Dowding, J.; Gawron, J. M.; Appelt, D.; Bear, J; Cherny, L.; Moore, R.; and Moran, D. (1993). Gemini: A Natural Language System for Spoken-Language Understanding. In Proceedings of ARPA Workshop on Human Language Technology, Princeton, New Jersey, Morgan Kaufmann Publisher, 43-48.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....multiple asynchronous communicating processes. The core of OAA is a facilitator which manages message passing between a number of software agents that specialize in certain tasks (e.g. speech recognition) Our system uses OAA to coordinate the following five agents: 1. The Gemini NLP system [4]. Gemini uses a sin gle unification grammar both for parsing strings of words into logical forms (LFs) and for gener ating sentences from LF inputs. The use of a symbolic grammar enables us to provide precise and reliable meaning representations (in contrast to shallow statistical approaches ....

Dowding, J., J. Gawron, D. Appelt, J. Bear, L. Cherny, R.C. Moore and D. Moran. Gemini: A natural language system for spoken-language understanding. Proceedings of the ARPA Workshop on Human Language Technology(1993).

....which interacts with their natural language processing component, DELPHI, through an N best interface. DELPHI uses a definite clause grammar formalism, and a parsing algorithm which uses a statistically trained agenda to produce a single best parse for an input utterance. SRI s Gemini system [11] takes a linguistic approach to natural language understanding, in which a constituent parser applies syntactic, semantic, and lexical rules to populate a chart with edges containing syntactic, semantic, and logical form information. 1 . 2 The IBM Approach The effort at IBM has only recently ....

J. Dowding et al., "Gemini: a natural language system for spoken-language understanding," in Proceedings of the DARPA Speech and Natural Language Workshop, (Harriman, NY) February 23--26 1992.

.... The feasibility of using Grok for natural language applications has been demonstrated in [4] It has also been used to create a natural language agent which has been connected to NASA s spoken natural language dialogue system ( 27] the standard natural language agent for this system is Gemini [13]) The material presented in this paper is part of the measures we are taking towards improving Grok as a practical basis for building robust NLP systems. We do not discuss Grok in great detail here, and instead refer the interested reader to the Grok homepage (http: grok.sourceforge.net) and ....

John Dowding, Jean Mark Gawron, Doug Appelt, Lynn Cherny, Robert Moore, and Douglas Moran. Gemini: A Natural Language System for Spoken Language Understanding. In Proceedings of the 31st Annual Meeting of the Association of Computational Linguistics, Columbus, OH, 1993.

....for various subtasks in the dialogue system (see Figure 2) Speech Recogniser: a wrapper to a Nuance speech recognition server 3 using a language model compiled directly from our Gemini grammar for UAV operator dialogues. NL ( natural language ) a wrapper to the SRI Gemini parser and generator [4] using our UAV grammar (version 1) 2 All are implemented in Java, but for the NL agent (Prolog) 3 Nuance: www.nuance.com Figure 1: A demonstration of dialogue system I Multi modal Utterances Dialogue Moves Operator (O) Where are the buildings Raise question UAV (U) Here. displays ....

John Dowding, Jean Mark Gawron, Doug Appelt, John Bear, Lynn Cherny, Robert Moore, and Douglas Moran, "GEMINI: a natural language system for spoken-language understanding," in Proc. 31st Annual Meeting of the ACL, 1993.

....Other parsers have been designed specifically for robustly analyzing transcriptions of spoken language. As we have seen robustness is required both to recover from recognition errors by the SR module and to handle performance errors by the human speaker. Examples include the Gemini parser at SRI [Dowding et al. 1993], the TINA parser at MIT [Seneff, 1992] the Phoenix [Ward, 1990] and GLR [Lavie, 1994] parsers at CMU, and the Delphi parser at BBN [Stallard and Bobrow, 1992] One technique commonly applied by these robust parsers is the use of domain knowledge to limit ambiguity and to thereby enable faster ....

