Stolcke, A., Shriberg, E., Tr, D., Tr, G. (1999). Modeling the Prosody of Hidden Events for Improved Word Recognition. In: Proceedings of Eurospeech, Budapest, Hungary, pp. 311-314.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....detection or correction. 10] report a recall of 83.4 and a precision of 93.9 in detecting the IP of a repair, but do not discuss the problem of finding the correct segmentation in detail. In addition their results are obtained on a corpus where every utterance contains at least one repair. [12] introduce hidden events to model the IPs of different classes of repairs in the speech recognition process. This reduces their recognition errors by about 0.9 absolute, but nothing is said about recall and precision of IP detection. Likewise they make no suggestion about getting the correct ....

A. Stolcke, E. Shriberg, D. Hakkani-Tur, and G. Tur. Modeling the prosody of hidden events for improved word recognition. In EUROSPEECH '99, volume 1, pages 307-- 310, Budapest, 1999.

....translated in any sentence T in a target language #e.g. German#. Toevery pair #S; T# a probability is assigned which re#ects the likelihood that a translator who sees S will produce T as the translation. The statistical machine translation problem is 4 Other triggers can be added as well. #Stolcke et al. 1999# for example integrate prosodic cues and an extended language model in a speech recognizer to detect IPs. 5 The classi#er is developed by the speech group of the IMMD 5. Special thanks to Anton Batliner, Richard Huber and Volker Warnke. 6 A more detailed introduction is given by #Brown et ....

A. Stolcke, E. Shriberg, D. Hakkani-Tur, and G. Tur. 1999. Modeling the prosody of hidden events for improved word recognition. In EUROSPEECH '99,volume 1, pages 307# 310, Budapest.

....translated in any sentence T in a target language (e.g. German) To every pair (S; T) a probability is assigned which reflects the likelihood that a translator who sees S will produce T as the translation. The statistical machine translation problem is 4 Other triggers can be added as well. (Stolcke et al. 1999) for example integrate prosodic cues and an extended language model in a speech recognizer to detect IPs. 5 The classifier is developed by the speech group of the IMMD 5. Special thanks to Anton Batliner, Richard Huber and Volker Warnke. 6 A more detailed introduction is given by (Brown et ....

A. Stolcke, E. Shriberg, D. Hakkani-Tur, and G. Tur. 1999. Modeling the prosody of hidden events for improved word recognition. In EUROSPEECH '99, volume 1, pages 307-- 310, Budapest.

.... that intonational information can reduce syntactic ambiguity for humans (Beach, 1991) and for computer parsers (Ostendorf, Wightman, and Veilleux, 1993) Other researchers have proposed segmenting speech into speech acts (Mast et al. 1996) or into linguistic clauses (i.e. Meteer and Iyer, 1996; Stolcke et al. 1999). There is no clear consensus as to the right approach, however. Although intonational phrases might not be the ideal unit for modeling interaction in dialogue, it does captures speaker intention and is a major component of any definition (Traum and Heeman, 1997) Now that we have introduced the ....

....These results illustrate the potential improvement that can be gained in language modeling from improving the modeling of speech repairs and intonational phrase boundaries. Part of this improvement can be obtained by improving the acoustic cues used to detect the spontaneous speech events (cf. Stolcke et al. 1999), and part due to improving the probability estimates for these events. 7. RELATIONSHIP TO OTHER WORK Although substantial research has been done in the area of intonational phrasing and speech repairs, very little work has been done on incorporating this research work with speech recognition. ....

[Article contains additional citation context not shown here]

Stolcke, A., E. Shriberg, D. Hakkani-Tur, and G. Tur. 1999. Modeling the prosody of hidden events for improved word recognition. In Proceedings of the 6th European Conference on Speech Communication and Technology.

No context found.

A. Stolcke et al. Modeling the prosody of hidden events for improved word recognition. In Proc. EUROSPEECH, vol. 1, pp. 307--310, Budapest, 1999.

No context found.

A. Stolcke, E. Shriberg, D. Hakkani-Tur, and G. Tur, "Modeling the prosody of hidden events for improved word recognition," in Proc. EUROSPEECH, Budapest, Sept. 1999, vol. 1, pp. 307--310.

No context found.

A. Stolcke, E. Shriberg, D. Hakkani-T ur, and G. T ur, "Modeling the prosody of hidden events for improved word recognition", in Proc. EUROSPEECH, vol. 1, pp. 307--310, Budapest, Sep. 1999.

....word stream. Instead, they correspond to the states of a hidden Markov model, and can be used to model linguistic events such as unmarked sentence boundaries. Optionally, these events can be associated with nonlexical likelihoods to condition the LM on other knowledge sources (e.g. prosody) [13]. A special type of hidden event LM can model speech disfluencies by allowing the hidden events to modify the word history; for example, a word deletion event would erase one or more words to model a false start [14] Skip language models In this LM, words in the history are probabilistically ....

