| { |
| "paper_id": "J75-4019", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T02:40:39.330496Z" |
| }, |
| "title": "", |
| "authors": [], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "", |
| "pdf_parse": { |
| "paper_id": "J75-4019", |
| "_pdf_hash": "", |
| "abstract": [], |
| "body_text": [ |
| { |
| "text": "and other computations are ~e q u i r e d in the process of p a s t i n g these t r e e s t o g t h e r i n a p p r o p r i a t e p l a c e s u n t i l a ' s i n g l e phrase marker I s a t t a i n e d w h i c h w i l l l e a d t o t h e s u r f a c e string. S i n c e we are g e n e r a t i n g from a s e m a n t i c network, aL1 t h e p a s t i n g t o g e t h e r i s a l r e a d y done. Grabbing t h e n e t w o r k by the node of i n t e r e s t and l e t t i n g t h e network dangl-e from i t gives a s t r u c t u r e m i c h may be searched a p p p o g r i a t e l y i n o r d e r t o g e n e r a t e t h e s u r f a c e s t r f n g directly i n l e f t t o r i g h t f a s h i o n . s e m a n t i c rietwork and its ATN t a k e s as i n p u t a l i n e a r list c o n t a i ning t h e semantic node and a generation p a t t e r n consisting o f a \" s e r i e s of c o n s t r a i n t s on t h e moclalltyfl (Simmons e t a l . , 1973, p . 9 2", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The g e n e r a t o r d . e s c r i b e d i n Schank e t a l . , 1973, t r a n s l a t e s from a \" c o n c e p t u a l s t r u c t u r e f 1 i n t o a network o f t h e form o f Simmons ' network w h i c h is t h e n g i v e n t o a v e r s i o n o f Simmons g e n e r a t i o n program. The two s t a g e s use d i f f e r e n t mechanisms. Our system amounts t o a u n i f i c a t i o o f these two s t a g e s .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The g e n e r a t o r , as d e s c r i b e d i n this p a g e r , as w e l l a s SNePS, a parser and an i n f e r e n c e mechanism have been written i n LISP 1 . 6 and a r e r u n n i n g I n t e r a c t i v e l y on a DEC system-10 on t h e I n d i a n a University Computing Network. The semantic network r e p r e s e n t a t i o n b e i n g used does n o t i n - a t e n s e can be g e n e r a t e d f o r an o u t p u t s e n t e n c e , b u t i t may be a d i f f e r e n t t e n s e t h a n that of t h e o r i g i n a l i n p u t s e n t e n c e i f time If we now consider g e n e r a t i n g i n these terms, we s e e t h a t t h e r e i s no simple n e x t i n p u t f u n c t i o n . The g e n e r a t o r will focus on some semantic node f o r a w h i l e , r e c u r s i v e l y s h i f t i n g its a t t e n t i o n to adjacent nodes and back. S i n c e there are s e v e r a l a d j a c e n t nodes, ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "(CHARLIE IS BELIEVING THAT A DOG KISSED SWEET YOUNG L U C Y ) , * (SNEG", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Representation i n t h e Semantic_Network", |
| "sec_num": null |
| }, |
| { |
| "text": "f a c e strings depending on t h e s e m a n t i c node i t i s g i v e n , and a single semantic node may be d e s c r i b e d by d i f f e r e n t s u r f a c e s t r i n g s depending on t h e grammar node i t is g i v e n t o . S i n c e g e n e r a t i o n from a semantPc network r a t h e r t h a n from d i s c o n n e c t e d p h r a s e markers,, t h e s u r f a c e s t r i n g may be g e n e r a t e d d i r e c t l y , l e f t t o r i g h t . I n t r o d u c t i o n I n t h i s p a p e r , w e d i s c u s s t h e approach b e i n g taken i n t h e E n g l l s h g e n e r a t i o n subsystem o f a n a t u r a l language understanding system presently under development a t I n d i a n a U n i v e r s i t y . The c o r e o f t h e u n d e r s t a n d e r i s a s e m a n t i c network p r o c e s s i n g s y s t e m , SNePS ( S h a p i r o , 1975), which i s a d e s c e n d a n t of t h e MENTAL s e m a n t i o subsystem ( S h a p i r o , 1971a, 1971b) o f the M I N D s y s t e m (Kay, 1973).The r o l e of t h e g e n e r a t o r 1 3 t o describe, i n E n g l i s h , any oP t h e nodes i n the sernantjc network, a l l o f which r e p r e s e n t concepts o f the understanding aystem.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "The a u t h o r i s indebted t o J o h n Lowrance, who lfiplemnted t h e", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgements", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": {}, |
