In this study, we analyzed prosodic features of accentual phrases and investigated their temporal changes to obtain cues for detecting boundaries at where turn-taking could occur in spontaneous conversations. The acoustic parameters used as prosodic features were the fundamental frequency, sound pressure level, and duration of accentual phrases in long utterance units. The results showed that t...

Methods for utterance veriication (UV) and their integration into statistical language modeling and spoken language understanding formalisms for a large vocabulary spoken understanding system are presented. The paper consists of three parts. First, a set of acoustic likelihood ratio based utterance veriication techniques are described and applied to the problem of rejecting portions of a hypoth...

To generate spontaneous utterances, it is important to omit some items from them appropriately. Previously, generation of elliptical utterances is discussed only from a view of whether an item in question is known or unknown in the context of dialogue. We point out that some items that are already known should not be omitted in some situation. For example, omitting too many items from utterance...

Francez and Berg (1994) introduced the term CROSS-SPEAKER ANAPHORA to indicate reference by conversational participant (CP) B to a discourse entity introduced by CP A. Crossspeaker anaphora is further discussed in (Groenendijk, Stokhof, and Veltman, 1997; Dekker, 1997), where it is noted that the phenomenon is subject to a number of constraints: e.g., it is generally not felicitous for B to fol...

This paper reports our recent development of a highly reliable call analysis technique that makes novel use of automatic speech recognition (ASR), speech utterance classification and non-speech features. The main ideas include the use the NGram filler model to improve the ASR accuracy on important words in a message, and the integration of recognized utterance with non-speech features such as u...

This paper describes classification of typed student utterances within AutoTutor, an intelligent tutoring system. Utterances are classified to one of 18 categories, including 16 question categories. The classifier presented uses part of speech tagging, cascaded finite state transducers, and simple disambiguation rules. Shallow NLP is well suited to the task: session log file analysis reveals si...

Previous research shows that two-year-olds’ third person singular -s and plural -s are produced more accurately in utterance-final compared to utterance-medial position. However, only the third person singular is affected by coda complexity. This study explores these effects with possessive -s. Acoustic analysis of twelve two-year-olds’ elicited imitations examined the use of simple versus comp...

This paper discusses the utterance generation module of a spoken dialogue system for transactions. Transactions are interesting because they involve obligations of both parties: the system should provide all relevant information; the user should feel committed to the transaction once it has been concluded. Utterance generation plays a major role in this. The utterance generation module works wi...