....The parser is all paths bottom up, uses subsumption checking, and is on line. However, it also extends the language analysis to recognize certain characteristic patterns of spoken utterances not generally included in a grammar and to recognize specific types of performance errors by the speaker [Dowding et al. 1993]. This is accomplished by an utterance grammar and utterance parser that are used to post process the chart. The SRI ATIS system incorporates this robust interpretation component that constructs queries out of grammatical fragments when an utterance cannot be analyzed as a single phrase or ....

J. Dowding, J. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran, "Gemini: A Natural Language System for Spoken-Language Understanding," In Proceedings of the ARPA Human Language Technology Workshop, pages 43--48, San Mateo, California, March 1993. ARPA, Morgan Kaufmann.

....that maximizes the posterior probability P(C S) Since the conceptual structure is represented as a hidden Markov model, the decoding of the conceptual content of a sentence can be carried out with the Viterbi algorithm. To understand a mixed language query, we cannot use statistical parsers [2,3] because it is difficult to train a parser for mixed language. The traditional direct channel models for semantic decoder, which is used in many spoken language understanding systems [4,5] cannot handle mixed language either. In addition, we cannot use independent statistical natural language ....

....results between our extended semantic decoder and the baseline semantic decoder are compared. For mixed language sentences, the average accuracy of extended semantic decoder is 26 higher on the average. Our methodology is described in section 2 in detail, and the experiments are shown in section 3. Finally, we conclude in section 4. 2. METHODOLOGY There are four main modules in the SALSA system[1] speech recognizer, spoken language understanding module, dialogue module and verbalization module. The module of mixed language recognizer converts the speech uttered by the user into text ....

[Article contains additional citation context not shown here]

John Dowding, Jean Mark Gawron, Doug Appelt, et al. GEMINI: A Natural Language System For Spokenlanguage Understanding. 31st ACL, Columbus, Ohio, June 1993, pp. 54-61

....Furthermore, as other acoustic cues are added into the model, the results should further improve. 5. COMPARISON This work expands on previous work on detecting and correcting speech repairs in Japanese. Sagawa et al. 15] proposed a parser first approach for detecting and correcting repairs (c.f. [4]) If the parser failed on an utterance, the 2 A repair is counted as correctly detected if the interruption point is correctly identified along with the type of repair. A repair is counted as properly corrected if the interruption point is correctly identified, and its editing terms and ....

J. Dowding, J. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. Gemini: A natural language system for spoken-language understanding. In Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics, pages 54--61, 1993.

....of the information, such as type of food, that she wants to retrieve. Figure 1 shows an example of a conversation that might take place in this domain. There has been some pioneering work on conversational interfaces for simple tasks in limited domains (Seneff et al. 1996; Allen et al. 1995; Dowding et al. 1993). These systems have reached the point where they are fairly robust for conversations that fit 1. U: Where should I eat today 2. S: What type of cuisine would you like 3. U: What are my choices 4. S: You can pick from types of food like Chinese, Indian, or Mexican. 5. U: Chinese. 6. S: What ....

....use. For example, the detection of a missing axiom in a proof attempt for a sub dialog initiates a dialogue interaction. However, their system is not truly mixed initiative, and is demonstrated in only one domain. Other systems for spoken language applications include (Seneff et al. 1996) and (Dowding et al. 1993). However, these systems primarily focus on the language understanding portion of conversational interfaces, and do not demonstrate them in the context of an adaptive interface. Early work on user modeling for interactive computer systems was based on predefined, hand coded knowledge about users ....

Dowding, J., Gawron, J., Appelt, D., Bear, J., Cherny, L., Moore, R., & Moran, D. (1993). Gemini: A natural language system for spoken-language understanding.

....for interruptions, as well incorporating a more general representation of the resulting plan. However, their An Adaptive Conversational Interface 15 system s initiative seems to focus on suggesting tasks that the user should next address, making it less directive than the approach we have taken. Dowding et al. 1993) and Seneff et al. 1998) have also developed conversational interfaces that give advice about air travel. Like the Place Advisor, their systems ask the user questions in order to reduce candidates, treating the choice of selecting airline flights as the interactive construction of database ....