| "ref_entries": { |
| "FIGREF0": { |
| "uris": null, |
| "text": "system b e a r s a s u p e r f i c i a l r e s e m b l a n c e t o that d e s c r i b e d fn Simmons and Slocum, 1972 and i n Simmons, 1973. T h a t s y s t e m , h o w e v e r , s t o r e s s u r f a c e i n f o r m a t i o n s u c h as t e n s e and v o i c e i n its", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF1": { |
| "uris": null, |
| "text": "Conceptual i n f o r m a t i o n derived from p a r s e d s e n t e n c e s o r deduced from other i n f o r m a t i o n ( o r i n p u t d i r e c t l y v i a t h e SNePS user's l a nguage) i s stored i n a s e m a n t i c network. The nodes i n the network represent c o n c e p t s which may b e d i s c u s s e d and reasoned abaat. The edges represent semantic b u t n o n -c o n c e p t u a l b i n a r x relations between nodes. There are a l s o a u x i l i a r y nodes w h i c h SNePS can use or which the user c a n use as SNePS v a r i a b l e s . (For a more complete diecussion of SNePS and the network n e e Z h a p i r o , 1975.)", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF2": { |
| "uris": null, |
| "text": ".t.6 be f e a t u r e s of' t h e s u r f a c e s t r i n g such as t e n s e , v o i c e o r main v s . r e l a t i v e c l a u s e . I h s t e a d of t e n s e , t e m p o r a l i n f o r m a t i o n i s s t o r e d P e I a t i v e t o a growing t i m e l i n e i n a manner s i m i l a r t o t h a t of B r u c e , 1 9 7 2 . From t h i s i n f o r m a t i o n", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF3": { |
| "uris": null, |
| "text": "has p r o g r e s s e d i r i %he 5 n t e r i m . The v o i c e o f a g e n e r a t e d s e n t e n c e i s u s u a l l y d e t e r m i n e d by t h e top l e v e l c a l l t o the g e n e r a t o r funct i o n . However, sometimes i t i s d e t e r m i n e d b y t h e g e n e r a t o r gramm a r . F o r example, when g e n e r a t i n g a r e l a t i v e c l a u s e , v o i c e i s d e t e r m i n e d by w h e t h e r t h e nodn b e i n g m o d i f i e d -i s t h e a g e n t o r object o f t h e a c t i o n d e s c r i b e d by the r e l a t i v e clause. The Main c l a w e of a g e n e r a t e d s e n t e n c e depends on which s e m a n t i c node i s g i v e n t o t h e g e n e r a t o r in t h e t o p l e v e l c a l l . O t h e r nodes conn e c t e d t o it may r e s u l t i n r e l a t i v e c l a u s e s b e i n g g e n e r a t e d . These r o l e s may b e r e v e r s e d i n o t h e r t o p l e v e l c a l l s t o t h e g e n e r a t o r . The g e n e r a t o r i s d r i v e n b y two s e t s of d a t a : t h e s e m a n t i c n e twork and a grammar i n t h e form of a r e c u r s i v e augmented t r a n s i t i o n network (ATN) similar t o t h a t o f Woods, 1973. The edges on o u r ATN a r e somewhat d i f f e r e n t from t h o s e of Woods s i n c e o u r view i s t h a t t h e g e n e r a t o r i s a t r a n d u c e r from a network i n t o a l i n e a r s t r i n g , whereas a p a r s e r I s a t r a n s d u c e r f r o m a l i n e a r string i n t o a t r e e o r network. The changes t h i s e n t a i l s a r e d i s c u s s e d below. During any p o i n t i n g e n e r a t i o n , t h e g e n e r a t o r i s working on some p a r t i c u l a r s e m a n t i c node. F u n c t i o n s on t h e edges of t h e ATN can