| { |
| "paper_id": "A83-1031", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T02:11:25.972510Z" |
| }, |
| "title": "INTERACTIVE NATURAL LANGUAGE PROBLEM SOLVING", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [ |
| "Pragmatic" |
| ], |
| "last": "Approach", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Duke University Durham", |
| "location": { |
| "settlement": "North Carolina" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Biermann", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Duke University Durham", |
| "location": { |
| "settlement": "North Carolina" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Rodman", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "B", |
| "middle": [ |
| "T" |
| ], |
| "last": "Betancourt", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Duke University Durham", |
| "location": { |
| "settlement": "North Carolina" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Heidlage", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Duke University Durham", |
| "location": { |
| "settlement": "North Carolina" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Fink", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Duke University Durham", |
| "location": { |
| "settlement": "North Carolina" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Gilbert", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Duke University Durham", |
| "location": { |
| "settlement": "North Carolina" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Gregory", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "I NTRODUCT ION A class of natural language processors is described which allow a user to display objects of interest on a computer terminal and manipulate them via typed or spoken English sentences.", |
| "pdf_parse": { |
| "paper_id": "A83-1031", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "I NTRODUCT ION A class of natural language processors is described which allow a user to display objects of interest on a computer terminal and manipulate them via typed or spoken English sentences.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": ", H. Deas , L.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "This paper concerns itself with the implementation of the voice input facility using an automatic speech recognizer, and the touch input facility using a touch sensitive screen.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "To overcome the high error rates of the speech recognizer under conditions of actual problem solving in natural language, error correction software has been designed and is described here. Also described are problems involving the resolution of voice input with touch input, and the identification of the intended referents of touch input.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "To measur~ system performance we have considered two classes of factors: the various conditions of testing, and the level and quality of training of the system user.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "In the paper a sequence of five different testing situations is observed, each one resulting in a lowering of system performance by several percentage points below the previous one.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "A training procedure for potential users is described. and an experiment is discussed which utilizes the training procedure to enable users to solve actual non-trivial problems using natural language voice communication. including arbitrarily deep nesting of noun groups, extensive conjunction processing, user defined imperative verbs, and looping and branching features.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "More recently, a domain independent abstraction of the NLC system has been constructed and now is being specialized to handle a text processlng task.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "In this system, text can be displayed and modified or formatted with natural language commands.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "Current work emphasizes the addition of voice input, voice output, and a touch sensitive display screen.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "Speech recognition is being done on an experimental basis with the Nippon Electric DP-200 Connected Speech Recognizer in both discrete and connected speech modes, and with the Votan Corporation V-SO00 Development System.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "The touch sensitive screen being used is a Carroll touch panel mounted on a 19-inch color monitor.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "Voice response is also provided by the Votan V-5000 which assembles and vocalizes digitally recorded human voice messages. The work has progressed to the point where OUr natural language matrix computer NLC is operative under voice control using the DP-200 and the text processing system is beginning to function using the V-5000 speech recognizer.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "The touch panel interface and voice response systems are still in the design phase.