In this thesis, we develop computational tools for analyzing conversations based on nonverbal auditory cues. We develop a notion of conversations as being made up of a variety of scenes: in each scene, either one speaker is holding the floor or both are speaking at equal levels. Our goal is to find conversations, find the scenes within them, determine what is happening inside the scenes, and then use the scene structure to characterize entire conversations. We begin by

This document reviews the literature on shallow discourse parsing, in particular the use of machine learning techniques. This is deliverable Y1.M6 of the Discourse Parsing White Paper which is part of the MDM IP of the IM2 project.

The lexical items that can serve as discourse markers (DMs) are often multi-functional. Like and well, in particular, play numerous other roles apart from DMs: for instance, the first one can also be a verb and the second one an adverb. The goal of the present study is the identification, on transcripts of multi-party dialogues, of the occurrences of like and well that play a discourse or pragmatic role. DM identification is a binary classification task over the set of all occurrences of tokens like and well. The importance of DMs to computational linguistics is first discussed, along with previous experiments in DM identification. Then, the data is briefly described, emphasizing the DM annotation procedure and an inter-annotator agreement study. The proposed method uses lexical, prosodic/positional and sociolinguistic features, together with machine learning algorithms, among which decision trees are preferred. The results obtained using a ten-fold cross-validation procedure are analysed at length, focussing first on overall performance, and then on the relevance of each type of features. Feature analysis using a range of techniques shows that lexical indicators are the most reliable features for DM identification, followed by prosodic/positional features. Sociolinguistic features are slightly correlated with the use of like as DM, while the dialogue act of the utterance containing a DM candidate

This article discusses the detection of discourse markers (DM) in dialog transcriptions, by human annotators and by automated means. After a theoretical discussion of the definition of DMs and their relevance to natural language processing, we focus on the role of like as a DM. Results from experiments with human annotators show that detection of DMs is a difficult but reliable task, which requires prosodic information from soundtracks. Then, several types of features are defined for automatic disambiguation of like: collocations, part-of-speech tags and duration-based features. Decision-tree learning shows that for like, nearly 70 % precision can be reached, with near 100 % recall, mainly using collocation filters. Similar results hold for well, with about 91 % precision at 100 % recall. 1

Turn-taking is usually considered to follow a simple set of rules, enacted through a perhaps more complicated system of signals. The most significant aspect of the turn-taking process is that, in most cases, it proceeds in a very smooth fashion. Speakers signal to each other that they wish to either yield or take the turn through syntactic, pragmatic, and prosodic means. In this paper, I explore how the turn-taking process develops in two different sets of Spanish conversations. In the first group of conversations, speakers take turns spontaneously, presumably as they would do in everyday situations. In the second group, turns were mechanically controlled, and communication was one-way. A comparison of the two types of conversation provides insights into the signals used in spontaneous turn-taking.

Abstract not found

This paper investigates the relation between the sound patterns of interjections and their functional realisation in the discourse process. It considers whether certain interjection functions tend to have particular sound distributions. In order to address these questions a classification scheme for American English nonlexical interjections in terms of discourse markers is also presented.

this paper, in textbooks by Allen (1995) and by Jurafsky and Martin (2000), in a survey by Cohen (1996), and in the websites of the Association for Computational Lin2 guistics' Special Interest Group on Discourse and Dialogue (SIGDial) (http://www.sigdial.org) and the Language Engineering Telematics project, MATE (http://mate.nis.sdu.dk/).

This thesis proposal describes completed and future studies on three particular aspects of prosodic variation in American English spoken dialogue: how information status affects the way humans decide which words to accent and which accent types to use; how humans produce potentially ambiguous single affirmative words like alright, okay, or yeah; and how humans manage to engage in synchronized conversation, anticipating the end of the speaker’s turn and starting their contributions at an appropriate time. The availability of these results should help improve the naturalness of speech synthesizers as well as the quality of human-machine interaction in voice response systems. As part of this work, we have designed, collected, and annotated a large corpus of spontaneous, task-oriented conversations, which we plan to use in a series of corpus analyses, perception studies and machine learning experiments to address these questions.

We present results of a series of machine learning experiments that address the classification of the discourse function of single affirmative cue words such as alright, okay and mm-hm in a spoken dialogue corpus. We suggest that a simple discourse/sentential distinction is not sufficient for such words and propose two additional classification sub-tasks: identifying (a) whether such words convey acknowledgment or agreement, and (b) whether they cue the beginning or end of a discourse segment. We also study the classification of each individual word into its most common discourse functions. We show that models based on contextual features extracted from the time-aligned transcripts approach the error rate of trained human aligners. Index Terms: cue words, discourse markers, spoken dialogue systems.