examine the network c o n n e c t e b t o t h i s node and fail o r s u c c e e d a c c o r d i n g l y . I n t h i s way, nodes o f t h e ATN can \" d e c i d e \" what surface form i s most a p p r o p r i a t e f o r d e s c r i b i n g a s e m a n t i c node, w h i l e d i f f e r e n t ATN n o d e s may g e n e r a t e different s u r f a c e foI?ms t o describe t h e same semantic node. A common assumption among l i n g u i s t s i s that g e n e r a t i o n b e g i n 3 w i t h a set of d i s c o n n e c t e d d e e p phrase markers. T r n n u f o~-m a t l o n~ LEX Semantic Network RepresentatLon for \"Charlie believes that a dog kissed sweet young Lucy,\" \"Charlie i s a person,\" and \"Lucy i s a person. ,.Af=rmation considered t o be features o f surface strings are n o t stored in the semantic network, but a r e used by the p a r s e r in cons t r u c t i n g t h e network rrom t h e i n p u t s e n t e n c e and by t h e g e n e r a t o r f o r generating a s u r f a c e s t r i n g from t h e network. For example, tense i s mapped into and from temporal r e l a t i o n s between a node r e p r e s e n t i n g that some a c t i o n has, i s , or will occur and a growing t i m e l i n e . Restrictive relative clauses are used by t h e p a r s e r to Identify a node being d i s c u s s e d , while n o n -r e s t r i c t i v e r e l a t i v e clauses may result i n new i n f o r m a t i o n b e i n g added t o t h e network. The example used i n t h i s p a p e r i s designed t o i l l u s t r a t e t h e generation issues being discussed. Although i t also i l l u s t r a t e s our general approach t o representational issues, some details w i l l *(SNEG M O O L b )", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF4": { |
| "uris": null, |
| "text": "i g u r e 2 : R e s u l t s o f calls t o t h e g e n e r a t o r with nodes from F i g u r e I* User i n p u t i s on l i n e s beginning with *. c e r t a i n l y change as work p r o g r e s s e s . F i g u r e 1 shows t h e s e m a n t i c network r e p r e s e n t a t i o n for the i n f o r m a t i o n i n the sefitencesb, \"Charlie b e l i e v e s that a dog k i s s e d sweet young L u c y , \" \" C h a r l i e is a p e r s o n , \" and \"Lucy i s a p e r s o n . \" Converse edges a r e n o t shown, b u t i n a l l c a s e s t h e label of a c o n v e r s e edge i s t h e l a b e l o f t h e f o rward edge with ' * ' appended e x o e p t f o r BEFORE, whose c o n v e r s e edge is l a b e l l e d AFTER. LEX p o i n t e r s p o i n t t o nodes c o n t a i n i n g ; l e x i c a l e n t r i e s . STIME p o i n t s t o the s t a r t i n g time o f a n a c t i o n and ETIME t o i t s e n d i n g time.Nodes r e p r e s e n t i n g i n s t a n t s of time a r e rel a t e d t o each o t h e r by t h e BEFORE/AFTER e d g e s . The auxiliary node NOW has a :VAL p o i n t e r to t h e c u r r e n t i n s t a n t o f t i m e .", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF5": { |
| "uris": null, |
| "text": "i g u r e 2 shows t h e g e n e r a t o r ' s o u t p u t f o r many o f t h e n a d c s o f F i g u r e 1. Figure 3 shows t h e Lcxicon uncd i n the e x a m p l e . (BELIEVE((CTGY.V)(I~BELIEVE) (PRES.BELIEVES)(PAST.BELIEVED)(PASTP.BELIEVED)(PRESP.BELIE~JING))) (CHARLIE ( (CTGY . NPR) (PI . CHARLIE) 1') (DO~'((CTQY.N)(SING.DOG)(PLUR.DOGS))) (K~SS((CTGY .v) CINF.KISS) ~PRES.KISSES)~PP.ST.KISSED~~PASTP.KISSED)(PRESP.KISSING~~) (LUCY-((CTGY.NPR)(PI.LUCY))) (PERSON((CTGY.N!(sING.PERSON)(PLUR.PEOPLE)~~ (SWEET(~CTGY.ADJ)(PI.SUEET) 1 ) (YO~NG((CTGY.ADJ)(PI.YOUNG))) The l e x i c o n used i n t h e example of Figures 1 and 2. Generation as P a r s i n g Normal p a~s i n g involves t a k i n g i n p u t from a linear s t r i n g and producing a tree o r network structure as o u t p u t . Viewing this in terns of an ATN grammar as described i n Woods, 1973, t h e r e i s a well-defined n e x t input f u n c t i o n which simply p l a c e s s u c c e s s i v e word6 into t h e * * r e g i s t e r . The o u t p u t f u n c t i o n , however, i s more complicated, uslng BUILDQ t o build p i e c e s of t r e e s , o r , as i n o u r parser, a BUILD function t o build p i e c e s o f network.", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF6": { |
