Any questions? Automatic question detection in meetings

Abstract

Abstract-In this paper, we describe our efforts toward the automatic detection of English questions in meetings. We analyze the utility of various features for this task, originating from three distinct classes: lexico-syntactic, turn-related, and pitchrelated. Of particular interest is the use of parse tree information in classification, an approach as yet unexplored. Results from experiments on the ICSI MRDA Corpus demonstrate that lexicosyntactic features are most useful for this task, with turn-and pitch-related features providing complementary information in combination. In addition, experiments using reference parse trees on the Broadcast Conversation portion of the OntoNotes release 2.9 data set illustrate the potential of parse trees to outperform word lexical features.

Identifying Teacher Questions Using Automatic Speech Recognition in Classrooms

“…Our best model achieved an overall F 1 score of 0.69 and an F 1 score of 0.59 for the question class. This represents a substantial 37% improvement in question detection accuracy over a recent state-ofthe-art model (Boakye et al, 2009) that reported an overall F 1 of 0.50; the authors do not report F 1 for the question class so the comparison is based on the overall F 1 .…”

“…Additionally, related work on question detection (see Section 1.1) suggested that acoustic, contextual, and temporal features (Boakye et al, 2009) may aid in the detection of questions. We will explore this in future work to determine if features capturing these properties will help improve our models for this task.…”

“…Boakye et al (2009) trained models to detect questions in office meetings. The authors used the ICSI Meeting Recorder Dialog Act (MRDA) corpus, a set of 75 hour-long meetings recorded with headset and lapel microphones.…”

Review spam detector with rating consistency check

“…This is the usual manner in which the reviewer tries to communicate with other reviewers or with the seller of the product. There are generally three types of questions [15]. An example of each is shown in Table 1.…”

Welcome Message from the Chairs

“…Unigram, bigram and trigrams, start and end utterance tags, parse tree representation of syntax, etc. have been used as textual features in addition to acoustic features in questions detection from English, French and Vietnamese utterances [9], [14], [21]. A recent study [18] on French questions detection has combined language model features extracted from speech text with acoustic features (duration, energy and pitch) with 75% accuracy.…”