Reithinger N. and Maier E.: Utilizing Statistical Dialogue Act Processing in Verbmobil, ACL-95, pp. 116-121, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....language systems is usually motivated by arguments concerning the system s robustness: 1) support for speech recognition and (2) providing necessary context for higher level dialogue management. Previous research has taken discourse information into account mainly in the form of speech acts [8, 10, 7], which seem to provide a suitable basis for dialogue models: they have been widely studied, and dealing with the speaker s intentions, they can be considered domain independent. However, speakers do not only produce intentionally linked utterances: they also aim at maintaining thematic coherence. ....

N. Reithinger and E. Maier. Utilizing statistical dialogue act processing in Verbmobil. In Proceedings of the 33rd Annual Meeting of the ACL, pp. 116--121, 1995.

....in this project, utterances are produced by actions that are executed with the goal of having some particular effect on the hearer. These actions are called speech acts. The task of an understanding system is to recognize which speech acts the speaker is performing with his or her utterances (Reithinger Maier 1995; Ros e et al. 1995; Wiebe et al. 1996) Consider the utterance 2 to 4 (dos a cuatro) a common type of utterance in the scheduling dialogs. The speaker might be suggesting that they meet from 2 to 4; they might be confirming that 2 to 4 is the time currently being discussed; they might, with ....

Reithinger, N. et al. 1995. Utilizing statistical dialogue act processing in verbmobil. In ACL, 116--122.

....in this project, utterances are produced by actions that are executed with the goal of having some particular effect on the hearer. These actions are called speech acts. The task of an understanding system is to recognize which speech acts the speaker is performing with his or her utterances (Reithinger Maier 1995; Ros e et al. 1995; Wiebe et al. 1996) Consider the utterance 2 to 4 (dos a cuatro) a common type of utterance in the scheduling dialogs. The speaker might be suggesting that they meet from 2 to 4; they might be confirming that 2 to 4 is the time currently being discussed; they might, with ....

Reithinger, N., and Maier, E. 1995. Utilizing statistical dialogue act processing in verbmobil. In Proceedings ACL, 116--122.

.... and verb argument structure (Brent, 1993) Recently, we have begun to see the first inkling of research on learning within the understanding agency: primitive methods have been described for learning make sense rules (Knight and Hatzivassiloglou, 1995) and for learning scripts (Miikkulainen, 1993; Reithinger and Maier, 1995). The closer we get to the understanding agency, the more training data and processing power we need in order to learn rules automatically. Thus, for English part of speech tagging, all the rules have basically been learned automatically. For verb argument structure, only a subset of the actual ....

Reithinger, N., and Maier, E. (1995). Utilizing statistical dialogue act processing in Verbmobil. In 33rd annual meeting of the Association for Computational Linguistics. pp. 116-121. San Francisco: Morgan Kaufmann.

....Whittaker and Stanton [19] 7. Empirical analysis of dialogue corpora to produce statistical measure of dialogue turn transitions based on a taxonomy of dialogue actsthe first empirical pragmatics In some ways, the provision of evidence for forms of (6) and so (4) e.g. Maier and Reithinger [20, 21]) Pattern matching approaches to bottom up dialogue analysis, providing input to some higher representational form and rejecting the possibility of effective dialogue grammar (Colby [22] ATIS [23] and CONVERSE [24] 8. Full treatment of empirical dialogue analysis, derived automatically from ....

N Reithinger and E Maier. Utilizing Statistical Dialogue Act Processing in Verbmobil. In Proceedings of ACL-95, 1995.

....mainly. Since the prediction of ten from a total number of seventeen is not restrictive enough, a statistical prediction method is used. The statistical module makes use of deleted interpolation to compute the probability of a sequence of dialogue acts to narrow down the analysis search [Reithinger and Maier 95] As reported in [Reithinger 96] the most significant results are obtained when the dialogue acts are annotated with information on the speaker: a reject by Speaker A to a question posed by A might be a clarification, explanation or correction; a reject by Speaker B would be a rejection of the ....

N. Reithinger and E. Maier. Utilizing Statistical Dialogue Act Processing in Verbmobil. In Proceedings of ACL, 1995.

....act recognition in Section 4. Lastly, we evaluate the effects of the proposed approaches on reducing cumulative error. 2 Related Work There has been much recent work on building a representation of the discourse context with a plan based or finite state automaton based discourse processor [1, 9, 3, 7, 8, 5]. Of these, the Verbmobil discourse processor [7] and our Enthusiast discourse processor are designed to be used in a wide coverage, large scale, spontaneous speech system. In these systems, the design of the dialogue model, whether plan based or a finite state machine, is grounded in a corpus ....

....of the proposed approaches on reducing cumulative error. 2 Related Work There has been much recent work on building a representation of the discourse context with a plan based or finite state automaton based discourse processor [1, 9, 3, 7, 8, 5] Of these, the Verbmobil discourse processor [7] and our Enthusiast discourse processor are designed to be used in a wide coverage, large scale, spontaneous speech system. In these systems, the design of the dialogue model, whether plan based or a finite state machine, is grounded in a corpus study that identifies the standard dialogue act ....

