Speaker turn characterization for spoken dialog system monitoring and adaptation

Abstract

This paper describes an utterance classification method based on a multiple decoding scheme. We use the Spoken Language Understanding (SLU) strategy proposed within the European project LUNA. The goal of this classification process is to characterize each speaker's turn, in a dialog context, according to different categories relevant from an SLU point of view: out-of-domain messages, requests not covered by the interpretation module, frequent requests,. . . . These categories are used for two purposes in an off-line mode: system monitoring for detecting changes in users' behaviour and system adaptation by selecting dialogs likely to contain some phenomenon poorly covered by the models for an active learning scheme. All the models and the evaluations are performed on the France Telecom FT3000 corpus.

Active learning for rule-based and corpus-based Spoken Language Understanding models

“…For example all the paths of the word/concept transducer can be scored thanks to a Hidden Markov Model tagger trained on the corpus collected from the deployed system. We have proposed in [6], [7] different strategies mixing statistical approaches modeling the plausibility of an interpretation with the rule-based approach of the deployed FT3000, based on a priori knowledge modeling the acceptability of an interpretation. We propose in this paper to build a statistical SLU module for replacing the logical composition rules of the deployed FT3000 SLU, using an active learning approach.…”