Multi-feature error detection in spoken dialogue systems

P.K. Lendvai, A. van den Bosch, E.J. Krahmer, M.G.J. Swerts

    Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

    80 Downloads (Pure)

    Abstract

    The present paper evaluates the role selected features and feature combinations play for error detection in spoken dialogue systems. We investigate the relevance of various, readily available features extracted from a corpus of dialogues with a train timetable information system, using RIPPER, a rule-inducing machine learning algorithm. The learning task consists of the identification of communication problems arising in either the previous turn or the current turn of the dialogue. Previous experiments with our corpus have shown that combining dialogue history and word-graph features is beneficial for detecting errors (in particular in the previous turn). Other researchers have reported that combining prosodic and ASR characteristics is helpful (primarily in the current turn). In this paper, we investigate the usefulness of large-scale combinations of these features for the above two tasks. We show that we are unable to reproduce the benefits of prosodic features for learning problematic situations, even though the overall prosodic trends in our corpus are similar to those earlier reported on. Moreover, the best results are obtained using just minimal combinations of two sources of information.
    Original languageEnglish
    Title of host publicationComputational Linguistics in the Netherlands 2001
    Subtitle of host publicationSelected Papers from the Twelfth CLIN Meeting
    EditorsM. Theune, A. Nijholt, H. Hondorp
    Place of PublicationAmsterdam
    PublisherRodopi
    Pages163-178
    Number of pages16
    ISBN (Print)9042009438
    Publication statusPublished - 2002

    Publication series

    NameLanguage and computers
    Number45

    Fingerprint

    Error detection
    Learning algorithms
    Learning systems
    Information systems
    Communication
    Experiments

    Cite this

    Lendvai, P. K., van den Bosch, A., Krahmer, E. J., & Swerts, M. G. J. (2002). Multi-feature error detection in spoken dialogue systems. In M. Theune, A. Nijholt, & H. Hondorp (Eds.), Computational Linguistics in the Netherlands 2001: Selected Papers from the Twelfth CLIN Meeting (pp. 163-178). (Language and computers; No. 45). Amsterdam: Rodopi.
    Lendvai, P.K. ; van den Bosch, A. ; Krahmer, E.J. ; Swerts, M.G.J. / Multi-feature error detection in spoken dialogue systems. Computational Linguistics in the Netherlands 2001: Selected Papers from the Twelfth CLIN Meeting. editor / M. Theune ; A. Nijholt ; H. Hondorp. Amsterdam : Rodopi, 2002. pp. 163-178 (Language and computers; 45).
    @inproceedings{aa7945328a624f969d20f5b5751c7394,
    title = "Multi-feature error detection in spoken dialogue systems",
    abstract = "The present paper evaluates the role selected features and feature combinations play for error detection in spoken dialogue systems. We investigate the relevance of various, readily available features extracted from a corpus of dialogues with a train timetable information system, using RIPPER, a rule-inducing machine learning algorithm. The learning task consists of the identification of communication problems arising in either the previous turn or the current turn of the dialogue. Previous experiments with our corpus have shown that combining dialogue history and word-graph features is beneficial for detecting errors (in particular in the previous turn). Other researchers have reported that combining prosodic and ASR characteristics is helpful (primarily in the current turn). In this paper, we investigate the usefulness of large-scale combinations of these features for the above two tasks. We show that we are unable to reproduce the benefits of prosodic features for learning problematic situations, even though the overall prosodic trends in our corpus are similar to those earlier reported on. Moreover, the best results are obtained using just minimal combinations of two sources of information.",
    author = "P.K. Lendvai and {van den Bosch}, A. and E.J. Krahmer and M.G.J. Swerts",
    note = "Pagination: 16",
    year = "2002",
    language = "English",
    isbn = "9042009438",
    series = "Language and computers",
    publisher = "Rodopi",
    number = "45",
    pages = "163--178",
    editor = "M. Theune and A. Nijholt and H. Hondorp",
    booktitle = "Computational Linguistics in the Netherlands 2001",

    }

    Lendvai, PK, van den Bosch, A, Krahmer, EJ & Swerts, MGJ 2002, Multi-feature error detection in spoken dialogue systems. in M Theune, A Nijholt & H Hondorp (eds), Computational Linguistics in the Netherlands 2001: Selected Papers from the Twelfth CLIN Meeting. Language and computers, no. 45, Rodopi, Amsterdam, pp. 163-178.

    Multi-feature error detection in spoken dialogue systems. / Lendvai, P.K.; van den Bosch, A.; Krahmer, E.J.; Swerts, M.G.J.

    Computational Linguistics in the Netherlands 2001: Selected Papers from the Twelfth CLIN Meeting. ed. / M. Theune; A. Nijholt; H. Hondorp. Amsterdam : Rodopi, 2002. p. 163-178 (Language and computers; No. 45).

    Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

    TY - GEN

    T1 - Multi-feature error detection in spoken dialogue systems

    AU - Lendvai, P.K.

    AU - van den Bosch, A.

    AU - Krahmer, E.J.

    AU - Swerts, M.G.J.

    N1 - Pagination: 16

    PY - 2002

    Y1 - 2002

    N2 - The present paper evaluates the role selected features and feature combinations play for error detection in spoken dialogue systems. We investigate the relevance of various, readily available features extracted from a corpus of dialogues with a train timetable information system, using RIPPER, a rule-inducing machine learning algorithm. The learning task consists of the identification of communication problems arising in either the previous turn or the current turn of the dialogue. Previous experiments with our corpus have shown that combining dialogue history and word-graph features is beneficial for detecting errors (in particular in the previous turn). Other researchers have reported that combining prosodic and ASR characteristics is helpful (primarily in the current turn). In this paper, we investigate the usefulness of large-scale combinations of these features for the above two tasks. We show that we are unable to reproduce the benefits of prosodic features for learning problematic situations, even though the overall prosodic trends in our corpus are similar to those earlier reported on. Moreover, the best results are obtained using just minimal combinations of two sources of information.

    AB - The present paper evaluates the role selected features and feature combinations play for error detection in spoken dialogue systems. We investigate the relevance of various, readily available features extracted from a corpus of dialogues with a train timetable information system, using RIPPER, a rule-inducing machine learning algorithm. The learning task consists of the identification of communication problems arising in either the previous turn or the current turn of the dialogue. Previous experiments with our corpus have shown that combining dialogue history and word-graph features is beneficial for detecting errors (in particular in the previous turn). Other researchers have reported that combining prosodic and ASR characteristics is helpful (primarily in the current turn). In this paper, we investigate the usefulness of large-scale combinations of these features for the above two tasks. We show that we are unable to reproduce the benefits of prosodic features for learning problematic situations, even though the overall prosodic trends in our corpus are similar to those earlier reported on. Moreover, the best results are obtained using just minimal combinations of two sources of information.

    M3 - Conference contribution

    SN - 9042009438

    T3 - Language and computers

    SP - 163

    EP - 178

    BT - Computational Linguistics in the Netherlands 2001

    A2 - Theune, M.

    A2 - Nijholt, A.

    A2 - Hondorp, H.

    PB - Rodopi

    CY - Amsterdam

    ER -

    Lendvai PK, van den Bosch A, Krahmer EJ, Swerts MGJ. Multi-feature error detection in spoken dialogue systems. In Theune M, Nijholt A, Hondorp H, editors, Computational Linguistics in the Netherlands 2001: Selected Papers from the Twelfth CLIN Meeting. Amsterdam: Rodopi. 2002. p. 163-178. (Language and computers; 45).