Context-dependent regression testing for natural language processing

Elaine Farrow, Myroslava O. Dzikovska

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    Regression testing of natural language systems is problematic for two main reasons: component input and output is complex, and system behaviour is context-dependent. We have developed a generic approach which solves both of these issues. We describe our regression tool, Contest, which supports context-dependent testing of dialogue system components, and discuss the regression test sets we developed, designed to effectively isolate components from changes and problems earlier in the pipeline. We believe that the same approach can be used in regression testing for other dialogue systems, as well as in testing any complex NLP system containing multiple components.
    Original languageEnglish
    Title of host publicationProceedings of the Workshop on Software Engineering, Testing, and Quality Assurance for Natural Language Processing
    EditorsP.G.T Healey, R. Pieraccini, D. Byron, S. Young, M. Purver
    Place of PublicationStroudsburg, PA, USA
    PublisherAssociation for Computational Linguistics
    Pages5-13
    Number of pages9
    ISBN (Electronic)978-1-932432-32-9
    Publication statusPublished - Jun 2009
    EventNAACL-HLT 2009 North American Chapter of the Association for Computational Linguistics - Human Language Technologies Conference - Boulder, Colorado, United States
    Duration: 31 May 20095 Jun 2009

    Conference

    ConferenceNAACL-HLT 2009 North American Chapter of the Association for Computational Linguistics - Human Language Technologies Conference
    CountryUnited States
    CityBoulder, Colorado
    Period31/05/095/06/09

    Fingerprint Dive into the research topics of 'Context-dependent regression testing for natural language processing'. Together they form a unique fingerprint.

    Cite this