an a t t e m~ t t o satisfy the need for phonological notation p o t e n t i a l l y capable of explicating both the discrete segmental elements of which the speech chain i s composed, and the coarticulatory transitions which connect them i n l i v e speech. The practical effect of the foxmulation is that one's attention is drawn not t o a yelatively limited s e t of radical phonological changes, but t o the co-articulatory effect of every junction on its operands, regardless of i t s subtlety. Thisis important :f high quality synthetic speech is to be achievkd.Delimiting ClustersI Both i n i t i a l and final syllable delimiters frequently c o n s i s t of c l u s t e r s of segments rather than discrete segments. An a n a l y s i s of such clusters shows that notable assimilative forces are involved. We view this as a form of articulatory subordination, and, consequently, use subjunction as the basic junction type f o r treating such clusters. The fact that a r t i c u l a~i o n trees are capable of showing a variety of compositional arrangements makes i t possible to give whatever internal structure f o r such c l u s t e r s as seems t o be operative. Thus f o r strand, where tr seem t o .be more c l o s e l y associated than st this can be e x p l i c i t l y represented. consider how m u l t i -s y l l a b l e words may be given in the form of a r t i c u l a t i o n trees. The p r o c e d u r e , briefly, i s a s follows, using B a m b i and Donna as the w o r d s to be diagrammed: The s y l l a b l e s are identified. M I -B I [baembi] DON-NA [da .n a ]", |
| "num": null |
| }, |
| "FIGREF51": { |
| "type_str": "figure", |
| "uris": null, |
| "text": "The i n * t i a l d e l i m i t e r of t h e more weakly-stressed s y l l a b l e becomes t h e i n t e r s e c t node. t e r e s t i n g r e s u l t of the not ,tion i s t h a t stress i s no longer a p r o p e r t y of vowels, but of e n t i r e syllables, i.e. t h e d e l i m i t e r s and the vnwe1. Further, stress reflects a r e l a t i o n between c o n s t i t u e n t s , s o t h a t no f e a t u r e s expressing stress values are necessary.Phraa es -Phrases are diagrammed by introducing prosodic c o n s t i t u e n t s (B) t o which word-trees a r e s u b o~d i n a t e d . (Refer t oFigure 5.) The ranking s y l l a b l e , 1.e. the pne receiving primary stress, j o i n s to t h e prosodic c o n s t i t u e n t .The n o t a t i o n i s intended t o r e f l e c t t h e simultaneous execution of segrneni-a1 and supra-segmental u n i t s during the a r t i c u l a t o r y process, i n a way comparable t o t h e multitudinous i n t e r n a l manipulations of an engine a s one turns a crank. The crank of tbe a r t i c u l a t o r y apparatus i s the diaphragm and o t h e r musculature which provide energy and assume o t h e r symboI c a l l y s i g n i f i c a n t s t a t e s a t c e r t a i n i n t e r v a l s during t h e executioh of the seementals.Prosodic c o n s t i t u e n t s r e s u l t i n the s p e c i f i c i n t o n a t i o n a l contours w e hear superimposed over syllables, words, and phrases. While both segmental and suprasegmental c o n s t i t u e n t s are coded i n the context of s e n a n t i c d a t a , we emphasize again t h a t A-trees c o n t a i n only a r t i c u l a t o r y d a t a . Thds, if A-trees a r e compared t o the customary r e p r e s e n t a t i o n s of generative phonology, as typified by those given by 9 Chomsky and Halle (cohpare Figures 5 and 9 ) , i t w i n b e noted that! the syntacto-semantic s u p e r s t r u c t u r e of the r e g u l a r t r e e s a r e replaced by an a r t f c u l a t o r p s s p e r s t r u c t u r e fn the A-trees, The r a t i o n a l e f o r t h i s departure from standard p r a c t i c e i s not only motivated by the requirement impased by t h e theory (that d a t a types n o t be intermingled), but a l s o by the observation t h a t the r e g u l a r t r e e s tend t o neglect prosodic a r t i c u l a t o r y phenomena. When in\u00a3 ormation . r e l a t i n g to these phenbmena is incorporated i n t o articulation t r e e s , i t replaces t h e usual superstructure of S ' s , NP's, and o t h e r similar l a b l e s i n a n a t u r a l way. The prosodic c o n s t i t u e n t s thus introduced are comparable i n their function t o the i n t o n a t i o n contours a s s o c i a t e d by r u l e with segmental sequences i n the systbm proposed by Leben. 