[Article contains additional citation context not shown here]

N. Reithinger and E. Maier. Utilizing Statistical Dialogue Act Processing in Verbmobil. In Proceedings of the ACL, 1995.

....Apart from rendering the system anything but user friendly, this approach leaves no room for user initiative in the course of the dialogue. The alternative approach of automatically discovering n grams of speech acts on the basis of corpora (e.g. Moeller, 1997; Passonneau and Litman, 1997; Reithinger and Maier, 1995 ] is by necessity too domainand corpus dependent to be of interest for our goals. The point is that the DMan should not have to be restructured or have more than one of its modules respecified every time it is customised to a new application and domain. Thus, a plausible answer seems to be the ....

N. Reithinger and E. Maier. Utilizing statistical dialogue act processing in VERBMOBIL. Technical Report VM Report 80, DFKI GmbH, Saarbruecken, Germany, May 1995.

....to particular constructs, and concerned with the maintenance of communication norms. The second line of work which can be seen as addressing this particular defect of speech act theory is the Conversational Games tradition: Power (1979) Houghton (1986) Kowtko, Isard, and Doherty (1992) Reithinger and Maier (1996). More of a descriptive framework than a theory, this tradition posits a set of conversational games or dialogue games each consisting of a set of moves, where an utterance may realise one or more moves. The important thing is that the games encompass both partners in dialogue: for example, a ....

....to the questions of how many games there are, how they vary according to the type of dialogue, and what constraints there are upon possible games These are the kinds of question routinely asked of every other level of linguistic formalism. For example, in the Verbmobil system as described in Reithinger and Maier (1996), games of much finer level of detail than in the Map Task are envisaged: e.g. arranging a time , or confirming a date . These are justified along exactly the same lines as those developed for the Map task, namely, intuitive agreement that a certain level of commonality exists between different ....

[Article contains additional citation context not shown here]

Reithinger, N. and E. Maier (1996). Utilizing statistical dialogue act processing in verbmobil. In Proceedings of 33rd ACL, Cambridge Mass., pp. 116--121.

....even the baseline result has to be considered a lower bound for the expected error in real conditions. In the Verbmobil project, a problem similar to that presented here is faced. In fact, a shallow translation is based on a set of DAs to be assigned to each input sentence. Interestingly, in [9] it is reported that the N gram based approach outperformed a more complex dialogue model, based on a finite state automaton. The results presented on the prediction power of N grams can be compared with those depicted in Figure 1. The adopted experimental set up is very similar to that ....

N. Reithinger and E. Maier. Utilizing Statistical Dialogue Act Processing in Verbmobil. In Proc. of Annual Meeting of the ACL, 1995.

....and suggesting that a n gram dialog act model may be the best approach for predicting DAMSL tags. DAMSL Searle s taxonomy of illocutionary acts (Searle 1975) has been highly influential on the set of dialog acts recognized by computational systems. Two examples are the VERBMOBIL dialog system (Reithinger Maier 1995) and the TRAINS dialog system (Ferguson, Allen, Miller 1996) Both have dialog acts fitting under Searle s Directives label such as Request Suggest in VERBMOBIL and Request in TRAINS. However, with other dialog acts such as VERBMOBIL s Digress act or TRAIN s Acknowledge Apology it is difficult ....

Reithinger, N., and Maier, E. 1995. Utilizing statistical dialogue act processing in verbmobil. In Proc. of the 33rd annual meeting of the Association for Computational Linguistics (ACL-95).

....a conflict . Disagreement . Suggestions of locations . Acknowledgement and assertion of a conflict . Demand for scheduling an appointment . A first time suggestion . Acknowledgement . Turns without any domain contribution Figure 4: Learned DDA classes reported by Reithinger and Maier [20]. They report a hit rate for the first prediction of 29 and 45 with the first and second together when considering every turn in the data. For learning we used again 187 dialogues with 4521 turns. Prediction rates based on an unseen test set of 79 dialogues with 1495 turns are given in Figure 5. ....

N. Reithinger and E. Maier "Utilizing Statistical Dialogue Act Processing in VERBMOBIL", Verbmobil VMReport

....and communicative obligations, not specific to particular constructs, and concerned with the maintenance of communication norms. The second line of work which can be seen as addressing this particular defect of speech act theory is the Conversational Games tradition: Pow79] Hou86] KID92] RM96] More of a descriptive framework than a theory, this tradition posits a set of conversational games or dialogue games each consisting of a set of moves, where an utterance may realise one or more moves. The important thing is that the games encompass both partners in dialogue: for example, a ....

....to the questions of how many games there are, how they vary according to the type of dialogue, and what constraints there are upon possible games These are the kinds of question routinely asked of every other level of linguistic formalism. For example, in the Verbmobil system as described in [RM96] games of much finer level of detail than in the Map Task are envisaged: e.g. arranging a time , or confirming a date . These are justified along exactly the same lines as those developed for the Map task, namely, intuitive agreement that a certain level of commonality exists between different ....