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "The goal of the project is to make possible voice and touch interactions of the following kind:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "Retrieve file Budget83. Prompts and error messages will be given by voice response, gystem design is aimed at allowing fast interactive control of the objects on the screen while the user maintains uninterrupted eye contact with th~ events as they happen.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "A continuous program of human factors testing has been maintained by the project in order to build a realistic view of potential users and to measure Progress in achieving usability.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "For example, in a test of the matrix computation system with typed input, twenty-three subjects solved problems similar to those that might be assigned in a first course in programming (Biermann, Ballard, and Sigmon [7] ). In this test, the NLC system correctly processed 81 percent of the sentences and users were quite satisfied with its general performance.", |
| "cite_spans": [ |
| { |
| "start": 196, |
| "end": 204, |
| "text": "Ballard,", |
| "ref_id": null |
| }, |
| { |
| "start": 205, |
| "end": 219, |
| "text": "and Sigmon [7]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "Other tests of the system are described in Fink [14] and Geist et el.", |
| "cite_spans": [ |
| { |
| "start": 48, |
| "end": 52, |
| "text": "[14]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "[IS].", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "In another test (Fineman [13]), a simulator for a voice driven office automation system was used to obtain data on user behaviors when problem solving is with discrete and slow connetted speech.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "It was found that users quickly adapted their speech to the required discipline of slow, methodical, and simple sentences which can be recognized by machine.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "Since the data obtained in any system test is heavily dependent on the amount and kind of training given to subjects, it is necessary to have a standardlzed training procedure. In the current work, a voice tutorial has been developed for training users to use a voice interactive system (Deas [Ii] ).", |
| "cite_spans": [ |
| { |
| "start": 293, |
| "end": 297, |
| "text": "[Ii]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Fineman t G. Bilbro", |
| "sec_num": null |
| }, |
| { |
| "text": "reports on the current status of these projects with emphasis on system design, speech input facilities and their performance, the touch input system and human factors considerations. predictable, e.g. \"by\" for \"five\", \"and\" for \"add\", or \"up\" for \"of\".", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "This paper", |
| "sec_num": null |
| }, |
| { |
| "text": "We If a given word slot does not contain the correct token, the substituted word can be added to the appropriate synophone set for that subject. Thereafter, if the same substitution error recurs during a session with that subject, the correct word will be included in the synophone list for that word slot.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "SYSTEM OVERVZEW", |
| "sec_num": null |
| }, |
| { |
| "text": "The occurrence of one or more rejections in a sentence almost always results in a request for repetition.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "However, we are designing a number of facilities to handle rejections.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "In some cases, the rejected word can be determined from context, and processing can continue uninterrupted.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Otherwise, the current plan is to handle a single rejection by returning an audio response that repeats all of the sentence with the word \"what\" in place of the rejected element. The speaker will then .be able to choose to repeat the rejected word or, in case other errors are apparent, to repeat the entire utterance.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "In cases of multiple rejection errors, the speaker is requested to repeat the entire utterance.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "In all cases previous utterances will not be. discarded. Move that there and cover it.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "with a point to the object to be moved and covered.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "A pointing ability would fit in very nicely with voice driven NLC and our pro-Ject includes a touch sensitive screen so that the user can say \"double this\", point to a row, and cause the processor to double every element in that row. More complex sentences such as Add this row to that row putting the results here. (with three touchee) also become possible. accompanied by three touches.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "The strategy here is to -air touches and utterances in the order given by the user.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "In the last example all touches func-tioned to establish focus or resol~=e no,~n group reference.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "If the emphasis function of touch is mixed in, a more difficult situation arises.