| "uris": null, |
| "text": "connected by variously l a b e l l e d e d g e s , t h e grammar a u t h o r must specify which edge t o f o l l o w when t h e g e n e r a t o r i s to move t o a n o t h e r semantic node. For t h e s e r e a s o n s , t h e same focal semantic node is used when traversing edges of the grammar network and a new semantic node isspecified by gl,ving a p a t h from t h e c u r r e n t semantic node when pushing t o a new grammar node. The r e g i s t e r SNODE is u s e d t o hold t h e current semantic node. The output f u n c t i o n of g e n e r a t i o n i s s t r a i g h t f o r w a r d , s i m p l ybeing concatenation o n t o a growing s t r i n g . Since t h e o u t p u t s t r i n g is analogous t o the parser's I n p u t s t r i n g , w e store i t in the reg-", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF7": { |
| "uris": null, |
| "text": "sexp form -: := w f o r m sform a c t ion : := (SETR regname f o m (ADDTO regname form* ) (ADDON regname f o m * ) sexp test : : (MEMS form form) (PAT23 sform sarc* sform) form sexp gnode : := <any LISP atom which r e p r e s e n t s a grammar node> word : := <any LISP atom> regname ::= <any non-numeric LISP atom used as a r e g i s t e r name> sarc ::= <any LISP atom used as a semantic a r c l a b e l > l f e a t : := <any LISP atom used as a l e x i c a l feature> sexp : := <any LISP s-expression> Syntax o f edged of g e n e r a t o r ATN grammars i s t e r *. When a pop oc c u r s , i t i s always t h e c u r r e n t v a l u e of * t h a t is r e t u~n e d .", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF8": { |
| "uris": null, |
| "text": "shows the s y n t a x of t h e g e n e r a t o r ATN grammar. Object language s y m b o l s are ) , ( , and elements i n c a p i t a l l e t t e r s . Metalanguage symbols are i n lower c a s e , Square b r a c k e t s e n c l o s e opt i o n a l elements. Elements followed by * may be repeated one o r more t j m e s . Angle b r a c k e t s e n c l o s e i n f o r m a l E n g l i s h d e s c r~p t i o n s .Semantics of Eage Functions In this section, the semantics of t h e grammar arcs, forms and t e s t s ape p r e a e h t e d and compared t o those of Woods1 ATNs.? The ---t A l l comparlsona arc with Woodo, 1973. NCL ') JUMP(SETR * @ ( / / / / NO GRAMMAR NODE F O U N D ) ) M E N D )The default e n t r y i n t o the grammar n e t w o~k .essential differences are those required by the differences between generating and p a r s f n g as discussed in the p r e v i o u s s e c t i o n . (TEST t e s t [action]*(TO gnode)) If the test i s successful ( e v a l u a t e s t o non-NIL), t h e a c t i o n s are performed and g e n e r a t i o n c o n t i n u e s at gnode. I f t h e test f a i l s , this edge i s n o t t a k e n . TESTis t h e same as WoodsT TST, w h i l e TEST(GETF s a r c ) i s analogous t o Woods' CAT. (JUMP action]*(^^ gnode)) Equivalent to (TEST T [actlon]*(TO gnode)). JUMP is simllar In use to Woods JUMP, b u t the d i f f e r e n c e from TEST T disappears s i n c e no edge r'consumesl' a n y t h i n g . (MEM wform (word*) t e s t [action]* (TO gnode) ) If the value of wform has a non-null intersection w i t h t h e list of words, the test is performed. If t h e test is also successf u l the actions are performed and g e n e r a t i o n continues a t gnode. if either the Intersection is null or the t e s t f a i l s , the edge Generation of subject of sub ject-verb-ob ject sentence. is not taken. This is similar in form to Woods1 MEM, but mainly used for testing r e g i s t e r s . (NOTMEM wform (word*) test [action]*(TO gnode)) This is e x a c t l y l i k e MEM except the intersection must be n u l l . (TRANSR ([regnamel] regname2 regname3) t e s t [ a c t i o n ] * gnode) ) If regnamel i s present, t h e c o n t e n t s o f regname2 are added on the end of regnamelo If regname is empty, t h e edge is not", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF9": { |