Dowding, J., Gawron, J., Appelt, D., Bear, J., Cherny, L., Moore, R., & Moran,D. (1993). Gemini: A natural language system for spoken-language understanding.

....added. It is easy to integrate different levels of NL understanding, depending upon the requirements of the system, just by plugging in an adequate engine. The available engines are two of our low end NL systems: Nuance s template slot tools and DCG NL, a Prologbased top down parser. SRI s GEMINI [5] and FASTUS [7] are more powerful tools, used for complex NL tasks. To design the dialog on the fly, a visual tool is under development (Figure 7) It simulates the behaviors of the NL engine and creates the necessary code and data for the final NL agent. Figure 7. Visual Design Tool 4.3. Speaker ....

J. DOWDING, J.M. GAWRON, D. APPELT, J. BEAR, L. CHERNY, R. MOORE and D. MORAN, "GEMINI: A natural language system for spoken-language understanding," Proc. of 31st Annual Meeting of the Association for Computational Linguistics (ACL'96): Columbus, USA, 1996.

....errors and characteristics of spontaneous speech. In research today, it is even far to be solved. A credible alternative would consist in introducing as soon as possible morphological, syntactical and semantical knowledge in the decoding process. 1.1. What recognition output Several works [8, 18, 16, 9, 17, 4] show that the best acoustic hypothesis is often insufficient. They emphasise it is necessary to keep the N best hypotheses or the word graph, to perform complementary processing and to improve the recognition rate. Indeed, a grammatical approach may become essential as soon as the application is ....

....are complex, i.e. when the user can ask a spontaneous request and not only give a sequence of keywords, this kind of analysis seems to be insufficient. If we want to introduce understanding, or to be able to lead an interaction, it is necessary to turn towards a complete linguistic analysis [18, 4]. Yet, it is not always possible to develop an exhaustive grammar in documentary base interrogation. 1.3. Spoken query analysis In the context of a document retrieval application, any grammar of queried topics should be very large and complex, since it should include the language of all the ....

[Article contains additional citation context not shown here]

Dowding J., Gawron JM., Appelt D., Bear J., Cherny L., Moore R. et Moran D. 1993. GEMINI : A natural language system for spoken language understanding. In proceedings of the 31 st annual meeting of the Association for Computational Linguistics, ACL'93, 54-61.

.... based on Prolog Definite Clause Grammars) However, in most settings, one wants to use the most powerful, flexible approaches available, and thus our efforts have been focused on the use of two very sophisticated systems developed at SRI: the Decipher [3] speech recognition system, and the Gemini [5] natural language understanding system, both of which have been used as agents in a number of OAA based systems. Consequently, the requirements for LEAP have largely been driven by these Figure 1: Using ProACT to generate source code for an agent. two systems. Although the Speech Recognition ....

J. Dowding, J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. Gemini: A natural language system for spoken-language understanding. In Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics, pages 54--61, Columbus, Ohio, June 1993.

....In addition to the Swedish version, there are versions for Japanese [Kameyama 1995] French and Spanish [Rayner et al. 1996] CLE grammars for small portions of German, Korean and Polish have also been designed. Most notably, there is a second English CLE, from SRI AI Center, Menlo Park, California [Dowding et al. 1993]. It started out as a version of the Cambridge CLE, but has now developed quite far away from the original. Just like the SLE, the French and Spanish versions are based on the Cambridge CLE, while Megumi Kameyama s Japanese Language Engine stems from the Menlo Park system. When the text in the ....

John Dowding, Jean Mark Gawron, Douglas Appelt, John Bear, Lynn Cherny, Robert Moore, and Douglas Moran. "Gemini: A Natural Language System for Spoken-Language Understanding". In Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics, pp. 54--61, Columbus, Ohio, 1993. ACL. BIBLIOGRAPHY 222

....the example demonstrates, look very text like. The main exception is the presence of speech repairs, which Bear, Dowding, and Shriberg (1992) report happen in 10 of sentences longer than nine words. Good results in understanding have been achieved in this domain, as evidenced by the results of Dowding et al. 1993), who report syntactic and semantic understanding of 86 of all utterances. So, this task gives the allusion that spontaneous speech understanding corresponds to our intuitions about understanding text. However, such spoken queries to not capture the problems inherit in natural dialogues. Unlike ....