A. Stolcke, E. Shriberg, D. Hakkani-Tur, and G. Tur, "Modeling the prosody of hidden events for improved word recognition", in Proc. EUROSPEECH, vol. 1, pp. 307--310, Budapest, Sep. 1999.

....studies that used prosody to help in syntactic disambiguation and understanding [1, 2] or to detect disfluencies and sentence boundaries [3, 4] Other efforts studied the effects of lexical stress, phone durations, or higher level prosodic information in ASR. Examples of this work can be found in [5, 6, 7, 8]. In this paper we investigate two of the above mentioned techniques plus a novel one, in order to integrate the different levels of prosodic information in an ASR system: # The first approach is an improved version of the wordduration model described in [6] Each word is represented by a ....

....(GMMs) are used to predict the word duration features from the hypothesized words and pauses. # The second approach addresses the interaction between words and the between word pauses, using an N gram model to predict the duration of the pauses from the word context. # In the third approach [8] prosodic features are used to predict certain hidden events in speech, such as segment boundaries and disfluencies. The interaction of the hidden events with the spoken words is modeled by an N gram language model (LM) All the above techniques were integrated in SRI s 2002 Switchboard system ....

[Article contains additional citation context not shown here]

A. Stolcke, E. Shriberg, D. Hakkani-T ur, and G. T ur, "Modeling the prosody of hidden events for improved word recognition," in Proc. EUROSPEECH, Budapest, Sept. 1999, vol. 1, pp. 307--310.

.... improved recognition in task oriented dialogs [14] but significant improvements in large vocabulary recognition remain elusive [11] We have had some success using the hidden event N gram model (previously introduced for sentence segmentation and disfluency detection) for word recognition [13]. As before, we computed prosodic likelihoods for each event type at each word boundary, and conditioned the word portion of the N gram on those events. The result was a small, but significant 2 relative reduction in Switchboard word recognition error. This improvement was surprising given that ....

A. STOLCKE,E.SHRIBERG,D.HAKKANI-T UR, AND G. T UR, Modeling the prosody of hidden events for improved word recognition, in Proceedings of the 6th European Conference on Speech Communication and Technology, vol. 1, Budapest, Sept. 1999, pp. 307--310.

.... improved recognition in task oriented dialogs [13] but significant improvements in large vocabulary recognition remain elusive [10] We have had some success using the hidden event N gram model (previously introduced for sentence segmentation and disfluency detection) for word recognition [12]. As before, we computed prosodic likelihoods for each event type at each word boundary, and conditioned the word portion of the N gram on those events. The result was a small, but significant 2 relative reduction in Switchboard word recognition error. This improvement was surprising given that ....

A. Stolcke, E. Shriberg, D. Hakkani-Tur, and G. Tur. Modeling the prosody of hidden events for improved word recognition. In Proc. EUROSPEECH, vol. 1, pp. 307--310, Budapest, 1999.

No context found.

Stolcke, A., Shriberg, E., Tr, D., Tr, G. (1999). Modeling the Prosody of Hidden Events for Improved Word Recognition. In: Proceedings of Eurospeech, Budapest, Hungary, pp. 311-314.

No context found.

A. Stolcke, et al., "Modeling the prosody of hidden events for improved word recognition," Proc. Eurospeech'99, Budapest, vol.1, pp.311--314, 1999.

No context found.

Stolcke, A., E. Shriberg, D. Hakkani-Tur, and G. Tur. "Modeling the prosody of hidden events for improved word recognition" In Proceedings of the 6th European Conference on Speech Communication and Technology. 1999

No context found.

A. Stolcke, E. Shriberg, D. Hakkani-Tur and G. Tur, "Modeling the prosody of hidden events for improved word recognition, " in Proc. Eurospeech, 1:311-314, 1999.

No context found.

Andreas Stolcke, Elizabeth Shriberg, Dilek Hakkani-T ur, and G okhan T ur, "Modeling the prosody of hidden events for improved word recognition," In Proc. of Eurospeech '99, pp. 311--314, 1999.

No context found.

Stolcke, A., Shriberg, E., Tr, D., Tr, G., "Modeling the Prosody of Hidden Events for Improved Word Recognition," Proc.Eurospeech, pp. 311-3314, 1999.

No context found.

Stolcke A., Shriberg E., Hakkani-Tr D., and Tr G., Modeling the Prosody of Hidden Events for Improved Word Recognition, Proc. 6 th European Conference on Speech Communication and Technology, Budapest, Hungary, Sept. 1999.

No context found.

A. Stolcke, E. Shriberg, D. Hakkani-Tr, and G. Tr, "Modeling the prosody of hidden events for improved word recognition," Proceedings of the 6th European Conference on Speech Communication and Technology, Budapest, Hungary, September 1999.

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