| "uris": null, |
| "text": "Otherwise, t h e f i r s t element i n regname is removed and 3 placed in regname2 arid t h e t e s t i s performed. If t h e t e s t f a i l s ,t h e edge is not taken, b u t if i t succeeds, the a c t i o n s are performed and generation continues at gnode. TRANSR i s used t o iterate t h r o u g h several nodes all in t h e same semantic r e l a t i o n w i t h t h e main semantic node. (GEN gnodel s f o r m [action]*regname [ a c t i o n ] * ( T O gnode*)) The f i r s t s e t of a c t i o n s are performed and t h e generation i s c a l l e d recursively with the semantic node that is the v a l u e of sform and a t t h e grammar node gnodel. If this g e n e r a t i o n is successful ( r e t u r n s non-NIL), the result i s placed i n t h e r e g i s t e r regname, t h e second s e t of a c t i o n s are performed and g e n e r a t i o n c o n t i n u e s at gnode*. If the generation fails, the edge i s not t a k e n . This I s the same as Woods1 PUSH but requires a semantic node t o be npecif i e d and allows any register to b e used t o h o l d t h e result. Instead o f having a POP edge, a r e t u r n automatically occurs when (ADDON * @WILL @HAVE) (ADDON * QWQULD) UI@(ADDON * @(///CANNOT COMPUTE TENSE)) TEST( MENS (GETR REF) ( * @NOW) ) P ( A D D 0 N * (LExLOOK PASTP(GETF VERB))) JuMP(ADDON * (LEXLOOK INF(GETF V E R B ) ) ) CTIINF ) Tense g e n e r a t i o n network. transfer is made to the node END. At t h a t point, the c o n t e n t s o f the register named * are r e t u r n e d . (CONCAT form form*) The forms are evaluated and concatenated i n the order g i v e n . Performs a r o l e analogous to t h a t of Woodst BUILDQ. (GETF sarc [sform]) Returns a list of a l l semantic nodes at the end of the semant i c arcs l a b e l l e d sarc from t h e aemantic node whl ch i s t h e value", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF10": { |
| "uris": null, |
| "text": "The tenses of \"breakn which t h e network o f F i g u r e 7 can generate. o f sfomn.If sform i s missing, SNODE is assumed. Returns N I L i f I t h e r e a r e no such s e m a n t i c nodes. It is similar in the Woods' GETF i n t h e l e x i c a l domatn.Returns the contents o f register regname. It i s essentially t h e same as Woodst GETR. o f the lexical feature, l f e a t , o f the l e x i c a l e n t r y a s s o c i a t e d w i t h the semantic node which i s t h e value o f sform. If sform is missing, SNODE i s assumed. If no lexical entry i s assoc i a t e d with t h e semantic node, NIL i s r e t u r n e d . LEXLOOK i s sirnilar to Woods1 GETR and as a l s o i n the l e x i c a l domain. (SETR regnarne form) The v a l u e of form is p l a c e d in the r e g i s t e r regname. It is the same as Woods1 SETR. (ADDTO Pegname form*) E q u i v a l e n t t o (SETR regname (CONCAT (GETR regname) form*)). Equivalent to (SETR regname (CONCAT form* (GETR regname))). (MEW f o r m form) Returns T If t h e values of the t w o forms have a non-null intersection, N I L otherwise. KLNP (GETF AGENT) REG (ADDON * (GETR 'REG) ) CpRGDOB3 GEN NCLNP (GETF OBJECT) REG (ADDON * (GETR REG ) ) I Generating the s u r f a c e o b j e c t . (PATH s f o m l sarc* sformp) Returns T i f a p a t h descr3bedl By t h e sequence o f semantic a r c s exists between t h e value of sfoml and sformp. If %he sequence i s sarcl s a r c 2 ... sarc,, the p a t h d e s c r i b e d i s t h e same as t h a t i n d i c a t e d by sarcl* sarc2* ... sarcni. If