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "If three touches accompany Add this entry to the first row and put the result here. then the second touch was presumably to emphasize the first row or even to establish a rhythm of touching.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "In any case the facility to match touches with nondeictic expressions iS needed. If only two touches accompany this last sentence then the focusing function should take precedence, and the touches should be matched with \"this entry\" and \"here.\"", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "The situation is made even more complex by the ability to establish focus verbally.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "In NLC the user can say", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Consider row four.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Double that row. and the expression \"that row\" will refer to row four. ~f the same utterance is accompanied by a touch to a row other than four a potential conflict results. Our strategy is to give precedence to touch, since it is the more immediate focussing mechanism.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Thus the sequence Consider row four. Double that row. (touching row three) will result in the doubling of row three. (1)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Lists of words are read in tests performed by the manufacturer.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "(2)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Lists of words are read in our laboratory.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "(3)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| }, |
| { |
| "text": "Sentences are read in our laboratory. (discrete or connected) ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Re~ection Errors", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Semantic and Procedural Processing for a Natural Language Programming System", |
| "authors": [ |
| { |
| "first": "B", |
| "middle": [ |
| "W" |
| ], |
| "last": "Ballard", |
| "suffix": "" |
| } |
| ], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "B.W. Ballard, \"Semantic and Pro- cedural Processing for a Natural Language Programming System,\" Ph.D.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Understandin~ Spoken Language", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [ |
| "E" |
| ], |
| "last": "Walker", |
| "suffix": "" |
| } |
| ], |
| "year": 1978, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "D.E. Walker (ed.), Understandin~ Spo- ken Language, Elsevier North-Holland, Ne\"'wYork, 1978.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "An English Language Ouestion Answering System for a Large Relational Database", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [ |
| "L" |
| ], |
| "last": "Waltz\u00b0", |
| "suffix": "" |
| } |
| ], |
| "year": null, |
| "venue": "Communications of the ACM", |
| "volume": "21", |
| "issue": "7", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "D.L. Waltz\u00b0 \"An English Language Ouestion Answering System for a Large Relational Database,\" Communications of the ACM, Vol. 21, No. 7, pp. 526-", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF6": { |
| "num": null, |
| "type_str": "table", |
| "text": "If the stress pattern were Add this row to that row.", |
| "html": null, |
| "content": "<table><tr><td colspan=\"13\">plus a single touch, where this bears prl-</td><td>and also so that the</td><td>project</td><td>can</td><td>inter-</td></tr><tr><td>mary</td><td/><td colspan=\"2\">stress</td><td/><td>and</td><td colspan=\"2\">that</td><td colspan=\"2\">bears</td><td colspan=\"3\">secondary</td><td>nally</td><td>Judge</td><td>its</td><td>own</td><td>improvements</td><td>over</td></tr><tr><td colspan=\"9\">stress, should flnd t--e~touch</td><td colspan=\"4\">referring to</td><td>time.</td><td>In voice language</td><td>processing,</td><td>one</td></tr><tr><td colspan=\"4\">\"this row.\" with primary</td><td colspan=\"3\">stress</td><td>on</td><td>Add,</td><td colspan=\"2\">the</td><td colspan=\"2\">touch</td><td>can word and sentence error attempt to measure performance by the rates. However, experience shows that these measures are highly dependent on two factors and that almost any level of performance can be</td></tr><tr><td colspan=\"2\">would</td><td colspan=\"11\">more llkely be assoc--~-ated with that</td><td>reached if those factors are appropriately</td></tr><tr><td>row. ~w</td><td colspan=\"12\">It is unfortunate that of no voice equipment sensitive to date we</td><td>adjusted.</td><td>Those factors are</td></tr><tr><td colspan=\"7\">enough to distinguish stress patterns.</td><td colspan=\"6\">between two such</td><td>(a)</td><td>the environment and type of test within which the measurement is</td></tr><tr><td/><td colspan=\"8\">Somewhat more complicated</td><td colspan=\"3\">cases</td><td>are</td><td>made, and</td></tr><tr><td colspan=\"3\">possible=</td><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>(b}</td><td>the level</td><td>of</td><td>training</td><td>of</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>system user.