Follow-up utterances in QA dialogue

Boris van Schooten, Rieks op den Akker

    Research output: Contribution to journalArticleAcademicpeer-review

    49 Downloads (Pure)


    The processing of user follow-up utterances by a QA system is a topic which is still in its infant stages, but enjoys growing interest in the QA community. In this paper, we discuss the broader issues related to handling follow-up utterances in a real-life "information kiosk" setting. With help of a corpus of follow-up utterances, we lay out a general framework for QA dialogue, and propose a general-coverage classication of real user reactions, based on their function in the dialogue, and their relation to the context of previous utterances. We discuss a dialogue manager we have developed that handles the identied types of follow-up utterances. It uses existing NLP and IR techniques to handle follow-up questions, and generates appropriate dialogue reactions to handle non-follow-up questions. Our main conclusion is that a large part of follow-up utterances can be handled by a set of simple techniques, even though we found it is not easy to determine the exact context in which follow-up questions are to be understood.
    Original languageEnglish
    Pages (from-to)181-206
    Number of pages26
    JournalTraitement Automatique des Langues (TAL)
    Issue number3
    Publication statusPublished - 29 Sep 2006


    • HMI-SLT: Speech and Language Technology
    • HMI-HF: Human Factors
    • METIS-248450
    • IR-63437
    • EWI-6938
    • follow-up questions
    • sentence rewriting
    • dialogue management
    • Question Answering


    Dive into the research topics of 'Follow-up utterances in QA dialogue'. Together they form a unique fingerprint.

    Cite this