no such p a t h exists, NIL i s r e t u r n e d . (Remember, * means r e p e a t one o r more times. ) Discussion o f an Example Orammar Network, The top level g e n e r a t o r f u n c t i o n , SNEG, Ls given as arguments a semantic node and, o p t i o n a l l y , a grammar node. If t h e grammar node i s not g i v e n , g e n e r a t i o n begins a t t h e node G 1 which should be a small disccirnination n e t t o choose t h e p r e f e r r e d d e s c r i p t i o n for t h e given semantic node, T h i s p a r t of t h e example grammar is shown i n Figure 5. Jn it w e see t h a t t h e preferred d e s c r i p t i o n f o r any semantic node i s a s e n t e n c e . If no sentence can b e formed a noun phrase w i l l be t r i e d . Those a r e t h e only p r e s e n t l y available options. Semadtic nodes w i t h an outgoing VERB edge can be d e s c r i b e d by a normal SUBJECT-VERB-OBJECT s e n t e n c e . (For this example, we have n o t used a d d i t i o n a l c a s e s . ) F i r s t t h e s u b j e c t i s g e n e r a t e d ,", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF11": { |
| "uris": null, |
| "text": "Generating t h e t h r e e \"non-regular\" s e n t e n c e s .which depends on whether t h e s e n t e n c e i s t o b e i n a c t i v e o r p a s s i v e voice. A l t e r n a t i v e l y , t h e c h o i c e could be e x p r e s s e d i n terms o f whether t h e a g e n t o r o b j e c t is t o be t h e t o p i c as s u g g e s t e d by Kay, 1975. Figure 6 shows Lhe network that generates t h e s u b j e c t . The r e g i s t e r DONE h a l d s semantic noaes f o r which s e n t e n c e s are b e i n g g e n e r a t e d for l a t e r checking too prevent i n f i n i t e r e c u r s i o n . WPthout i t , node MOO23 o f F i g u r e I would be d e s c r i b e d as, \"A dog which k i s s e d young sweet Lucy who was k i s s e d b y a dog w h i c h k i s s e d . ..\"The i n i t l a 1 p a r t of t h e PRED network i s concerned with generating t h e tense. T h i s depends on t h e BEFORE/AFTER p a t h between t h e s t a r t i n g and/or endinb time of t h e a c t i o n and t h e c u r r e n t v a l u e o f NOW, w h i c h , i s given by t h e form ( # @NOW).Figure 7 shows t h e t e n s e g e n e r a t i o n network. Figure 8 shows t h e t e n s e s t h i s network i s a b l e t o generate. A f t e r the verb group is g e n e r a t e d , the surface o b j e c t is genera t e d by d e s c r i b i n g e i t h e r t h e s e m m t i c agent o r object. F i g u r e 9 skuws this part of t h e network The o t h e r t h r e e kinds of s e n t e n c e s a r e lor d e s c r i b i n g nodes r e p r e s e n t i n g : (1) t h a t something has a p a r t i c u l a r adjective a t t r i b uable t o it, ( 2 ) that something has a name, ( 3 ) that something i s a member of some c l a s s . The networks f o r t h e s e a r e shpwn i n F i g u r e 1 0 . Again, t h e DONE r e g i s t e r i s used t o p r e v e n t such s e n t e n c e s as \"Sweet young Lucy I s sweet,\" \"Charlie i s Charlie.\" and \"A dog is a dog.\" GEN 3 (GETF OBJECT(GETF VERB*))REG(ADDON * @THAT(GETR R E G ) ) cICL>GEN SREG SNODE *(ADDTO * @THAT) Generating norninalized verbs and sentences. Pugure 5 showed three b a s i c kinds of noun phrases t h a t can b e generated: the noun clause or nominalized sentence, s u c h as \" t h a t a d o f~ kissed sweet young Lucyt'; the nominalized verb, such as \"the kisslng of sweet young Lucy by a dogn; the r e g u l a r noun phrase. The first two of these are generated by the network shown in Figure 11. Bere DONE is ased t o prevent, f o r example, \" t h e kissing of sweet young Lucy who was kiesed by a dog by a dog.\" The regular noun phrase network begins w l t h another descriminat i o n net which has the following p r i o r i t i e s : use a name of the o b j e c t ; use a class the obJect belongs to; use something e l s e known about *he obJect. A lower p r i o r i t y description will be used if all h i g h e r priority descriptions are a l r e a d y in DONE. Figure 12 shows t h e beglnnlrrg of the noun phrase network. A d j e c t i v e s are added before t h e mame or before tkre class name and a relative clause is added after. GEN ADJS SNODE *(ADDTO * (LEXLOOK SING(GETE CLASS(GETF MEMBER*))))", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF12": { |