</td></tr><tr><td/><td colspan=\"9\">Consider row three. ~ld this row to that row and</td><td/><td/><td>Type O~f Testln~ Environment</td></tr><tr><td/><td colspan=\"11\">put the result in the first row.</td><td>Considering (a), we tend to</td><td>classify</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>the type of test for a recognizer into one</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>of the following five</td><td>categories</td><td>and</td><td>we</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>expect</td><td>significant</td><td>differences in device</td></tr><tr><td colspan=\"4\">accompanied</td><td>by</td><td colspan=\"2\">two</td><td colspan=\"2\">touches.</td><td/><td colspan=\"2\">Since</td><td>we</td><td>response in each case.</td></tr><tr><td colspan=\"2\">allow</td><td colspan=\"3\">a touch</td><td colspan=\"8\">to occur with expressions</td></tr><tr><td colspan=\"11\">such as \"the first row,\" and since</td><td>it</td><td>is</td></tr><tr><td colspan=\"2\">possible</td><td/><td colspan=\"10\">to disregard the element in ver-</td></tr><tr><td colspan=\"13\">bal focus altogether, such a case produces</td></tr><tr><td colspan=\"7\">multiple ambiguities.</td><td colspan=\"6\">Although we foresee</td></tr><tr><td colspan=\"7\">being able to resolve</td><td colspan=\"2\">these</td><td colspan=\"4\">ambiguities</td></tr><tr><td colspan=\"4\">effectively,</td><td colspan=\"2\">and</td><td colspan=\"7\">ca~ always fall back on</td></tr><tr><td colspan=\"13\">our \"backup\" facility in case of mistakes,</td></tr><tr><td>we</td><td colspan=\"2\">also</td><td colspan=\"3\">believe</td><td colspan=\"7\">that such complex cases</td></tr><tr><td colspan=\"8\">will be extremely rare.</td><td>No</td><td colspan=\"3\">sentence</td><td>of</td></tr><tr><td>such</td><td colspan=\"12\">complexity was produced in our solu-</td></tr><tr><td colspan=\"2\">tions</td><td>to</td><td colspan=\"2\">the</td><td colspan=\"2\">eight</td><td colspan=\"6\">problems mentioned</td></tr><tr><td colspan=\"2\">above.</td><td/><td>With</td><td/><td colspan=\"8\">a voice and touch facility,</td></tr><tr><td colspan=\"13\">sentences tend to be shorter and simpler.</td></tr><tr><td/><td colspan=\"9\">NLC has implemented plurals,</td><td/><td>but</td><td>we</td></tr><tr><td>have</td><td colspan=\"2\">not</td><td colspan=\"4\">considered</td><td colspan=\"2\">their</td><td colspan=\"4\">use in touch</td></tr><tr><td colspan=\"2\">input.</td><td colspan=\"6\">Such sentences as</td><td/><td/><td/><td/><td>When both verbal and touch focus</td><td>are</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>present,</td><td>nearly</td><td>unresolvable ambiguities</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>may result.</td><td>The sequence</td></tr><tr><td/><td colspan=\"9\">Multiply these elements by</td><td/><td/></tr><tr><td/><td colspan=\"5\">this element.</td><td/><td/><td/><td/><td/><td/><td>Consider row four.</td></tr><tr><td>or</td><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>Add this row to that row.</td></tr><tr><td/><td colspan=\"8\">Add these elements up.</td><td/><td/><td/></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>accompanied by one touch,</td><td>gives</td><td>rise</td><td>to</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>the problem as to which demonstrative noun</td></tr><tr><td colspan=\"7\">with multiple touches,</td><td colspan=\"2\">would</td><td>be</td><td/><td colspan=\"2\">useful.</td><td>group to</td><td>associate</td><td>with</td><td>row</td><td>four,</td><td>and</td></tr><tr><td>In</td><td>the</td><td/><td colspan=\"2\">trial</td><td colspan=\"8\">run of eight problems, the</td><td>which</td><td>to</td><td>associate</td><td>with the touch.</td><td>One</td></tr><tr><td colspan=\"11\">introduction of plurality resulted</td><td colspan=\"2\">in up</td><td>strategy is to associate with a demonstra-</td></tr><tr><td>to</td><td colspan=\"2\">fifty</td><td colspan=\"3\">percent</td><td colspan=\"7\">reduction in number of</td><td>tive</td><td>noun</td><td>group</td><td>the touch that occurred</td></tr><tr><td colspan=\"10\">words needed and sentence length.</td><td/><td/><td>closest to the time of utterance.</td><td>Another</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>possible</td><td>strategy</td><td>is</td><td>to assume that the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>expression with that refers</td><td>to</td><td>the</td><td>more</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>distant</td><td>element</td><td>in</td><td>focus (the one esta-</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>blished</td><td>verbally</td><td>in</td><td>this</td><td>case).</td><td>This</td></tr><tr><td/><td/><td colspan=\"9\">MEASURING SYSTEM PERFORMANCE</td><td/><td>takes</td><td>advantage of the ~act that this and</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>that can be distinguished in English gram-</td></tr><tr><td/><td colspan=\"8\">Progress in any endeavor</td><td colspan=\"4\">is greatly</td><td>mar by a simple change by the feature +NEAR. iF stress Unfortunately pattern a</td></tr><tr><td colspan=\"13\">aided be measured in some meaningful way. if the level of accomplishment can It is desirable to give a figure of merit for a</td><td>speaker can undo this fairly weak regular-ity. Thus the sequence</td></tr><tr><td colspan=\"13\">system both so that a project can indicate to the world the degree of the achievement</td><td>Consider row four. ~dd th{s row to that row.</td></tr></table>" |