....repair replaces text of the same category. However, Hindle assumed an edit signal would mark the interruption point, a signal that has yet to be found. Another approach, taken by Bear, Dowding, and Shriberg (1992) uses a pattern matcher to look for patterns of matching words. In related work, Dowding et al. 1993) employed a parser first approach. If the parser and semantic analyzer are unable to make sense of an utterance, they look for speech repairs using the pattern matcher just mentioned. A number of researchers have also looked at intonational characteristics of speech repairs (O Shaughnessy, 1992; ....

[Article contains additional citation context not shown here]

Dowding, John, Jean Mark Gawron, Doug Appelt, John Bear, Lynn Cherny, Robert Moore, and Douglas Moran. 1993. Gemini: A natural language system for spoken-language understanding. In Proceedings of the 31 th Annual Meeting of the Association for Computational Linguistics, pages 54--61.

....is the simplest of those integrated into the OAA. It is written in Prolog using Definite Clause Grammars, and supports a distributed vocabulary; each agent dynamically adds word definitions as it connects to the network. A current project is underway to integrate the Gemini natural language system [8], a robust bottom up parser and semantic interpreter specifically designed for use in Spoken Language Understanding projects. Database Agents: Database agents can reside at local or remote locations and can be grouped hierarchically according to content. Micro agents can be connected to ....

Dowding, J., Gawron, J.M., Appelt, D., Bear, J., Cherny, L., Moore, B. and Moran D., "Gemini: A natural language system for spoken-language understanding", Technical Note 527, AI Center, SRI International, 333 Ravenswood Ave., Menlo Park, CA 94025, April 1993.

....must address the problem of recognition errors, utterance disfluencies and lack of cover of the interpretation process. 1.1. What recognition output As far as recognition errors are concerned, most experimentations emphasize the need of using more than the first best acoustic hypothesis [6, 17, 15, 7, 16, 3], so that either N best hypotheses or word graphs should be kept for further processing. 1.2. Grammar for post processing A critical phase is to retrieve the utterance meaning in the context of the application. A straightforward approach to achieve robustness is to focus only on informative ....

....some information in the sentence like negation and its scope and is strongly misled by errors on informative words. A more controlled robustness, and also an increased recognition rate, is achieved when attempting a full span linguistic analysis, backed off by optimal island analysis [17, 15, 7, 16, 3]. However, grammatical approaches require very significant development efforts 1.3. Our specific case In the context of a document retrieval application, any grammar of queried topics should be very large and complex, since it should include the language of all the documents. It is thus illusory ....

[Article contains additional citation context not shown here]

J. Dowding, JM. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. GEMINI: A natural language system for spoken language understanding. In proceedings of the 31 st annual meeting of the Association for Computational Linguistics, ACL'93, pages 54--61, June 1993.

....added. It is easy to integrate different levels of NL understanding, depending upon the requirements of the system, just by plugging in an adequate engine. The available engines are two of our low end NL systems: Nuance s template slot tools and DCG NL, a Prolog based top down parser. SRI s GEMINI [6] and FASTUS [8] are more powerful tools, used for complex NL tasks. 3.3 Pen Recognition and Annotation Pen modality is used on different supports such as a map or live video to enter data such as handwriting and gestures. The handwriting recognizer is provided by Communications Intelligence ....

Dowding, J., Gawron, J., M. Appelt, D., Bear, J., Cherny, L., Moore, R. and Moran, D. GEMINI: A natural language system for spoken-language understanding. 31st Annual Meeting of the Association for Computational Linguistics. Pp. 54-61. Colombus, USA, 1996

....by the end of the string. Such an algorithm could precede a parser, or even operate in lockstep with it. An ulterior motive for not using higher level syntactic or semantic knowledge is that the coverage of parsers and semantic interpreters is not sufficient for unrestricted dialogs. Recently, Dowding et al. 1993) reported syntactic and semantic coverage of 86 for the DARPA Airline reservation corpus. Unrestricted dialogs will present even more difficulties; not only will the speech be less grammatical, but there is also the problem of segmenting the dialog into utterance units (cf. Wang and Hirschberg, ....