| "uris": null, |
| "text": "The beginning of the noun phrase network.", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF13": { |
| "uris": null, |
| "text": "i g u r e 13 shows the a d j e c t i v e s t r i n g g e n e r a t o r and Figure 1 4 shows t h e r e l a t i v e c l a u s e g e n e r a t o r . Notice t h e use o f t h e TRANSR edges f o r i t e r a t i n g . A t t h i s time, we have no t h e o r y f o r determining t h e number o r which a d j e c t i v e s and r e l a t i v e c l a u s e s t o generate, s o a r b i t r a r i l y we generate a l l a d j e c t i v e s n o t a l r e a d y on DONE but only one relative c l a u s e . W e have n o t y e t implemented any o r d e r i n g of adjectives. It i s merely f o r t u i t o u s t h a t \"sweet young L u c y t is eenerated rather than \"young sweet Lucyft. The network i s w r i t t e n s o t h a t a r e l a t i v e clause f o r which t h e noun i s t h e deep agent is preferred over one i n which t h e noun i s t h e deep o b j e c t . Notice that t h i s choice determines t h e v o i c e of t h e embedded clause. The fomn (STRIP(FIND MEMBER (1. SNODE) CLASS (FIND LEX PERSON))) is a call t o a SNePS f u n c t i o n t h a t determines i f t h e o b j e c t i s known t o be a person, i n ~h i c h c a s e \"WHO\" i s used r a t h e r than \"'WHICHft. This determination i s made by r e f e r r i n g t o t h e semantic network r a t h e r t h a n by i n c l u d i n g a HUMAN f e a t u r e on t h e l e x i c a l e n t r i e s f o r LUCY and CHARLIE.", |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF14": { |
| "uris": null, |
| "text": "The network f o r g e n e r a t i n g a s t r i n g o f a d j e c t i v e s .Notice that any i n f o r m a t i o n a b o u t t h e object being described by a noun phrase may be used to construct a relatfve clause even i f that Lnfomnation derived from some main clause. Also, while the g e n e r a t o r i s examining a semantic node all t h e i n f o r m a t i o n aboutthat node i s reachable from i t and may be used d i r e c t l y . There i s no need t o examine disjoint deep phrase markers t o d i s c o v e r where they can be attached t o each o t h e r s o that a complex sentence can be derived.Future Work Additional work needs t s be done in developing t h e s t y l e of g e n e r a t i o n d e s c r i b e d i n this p a p e r . Experience with larger and richer networks will lead to the f o l l o w i n g i s s u e s : d e s c r i b i n g a node by a pronoun when that node has been described e a r l i e r i n the string; r e g u l a t i n g v e r b o s i t y and complexity, p o s s i b l y by the u s e of r e s o u r c e bound8 simulating the limitations of s h o r t term memory* keeping sub-, ordinate clauses and descriptions to t h e point of t h e c o n v e r s a t i o n p o s s l b l y by the use of a TO-DO register h o l d i n g t h e n o d e s t h a t are t o b e included in t h e string. In thla paper, only i n d e f i n i t e descriptions were g e n e r a t e d . W e are working on a routine t h a t w l l l I d e n t i f y t h e p r o p e r subnet o f t h e semantic network t o j u s t i f y a d e f i n i t e d e s c r i p t i o n . T h i s must b e such that it uniquely i d e n t i f l e e t h e node being d e s c r i b e d . ? !*I!JD :.lEMEEFl(f SI\\l; DE)CLASS (YT?ID E X PERSON) ) ) The r e l a t i v e clause g e n e r a t o r .", |
| "num": null, |
| "type_str": "figure" |
| } |
| } |
| } |
| } |
