We describe a statistical approach for modeling dialogue acts in conversational speech, i.e., speec-act-like

A crucial step in processing speech audio data for information-extraction, topic detection, or browsing/playback is to segment the input into sentence and topic units. Speech segmentation is challenging, since the cues typically present for segmenting text (headers, paragraphs, punctuation) are absent in spoken language. We investigate the use of prosody (informationgleaned from the timing and melody of speech) for these tasks. Using decision tree and hidden Markov modeling techniques, we combine prosodic cues with word-based approaches, and evaluate performance on two speech corpora, Broadcast News and Switchboard. Results show that the prosodic model alone performs on par with, or better than, word-based statistical language models—for both true and automatically recognized words in news speech. The prosodic model achieves comparable performance with significantly less training data, and requires no hand-labeling of prosodic events. Across tasks and corpora, we obtain a significant improvement over word-only models using a probabilistic combination of prosodic and lexical information. Inspection reveals that the prosodic models capture language-independent boundary indicators described in the literature. Finally, cue usage is task and corpus dependent. For example, pause and pitch features are highly informative for segmenting news speech, whereas pause, duration and word-based cues dominate for natural conversation.

Identifying whether an utterance is a statement, question, greeting, and so forth is integral to effective automatic understanding of natural dialog. Little is known, however, about how such dialog acts (DAs) can be automatically classified in truly natural conversation. This study asks whether current approaches, which use mainly word information, could be improved by adding prosodic information. The study is based on more than 1000 conversations from the Switchboard corpus. DAs were handannotated, and prosodic features (duration, pause, F0, energy, and speaking rate) were automatically extracted for each DA. In training, decision trees based on these features were inferred

We describe a new approach for statistical modeling and detection of discourse structure for natural conversational speech. Our model is based on 42 'Dialog Acts' (DAs), (question, answer, backchannel, agreement, disagreement, apology, etc). We labeled 1155 conversations from the Switchboard (SWBD) database (Godfrey et al. 1992) of human-to-human telephone conversations with these 42 types and trained a Dialog Act detector based on three distinct knowledge sources: sequence of words which characterize a dialog act, prosodic features which characterize a dialog act, and a statistical Discourse Grammar. Our combined detector, although still in preliminary stages, already achieves a 65% Dialog Act detection rate based on acoustic waveforms, and 72% accuracy based on word transcripts. Using this detector to switch among the 42 Dialog-Act-Specific trigram LMs also gave us an encouraging but not statistically significant reduction in SWBD word error.

We describe a new approach for statistical modeling and detection of discourse structure for natural conversational speech. Our model is based on 42 `Dialog Acts' (DAs), (question, answer, backchannel, agreement, disagreement, apology, etc). We labeled 1155 conversations from the Switchboard (SWBD) database (Godfrey et al. 1992) of human-to-human telephone conversations with these 42 types and trained a Dialog Act detector based on three distinct knowledge sources: sequences of words which characterize a dialog act, prosodic features which characterize a dialog act, and a statistical Discourse Grammar. Our combined detector, although still in preliminary stages, already achieves a 65% Dialog Act detection rate based on acoustic waveforms, and 72% accuracy based on word transcripts. Using this detector to switch among the 42 dialog-act-specific trigram LMs also gave us an encouraging but not statistically significant reduction in SWBD word error. 1 Introduction The ability to model and...

Automatic summarization of open-domain spoken dialogues is a relatively new research area. This article introduces the task and the challenges involved and motivates and presents an approach for obtaining automatic-extract summaries for human transcripts of multiparty dialogues of four different genres, without any restriction on domain. We address the following issues, which are intrinsic to spoken-dialogue summarization and typically can be ignored when summarizing written text such as news wire data: (1) detection and removal of speech disfluencies; (2) detection and insertion of sentence boundaries; and (3) detection and linking of cross-speaker information units (question-answer pairs). A system evaluation is performed using a corpus of 23 dialogue excerpts with an average duration of about 10 minutes, comprising 80 topical segments and about 47,000 words total. The corpus was manually annotated for relevant text spans by six human annotators. The global evaluation shows that for the two more informal genres, our summarization system using dialoguespecific components significantly outperforms two baselines: (1) a maximum-marginal-relevance ranking algorithm using TF*IDF term weighting, and (2) a LEAD baseline that extracts the first n words from a text. 1.

Function words, especially frequently occurring ones such as (the, that, and, and of), vary widely in pronunciation. Understanding this variation is essential both for cognitive modeling of lexical production and for computer speech recognition and synthesis. This study investigates which factors affect the forms of function words, especially whether they have a fuller pronunciation (e.g., , , 22 , ) or a more reduced or lenited pronunciation (e.g., ). It is based on over 8000 occurrences of the ten most frequent English function words in a four-hour sample from conversations from the Switchboard corpus. Ordinary linear and logistic regression models were used to examine variation in the length of the words, in the form of their vowel (basic, full, or reduced), and whether final obstruents were present or not. For all these measures, after controlling for segmental context, rate of speech, and other important factors, there are strong independent effects that made high-frequency monosyllabic function words more likely to be longer or have a fuller form (1) when neighboring disfluencies (such as filled pauses uh and um) indicate that the speaker was encountering problems in planning the utterance; (2) when the word is unexpected, i.e less predictable in context; (3) when the word is either utterance-initial or utterance-final. Looking at the phenomenon in a different way, frequent function words are more likely to be shorter and to have less full forms in fluent speech, in predictable positions or multi-word collocations, and utterance-internally. Also considered are other factors such as sex (women are more likely to use fuller forms, even after controlling for rate of speech, for example), and some of the dif...

We investigate a new approach for using speech prosody as a knowledge source for speech recognition. The idea is to penalize word hypotheses that are inconsistent with prosodic features such as duration and pitch. To model the interaction between words and prosody we modify the language model to represent hidden events such as sentence boundaries and various forms of disfluency, and combine with it decision trees that predict such events from prosodic features. N-best rescoring experiments on the Switchboard corpus show a small but consistent reduction of word error as a result of this modeling. We conclude with a preliminary analysis of the types of errors that are corrected by the prosodically informed model. 1.

We describe an integrated approach for statistical modeling of discourse structure for natural conversational speech. Our model is based on 42 `dialog acts' (e.g., Statement, Question, Backchannel, Agreement, Disagreement, Apology), which were hand-labeled in 1155 conversations from the Switchboard corpus of spontaneous human-to-human telephone speech. We developed several models and algorithms to automatically detect dialog acts from transcribed or automatically recognized words and from prosodic properties of the speech signal, and by using a statistical discourse grammar. All of these components were probabilistic in nature and estimated from data, employing a variety of techniques (hidden Markov models, N-gram language models, maximum entropy estimation, decision tree classifiers, and neural networks). In preliminary studies, we achieved a dialog act labeling accuracy of 65% based on recognized words and prosody, and an accuracy of 72% based on word transcripts. Since humans achiev...

Disfluencies occur frequently in spontaneous speech. Detection and correction of disfluencies can make automatic speech recognition transcripts more readable for human readers, and can aid downstream processing by machine. This work investigates a number of knowledge sources for disfluency detection, including acoustic-prosodic features, a language model (LM) to account for repetition patterns, a part-of-speech (POS) based LM, and rule-based knowledge. Different components are designed for different purposes in the system. Results show that detection of disfluency interruption points is best achieved by a combination of prosodic cues, word-based cues, and POS-based cues. The onset of a disfluency to be removed, in contrast, is best found using knowledge-based rules. Finally, specific disfluency types can be aided by the modeling of word patterns. 1.