....correction recall of 43 and correction precision of 50 . They also tried combining syntactic and semantic knowledge in a parser first approach first try to parse the input and if that fails, invoke repair strategies based on word patterns in the input. In a test set containing 26 repairs (Dowding et al. 1993), they obtained a detection recall rate of 42 and a precision of 84.6 ; for correction, they obtained a recall rate of 30 and a recall rate of 62 . Nakatani and Hirschberg (1993) investigated using acoustic information to detect the interruption point of speech repairs. In their corpus, 74 of ....

[Article contains additional citation context not shown here]

Dowding, John, Jean Mark Gawron, Doug Appelt, John Bear, Lynn Cherny, Robert Moore, and Douglas Moran. 1993. Gemini: A natural language system for spoken-language understanding. In Proceedings of the 31 th Annual Meeting of the Association for Computational Linguistics, pages 54--61.

....the recognition process by recovering correct hypotheses. The previous experiments suggest to exploit the grammar ability to identify recognition outputs that need further processing. The triggered process Nbest ranking or lattice decoding might in turn exploit linguistic knowledge as in [8, 6, 20, 19]. 2.2 Robustness Requirement The parsing of SR hypothesis is faced with utterance disfluencies, lack of grammatical cover and recognition errors substitutions, deletions, insertions. To achieve its goals, the grammatical parsing is thus required to accept main phenomena of spoken language, ....

....and grammatical extensions [15, 9] The latter strategies use several systematic relaxation rules. Dialogue transcription parsing strategies also use relaxation in their parses [7, 3] but in a less systematic way, in fact, only word skip is allowed. Word graph (or N best) parsing strategies [20, 8] may use the word skip relaxation, but they restrict it further by applying it only after all partial parses have been built, so that word skip is allowed between two grammatical categories but not within a grammatical category. The relaxation stage can be seen, in a dual manner, as a way to put ....

J. Dowding, JM. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. GEMINI: a natural language system for spoken language understanding. In proceedings of the 31 st annual meeting of the Association for Computational Linguistics, ACL'93, June 1993.

.... in a number of speech understanding systems for the airline reservation task, including Carnegie Mellon s ATIS system based on the Phoenix 14 grammar and robust parser [Ward91] MIT s Voyager system is a speech understanding system for a tourist s city guide task [Zue90] SRI s GEMINI system [Dowding93] is one of the natural language understanding components of the Open Agent Architecture (see section 2.4.4) Strictly speaking, the JANUS speech translation system [Waibel91] Waibel96b] is not exactly a speech understanding system. However, JANUS translates between languages by converting ....

Dowding, J., Gawron, J.M., Appelt, D., Bear, J., Cherny, L., Moore, R., and Moran, D. GEMINI: A Natural Language System for Spoken Language Understanding. In Proceedings of Annual Meeting of the Association for Computational Linguistics, 54-61. (Columbus, OH) June 1993. 175

....speech recognition errors. On the other hand, the good thing is that spoken language tend to contain less complex structures than written language. Several methods have been suggested compensate for these speech related problems: e.g. score and penalties, probabilistic rules, and skipping words [5, 15, 11, 8]. A small community have experimented with either purely statistical approaches[2, 14] or connectionist based approaches [1, 12, 9, 16] Their main advantages are learnability and robustness. All connectionist approaches to our knowledge, have suffered from one or more of the following problems: ....

J. Dowding, J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. Gemini: A Natural Language System for Spoken-Language Understanding. In Proceedings ARPA Workshop on Human Language Technology, pages 43--48, Princeton, New Jersey, March 1993. Morgan Kaufmann Publisher.

....abridged repairs from modification repairs, which should make both types of repairs easier to correct. An ulterior motive for not using higher level syntactic or semantic knowledge is that the coverage of parsers and semantic interpreters is not sufficient for unrestricted dialogs. Recently, Dowding et al. 1993) reported syntactic and semantic coverage of 86 for the Darpa Airline reservation corpus. Unrestricted dialogs will present even more difficulties; not only will the speech be more ungrammatical, but there is also the problem of segmenting the dialog into utterance units (c.f. Wang and ....