| }, |
| "TABREF7": { |
| "num": null, |
| "type_str": "table", |
| "text": "has been refined to the point that it could actually support user interactions in real time as we are attempting to do. many of the same technical problems arise regardless of application area. For the most part, the imperative sentence structures we are dealing with are simpler than the question forms recognized by the database systems cited above,", |
| "html": null, |
| "content": "<table><tr><td/><td/><td colspan=\"7\">In conclusion, we have at this</td><td colspan=\"2\">time</td></tr><tr><td colspan=\"7\">as the sequential parts of only fragmentary information each</td><td/><td colspan=\"3\">sentence regarding</td><td>the</td><td>recognition</td><td>system.</td><td>~he second part</td></tr><tr><td>are</td><td colspan=\"3\">voiced. what levels</td><td colspan=\"3\">Training of performance samples</td><td/><td colspan=\"2\">based on can</td><td>be</td><td>is administered very mechanically by read-</td></tr><tr><td colspan=\"11\">reading lists of vocabulary items tend be inaccurate templates for words spoken to in context. When sentences are spoken in a problem solving environment, situation (4), these effects increase and other achieved. HOwever, we have developed some tools for making measurements and will report the results as they become avail-able.</td><td>ing requesting the a tutorial tences. This introduces the system's capabilities and is specifically document to the user and utterance of trial sen-portion of the training user to the interactive Our project uses well developed speaker dependent voice recognition equip-ment with a small enough vocabulary to achieve usable accuracy rates.</td></tr><tr><td colspan=\"2\">aspects</td><td>of</td><td colspan=\"2\">word</td><td colspan=\"2\">pronunciation</td><td/><td colspan=\"2\">change.</td><td/><td>designed</td><td>to</td><td>be</td><td>administered</td><td>by</td><td>the</td></tr><tr><td colspan=\"10\">When voice control stops being the central</td><td/><td>machine.</td></tr><tr><td colspan=\"2\">concern</td><td colspan=\"8\">of the speaker, largeT variations OTRER kK)RK</td><td/></tr><tr><td colspan=\"6\">in speech are bound to occur</td><td colspan=\"2\">with</td><td colspan=\"2\">accom-</td><td/><td>Some Performance Data</td></tr><tr><td colspan=\"7\">panying larger error rates. Much of the applied work</td><td colspan=\"4\">in natural</td><td>r.Zl</td><td>An experiment was run during the sum-</td></tr><tr><td/><td colspan=\"10\">The most difficult situation language processing has concerned database of all</td><td>mer</td><td>of</td><td>1982 to obtain DP-200 performance</td></tr><tr><td colspan=\"11\">occurs query (Bronnenberg et al. in situation (5) might not even be a person Harris[17,18], Hendrix[22], C8], CoddC9], where the user who could be MylO-</td><td>data in an described</td><td>environment above. Beca~ise of</td><td>type no</td><td>(4) voice as</td></tr><tr><td colspan=\"11\">brought poUlOe[27], Plath[29], Thompson and Thomp-into a voice laboratory. In this</td><td>interactive system was</td><td>yet</td><td>available,</td><td>a</td></tr><tr><td colspan=\"3\">case, son[32], the</td><td colspan=\"8\">user Weltz[35], has only and Woods one concern, et el.</td><td>system</td><td>simulation</td><td>was</td><td>used.</td><td>After the</td></tr><tr><td colspan=\"11\">achieving the desired machine performance. Encouragement to speak carefully could met with impatience, and a few system several others have been successfully used be [36]). At least one such system is being marketed (namely INTELLECT [18]), while</td><td>the subject was placed in a room behind first part of the training session which the voice samples were collected, a in [2]</td></tr><tr><td colspan=\"11\">errors could result in even in pilot studies. (Damerau[10], F.~ly and worse speech Wescourt[12], Hershman et el. [24], quality and further degraded performance. Krause[25], Tennant[31]).</td><td>display microphone. terminal The voice tutorial with a head to the subject through a loudspeaker at mounted was read [3]</td></tr><tr><td colspan=\"11\">Our experience rates As described increase from about three to has been that word in this paper, our ini-seven percent as one moves error to each more tial work with N-LC involved programming as</td><td>the terminal introducing the of the simulated system and the types of capabilities voice commands that could be executed.