10 1 # # #tele+graph#ic # # ~lcomunicatei!! ion 8 # B # u n c t i o n a l Versus CategorLal Informatioh -Tha proposed system of phonological d e s c r i p t i o n makes p o s s i b l e m Lntexesting hypothesis regarding many of t h e f e a t u r e s used i n c u r r e n t i e s c r i p t i o n s . S p e c i f i c a l l y , i f A-trees a r e i n some senge a r e f l e c t i o n of a c t u a l a r t i c u l a t o r y processes, then phonological r e p r e s e n t a t i o n s whfch do not use t r e e s wili c o n s i s t of an i n t e r m i x t u r e of f u n c t i o n a l and c a t e g o r i a l l a b l e s ( f e a t u r e s ) . For exaxriple, ff t r e e s are used t o r e p r e s e n t t h e r e l a t i o n s bktween s u b j e c t , verb, and o b j e c t , i t i s not necessary t o label t h e s u b j e c t as such o r t h e o b j e c t a s such, s i n c e s t r u c t u r a l r e l a t i o n s make t h e s e n o t i o n s e x p l f c i t . I f t r e e s were n o t used t o r e p r e s e n t sentence s t r u c t u r e , however, f u n c t i o n a l l a b e l s would have cp be used. S i m i l a r l y , i t follows t h a t i f t r e e s a r e an appropriate medium f o r phonological d e s c r i p t i o n , but have n o t been used, then f u n c t i o n a l and c a t e g o r i a l information are intermingled h c u r r e n t d e s c r i p t i o n s . If this i s t r u e , then i t should be p o s s i b l e t o a b s t r a c t f u n c t i o n a l information away (and consequently not write i t in f e a t u r e form) by e l a b o r a t i n g A-tree n o t a t i o n . While t h e proposed system i s s t i l l i n i t s infancy, s o t o speak, some i n t e r e s t i n g i n i t i a l observations in t h i s regard can be made a t t h i s t i m e . F i r s t , major category f e a t u r e s become node l a b e l s i n a n a t u r a l way, t h u s suggesting why t h e formal i l l u s i o n exists t h a t a change, f o r example, of [+cons] + [-cons] i s equal i n magnitude t o a change of i-hroice] -[-voice] Second, [ t s y l l a b i c ] ([?consonantal] and [&vocalic] a r e a l s o used i n some systems) are f u n c t i o n a l l a b e l s and need not be wr'itten i f s y l l a b l e s are given as tree s t r u c t u r e s . Third, stress a t the segmental l e v e l and unmarked p i t c h a t t h e prosodic l e v e l become i m p l i c i t i n s t r u c t u r e i n terms of the rank of operand's in articulakory subjunction and need not be s p e c i f i e d by feature. While it is beyond the ecop8 of t h i s paper to elaborate t h i s point further, i t is without doubt the most interesting and provocative consequence of the research to date.", |
| "num": null |
| }, |
| "TABREF0": { |
| "type_str": "table", |
| "html": null, |
| "num": null, |
| "text": ". . . . METHOD . . . . . . . . . . . . . . . . . . . . . . . . 31", |
| "content": "<table><tr><td>INTRODUCTION AND SURVEY OF RESEARCH IN PITCH CONTOURS</td><td>6</td></tr><tr><td>S e c t i o n</td><td/></tr></table>" |
| }, |
| "TABREF1": { |
| "type_str": "table", |
| "html": null, |
| "num": null, |
| "text": "", |
| "content": "<table><tr><td colspan=\"2\">Now consider the q u e s t i o n \"t h a t you heard someone come in but you d i d n o t see</td></tr><tr><td/><td>What happened?</td></tr><tr><td>Ib John drove t o the store.</td><td>Who drove t o the s t o r e ?</td></tr><tr><td>l c John drove t o the s t o r e .</td><td>Row d i d John p e t t o t h e store.</td></tr><tr><td>I d John drove t o t h e store.</td><td>Where d i d John d r i v e ?</td></tr><tr><td>ie John drove to the s t o r e ?</td><td>John drove to the -,tore, you know.</td></tr><tr><td>(Are you s u r e that's what</td><td/></tr><tr><td>you meant t o say?)</td><td/></tr></table>" |
| }, |
| "TABREF2": { |
| "type_str": "table", |
| "html": null, |
| "num": null, |
| "text": "", |
| "content": "<table><tr><td>6.</td><td>John or Mary come?</td></tr><tr><td>7.</td><td>Did John o r Mary come?</td></tr><tr><td>11.</td><td>The boys w h o s t u d y get</td></tr><tr><td/><td>good grades.</td></tr><tr><td>12.</td><td>The boys who study get</td></tr><tr><td/><td>good g r a d e s ,</td></tr><tr><td>16.</td><td>They are eating a p p l e s .</td></tr><tr><td>1 7 .</td><td/></tr></table>" |
| }, |
| "TABREF4": { |
| "type_str": "table", |
| "html": null, |
| "num": null, |
| "text": "", |
| "content": "<table><tr><td>AVERAGE SCORES</td></tr></table>" |
| } |
| } |
| } |
| } |