....of 43 and correction precision of 50 . They also tried combining syntactic and semantic knowledge in a parser first approach first try to parse the input and if that fails, invoke repair strategies basedon their pattern matching technique. In a test set of 756 utterances containing 26 repairs (Dowding et al. 1993), they obtained a detection recall rate of 42 and a precision of 84.6 ; for correction, they obtained a recall rate of 30 and a precision rate of 62 . Nakatani and Hirschberg (1993) investigated using acoustic information to detect the interruption point of speech repairs. In their corpus, 74 ....

[Article contains additional citation context not shown here]

Dowding, J., Gawron, J. M., Appelt, D., Bear, J., Cherny, L., Moore, R., and Moran, D. (1993). Gemini: A natural language system for spoken-language understanding. In Proceedings of the 31 th Annual Meeting of the Association for Computational Linguistics, pages 54--61.

....provide the best parse or the N best parses. Therefore, a mixture of hard and soft rules (scores and penalties, probabilistic rules, and constraints) is applied. In most parsers, the core consists of hand modeled rules. With great success, these rules have been annotated with soft information[3, 10, 8, 6]. In this paper, we present a parser, FeasPar, that learns to parse, instead of having hand modeled rules. The FeasPar architecture consists of neural networks and a search. The search finds the best feature structure based on the neural network outputs, and feature structure constraints. FeasPar ....

J. Dowding, J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran, `Gemini: A Natural Language System for Spoken-Language Understanding', in Proceedings ARPA Workshop on Human Language Technology, pp. 43-- 48, Princeton, New Jersey, (March 1993). Morgan Kaufmann Publisher.

No context found.

Dowding, J.; Gawron, J. M.; Appelt, D.; Bear, J.; Cherny, L.; Moore, R.; and Moran, D. 1993. GEMINI: a natural language system for spoken-language understanding. In Proceedings 31st Annual Meeting of the ACL.

No context found.

Dowding, J.; Gawron, J.; Appelt, D.; Bear, J.; Cherny, L.; Moore, R.; and Moran, D. 1993. Gemini: a natural language system for spoken-language understanding. In Proc.

.... of the SRI CommandTalk system [10,14] The system comprises a suite of about 20 agents, connected together using the SRI Open Agent Architecture [9] Speech recognition is performed using a version of the Nuance recognizer[12] Initial language processing is carried out using the SRI Gemini system [6], using a domainindependent unification grammar and a domain specific lexicon. The language processing grammar is compiled into a recognition grammar using the methods of [7] the net result is that only grammatically well formed utterances can be recognized. Output from the initial ....

Dowding, J., Gawron, M., Appelt, D., Cherny, L., Moore, R. and Moran, D. 1993. Gemini: A natural language system for spoken language understanding. In Proceedings for the Thirty-First Annual Meeting of the Association for Computational Linguistics.

....collect a sufficient corpus for training a statistical language model can be a significant problem. For CommandTalk, we did not create a statistical language model. Instead, with information gathered from interviews of subject matter experts (SME s) we developed a handwritten grammar using Gemini (Dowding et al. 1993), a unification based grammar formalism. We used this unification grammar for both natural language understanding and generation, and, using a grammar compiler we developed, compiled it into a context free form suitable for the speech recognizer as well. The effects of this single grammar ....

J. Dowding, J. Gawron, D. Appelt, L. Cherny, R. Moore, and D. Moran. 1993. Gemini: A Natural Language System for Spoken Language Understanding.

....of CommandTalk are: ffl Speech Recognition based on Nuance, a commercial continuous speech speaker independent speech recognizer developed from SRI technology. Its language model is compiled [5] from the Gemini grammar used by the Natural Language agent. ffl Natural Language based on Gemini [2], performs syntactic and semantic analysis [3] and text generation [7] ffl Contextual Interpretation (CI) This agent is described in detail in section 3. ffl Speech Synthesis based on the Festival speech synthesizer 1 developed by the Centre for Speech Technology Research (CSTR) at the ....