</td></tr><tr><td colspan=\"11\">difficult an application area, while our more recent situation type depending on the</td><td>The</td><td>subject's commands were recognized by</td></tr><tr><td colspan=\"11\">vocabulary, t~e equipment, and other interest has shifted toward tors. Consequently, we tend to distrust office fac-any figures gathered in the easier classes of environments and attempt to do our own testing in the more difficult and more interesting situations. Most of our recent data is of type (4) and we hope to gain some type (5) experience in the com-ing year. Training the System User The second major factor affecting of training of the system user. Humans are extremely adaptive and capable of and pronoun resolution. be used in handling matters such as focus voice recognition performance is the level domains. However, as Petrick[2R] observes, while our noun phrases tend to exhibit more ela-borate structures. Furthermore, whereas typical database sy-tems process each input separately, or perhaps seek to han-semantic context as a session proceeds to dle ellipsis by consulting the immedlately preceding input, we build up a richer</td><td>Sentences are uttered in a prob-lem solving situation in our laboratory. (discrete or con-nected) Sentences are uttered in a prob-lem solving situation in the user environment. (discrete or con-nected) the DP-200 and executed by the simulation. (5) It was fo,lnd that the DP-~00 error rate rose to about 20 percent in word Thus each user command resulted in either appropriate action visible on the screen or a voice error message. In the final portion of the experiment, the subject was asked to solve an invoice pro61em that involved computing costs for a series of individual items and finding the tax and total. The experiment gave a [4] [s] reasonably accurate simulation of the expected NLC system behavior when it becomes completely voice interactive. The experiment attempted to simulate a syntactic level of voice error correction but nothing deeper. [ 6]</td></tr><tr><td colspan=\"11\">learning behaviors to perfection. Thus the designer of a voice a high degree of system might, over the years, learn to chat with it like an old friend whereas others might not be able to use the system at all. ~gain, almost any level of system performance can be observed depending on the quality of training of the user. Our approach to controlling this fac-tor has been to develop a standardized training procedure and to only report statistics on uninitiated users The most distinctive features of our present work are (a) the inclusion of voice input and output facilities, and (b) an attempt to deal with relatively \"deep\" relationships among domain objects. A more detailed discussion of the domain-independent mechanisms appears in Bier-mann[5], and as described in Ballard [2] the related LDC project being conducted in our laboratory is built around many of these techniques. Similar research pro-jects which are moving away from a fixed whose experience with the system is limited to this procedure. Ideally this procedure would be administered by machine to obtain database setting include work by Haas and Hendrix[16], Reldorn[20], Hendrix and Lewis[231, and Thompson and Thompson [33].</td><td>In the first is interested situation, in advertising the a manufac-best performance turer achievable. Tests are performed in controlled conditions with microphone placement and all system param-eters set for optimum performance, and an expert speaker is used. In our labora-tory, we are not interested in the best possible system performance but rather what we can realistically expect. The parameters are set at medium levels, there is some ambient noise, the microphone ~ay move during the test, and the user wil] be anyone we happen to bring in regardless of their speech characteristics. this test with about 14 of the 20 percent being automatically correctable. The [71 vocabulary size was 80, with three samples of most words, and six samples of a few of the difficult words, stn1=ed in the DP-200. This means that roughly every two to four sentences will have a single word error not correctable at shallow levels. This data comes from the first two hours r87 o~ usage for these subjects and we expect significant improvement as usage experi-ence increases over time. More recently, the ~LC system has become operative in a voice driven mode and subject testing has begun using the [93</td></tr><tr><td colspan=\"11\">maximum has not yet been possible. uniformity in training but this During the 197O's a number of speech understanding systems were developed under ARPA support (Lea [26], Reddy C30], Walker</td><td>As become same training procedure. soon as the sequential organized as sentences, situation words It is too early to report results but it appears that the performance predicted in the simulatiou [i0]</td></tr><tr><td colspan=\"11\">The training procedure has two parts. [34], Woods [37]) and currently some sys-</td><td>(3), the speaker begins to impose will be approximately achieved.</td><td>inflec-This</td></tr><tr><td colspan=\"11\">The tems ace being built in other first part is an informal session in countries,</td><td>tions experiment on</td><td>the will utterance that will affect include longer usage by</td></tr><tr><td colspan=\"11\">which the user is told how to speak for example [19]. Rowever, none of these indi-</td><td>recognition. the subjects and thus Certain</td><td>words indicate</td><td>may how</td><td>be much</td></tr><tr><td colspan=\"11\">vidual words to the system and examples of</td><td>stressed, and intonation may rise an~] fall error rates decrease over time.</td></tr><tr><td colspan=\"6\">the complete vocabulary are</td><td colspan=\"3\">collected</td><td colspan=\"2\">by</td><td>t89</td></tr></table>" |
| } |
| } |
| } |
| } |