[Article contains additional citation context not shown here]

N. Reithinger and E. Maier. Utilizing statistical dialogue act processing in verbmobil. In Proceedings of 33rd ACL, Cambridge Mass., pages 116--121, 1996.

....sources. The finite state approach provides a fast and efficient alternative to the more time consuming plan based approach. Currently, the two discourse processors are used separately. We intend to combine these two approaches with a layered architecture, similar to the one proposed for Verbmobil [6], in which the finite state machine would constitute a lower layer providing an efficient way of recognizing speech acts, while the plan based discourse processor, at a higher layer, would be used to handle more knowledge intensive processes, such as recognizing doubt or clarification ....

N. Reithinger and E. Maier. Utilizing statistical dialogue act processing in Verbmobil. In Proceedings of the ACL, 1995.

....against representing such sequences as speech acts; however, as in the computational work cited above, we have used notion of discourse level speech act to represent the functional relationship between the surface form of an utterance, the context, and the attitudes expressed by the speaker. 9 Reithinger and Maier (1995) have used n gram dialogue act probabilities to induce the adjacency pairs from a corpus of dialogues for appointment scheduling. 10 Communication can occur despite such differences because speakers with similar linguistic experiences presumably will develop similar expectations about how ....

....higher priority on the assumption that memory for suppositions is stronger than expectation. 32 It is possible that the same surface form might accomplish several different discourse acts, in which case it might be desirable to evaluate the likelihood of alternative choices. The work discussed by Reithinger and Maier (1995), for example, found statistical regularities in the misinterpretations that occurred in their corpus of appointment scheduling dialogues. McRoy and Hirst The Repair of Misunderstandings The defaults that characterize misunderstandings have a lower priority than the metaplans, because speakers ....

Reithinger, Norbert and Elisabeth Maier. 1995. Utilizing statistical dialogue act processing in verbmobil. In Proceedings, 33rd annual meeting of the Association for Computational Linguistics, pages 116--121, Cambridge, MA, June.

....when an incorrect hypothesis is chosen and incorporated into the context, thus providing an inaccurate context from which subsequent context based predictions are made. For example, in Enthusiast, we model the discourse context using speech acts to represent the functions of dialogue utterances [1, 3, 6, 7]. Speech act selection is strongly related to the task of determining how the current input utterance relates to the discourse context. When, for instance, a plan based discourse processor is used to recognize speech acts, the discourse processor computes a chain of inferences for the current ....

....speech act recognition in section 4. Lastly, we evaluate the effects of the proposed approaches on reducing cumulative error. 2 Related Work There has been much recent work on building a representation of the discourse context with a plan based or finite state automaton based discourse processor [1, 10, 3, 6, 8, 5]. Of these, the Verbmobil discourse processor [6] and our Enthusiast discourse processor are designed to be used in a wide coverage, large scale, spontaneous speech system. In these systems, the design of the dialogue model, whether plan based or a finite state machine, is grounded in a corpus ....

[Article contains additional citation context not shown here]

N. Reithinger and E. Maier. Utilizing statistical dialogue act processing in Verbmobil. In Proceedings of the ACL, 1995.

....those from the non context based parser approach. In developing our discourse processor for disambiguation we needed to address three major issues. First, most plan based or finite state automaton based discourse processors (Allen and Schubert, 1991; Smith, Hipp, and Biermann, 1995; Lambert, 1993; Reithinger and Maier, 1995), including the one we initially developed (Ros e et al. 1995) only take one semantic representation as input at a time: thus, we had to extend the discourse processor so that it can handle multiple hypotheses as input. Secondly, we needed to quantify the disambiguating predictions made by the ....

Reithinger, N. and E. Maier. 1995. Utilizing statistical dialogue act processing in Verbmobil. In Proceedings of the ACL.

....has been much recent work on using context to constrain spoken language processing. Most of this work involves making predictions about possible sequences of utterances and using these predictions to limit the search space of the speech recognizer or some other component (See [2] 3] 4] 5] [6], 7] 8] 9] The goal of such an approach is to increase the accuracy of the top best hypothesis of the speech recognizer, which is then passed on to the language processing components of the system. The underlying assumption being made is that design and complexity considerations require ....

N. Reithinger, E. Maier. Utilizing Statistical Dialogue Act Processing in Verbmobil In Proceedings of the ACL, 1995.

No context found.

Reithinger N. and Maier E.: Utilizing Statistical Dialogue Act Processing in Verbmobil, ACL-95, pp. 116-121, 1995.

No context found.

Reithinger, N. and Maier, E. 1995. Utilizing Statistical Dialogue Act Processing in Verbmobil. In Proceedings of the 33rd Annual Meeting of the Association for Computational Linguistics, Cambridge, MA.

No context found.

Rei95. Reithinger, N. and Maier, E. Utilizing Statistical Dialogue Act Processing in VERBMOBIL. Tech. Rept. Report 80, Verbmobil, 1995.

No context found.

Enschede. Reithinger, N. and Maier, E. 1995. Utilizing Statistical Dialogue Act Processing in VERBMOBIL. Verbmobil, VM-Report 80.

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