John Dowding, Jean Mark Gawron, Doug Appelt, Lynn Cherny, Robert Moore, and Douglas Moran. Gemini: A Natural Language System for Spoken Language Understanding. In Proceedings of the Thirty-First Annual Meeting of the ACL, Columbus, OH, 1993. Association for Computational Linguistics.

....that can be parsed in sync; that is, it guarantees that every word string that can be parsed by the natural language component is a potential recognition hypothesis, and vice versa. 2. 2 Natural Language The natural language (NL) agent consists of a thin agent layer on top of SRI s Gemini (Dowding et al. 1993, 1994) a natural language parsing and semantic interpretation system based on unification grammar. Unification grammar means that grammatical categories incorporate features that can be assigned values; so that when 4 grammatical category expressions are matched in the course of parsing or ....

Dowding, J., J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran (1993) "Gemini: A Natural Language System for Spoken-Language Understanding, " in Proceedings 31st Annual Meeting of the Association for Computational Linguistics, Columbus, Ohio, pp. 54--61.

....approximately 23,000 utterances of ATIS spontaneous speech data. 4. NATURAL LANGUAGE UNDERSTANDING The natural language understanding component accepts word hypotheses from the recognizer, and produces two outputs, the simplified logical form (SLF) and the answer to the query. The Gemini system [3, 4] has been incorporated into this application without significant modification. Gemini uses a word synchronous bottom up parser to process speaker utterances. An implication of this approach is that it is possible to start syntactic and semantic processing of the utterance before the entire ....

J. Dowding, J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran, "Gemini: A Natural Language System for Spoken-Language Understanding, " in Proceedings ARPA Workshop on Human Language Technology, Merrill Lynch Conference Center, Princeton, New Jersey, pp. 43--48 (21--24 March 1993).

....response to a linguistic act, in general the world can change for a variety of reasons, including the operator s activities on the GUI or the activities of other operators. 3 Language Interpretation and Generation The language used in CommandTalk is derived from a single grammar using Gemini (Dowding et al. 1993), a unification based grammar formalism. This grammar is used to provide all the language modeling capabilities of the system, including the language model used in the speech recognizer, the syntactic and semantic interpretation of user utterances (Dowding et al. 1994) and the generation of ....

J. Dowding, J. Gawron, D. Appelt, L. Cherny, R. Moore, and D. Moran. 1993. Gemini: A Natural Language System for Spoken Language Understanding. In Proceedings of the Thirty-First Annual Meeting of the ACL, Columbus, OH. Association for Computational Linguistics.

....even the termination, of the processing strategy) The following example shows a rule written in the unification based formalism used in the Core Language Engine (Alshawi, 1992) developed by SRI International s research group in Cambridge, England. This formalism is also used in the Gemini system (Dowding et al. 1993, 1994) developed by SRI in Menlo Park, California: S: tensed=yes] NP: person=P,num=N] VP: tensed=yes,person=P,num=N] What unification grammar adds to context free grammar is the notion of feature constraints. In the notation presented here, grammatical categories are specified in terms of a ....

Dowding, J., J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran (1993) "Gemini: A Natural Language System for Spoken-Language Understanding," in Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics, Columbus, Ohio, pp. 54--61.

....spoken language interface that allows commanders to use natural, spoken English commands to control simulated forces. SRI s Open Agent Architecture (OAA, Fig. 2, 2] was used to rapidly put the needed pieces together, i.e. the Nuance speech recognizer [3] the Gemini natural language (NL) parsing [4] and interpreting system and the ModSAF simulator, in order to make them talk to each other. Figure 2: OAA configuration for ModSAF 3. Experience We also developed, using OAA, another map based system, Multimodal Maps (MMAP, 5] which provides, in addition to the speech, other natural modalities ....

J. Dowding, J.M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore and D. Moran. Gemini: A Natural Language System for Spoken-Language Understanding. 31 st Annual meeting of the Association for Computational Linguistics, Columbus, Ohio, 1993.

....it provides simple mix and match for the components. In developing systems, we have used three different natural language (NL) systems: a simple one, based on Prolog DCG (Definite Clause Grammar) then an intermediate one, based on CHAT [16] and finally, our most capable research system GEMINI [6, 7]. The ability to trivially substitute one natural language agent for another has been very useful in rapid prototyping of systems. The DCG based agent is used during the early stages of development because grammars are easily written and modified. Writing grammars for the more sophisticated NL ....

....the utility of the concept, a more extensive analysis was conducted of the task and the commands used, and more capable prototypes were developed. One of the significant enhancements was the replacement of our simplest natural language agent (DCG based) with our most sophisticated (based on GEMINI [6, 7]) Summarization of Conversation A system that summarizes conversations provided a novel opportunity to use two instances of a speech recognition agent, in conjunction with a single instance of a text processing agent ( 10] In this system, MIMI, two Japanese speakers engage in a conversation, ....

John Dowding, J. Mark Gawron, Douglas Appelt, John Bear, Lynn Cherny, Robert Moore, and Douglas Moran. GEMINI: A natural language system for spokenlanguage understanding. In Proc. of the 31st Annual Meeting of the Association for Computational Linguistics, pages 54--61, Ohio State University, Columbus, Ohio, 22--26 June 1993.

No context found.

Dowding, J.; Gawron, J. M.; Appelt, D.; Bear, J; Cherny, L.; Moore, R.; and Moran, D. (1993). Gemini: A Natural Language System for Spoken-Language Understanding. In Proceedings of ARPA Workshop on Human Language Technology, Princeton, New Jersey, Morgan Kaufmann Publisher, 43-48.

No context found.

J. Dowding, M. Gawron, D. Appelt, L. Cherny, R. Moore, and D. Moran. 1993. Gemini: A natural language system for spoken language understanding. In Proceedings of the Thirty-First Annual Meeting of the Association for Computational Linguistics.

No context found.

J. Dowding, J.M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore and D. Moran, "Gemini: A Natural Language System for SpokenLanguage Understanding," ACL '93, 54-61.

No context found.

J. Dowding, J.M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. 1993. Gemini: a natural language system for spoken-language understanding. In Proc. ACL 93.

No context found.

Dowding, J., Gawron, J.M., Appelt, D., Bear, J., Cherny, L., Moore, R., Moran, D.: Gemini: A natural language system for spoken language understanding. In: Proc. 31st Annual Meeting of the Association for Computational Linguistics. (1993)

No context found.

J. Dowding, J. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran, "Gemini: A Natural Language System for Spoken Language Understanding," Proc. ARPA Workshop on Human Language Technology, 21--24, 1993.

No context found.

J. Dowding, J. M. Gawron, D. Appelt, and J. Bear. 1994. Gemini: A natural language system for spoken-language understanding. In Proc. ARPA Human Language Technology Workshop '93, pages 43--48, Princeton, NJ.

No context found.

J. Dowding, J. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, , and D. Moran. Gemini: A natural language system for spoken-language understanding. Communications of the ACM, 39(1):51 -- 87, 1993.

No context found.

J. Dowding, J. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. GEMINI: A natural language system for spoken-language understanding. In ACL-93, pages 54--61, 1993.

No context found.

J. Dowding, J. M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran. Gemini: A Natural Language System for Spoken-Language Understanding. In Proceedings ARPA Workshop on Human Language Technology, pages 43#48, Princeton, New Jersey, March 1993. Morgan Kaufmann Publisher.

No context found.

J. Dowding, J.M. Gawron, D. Appelt, J. Bear, L. Cherny, R. Moore, and D. Moran: "GEMINI: A natural language system for spoken-language understanding", (http://www.ai.sri.com/pubs/papers/Dowd93-54:GEMINI/document.ps. gz), 1993.

No context found.

John Dowding, J. Mark Gawron, Douglas Appelt, John Bear, Lynn Cherny, Robert Moore, and Douglas Moran. GEMINI: A natural language system for spoken-language understanding. pages 54-61, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC