Linkping University Electronic Press

Title:	Tracking Morphological and Semantic Co-occurrences in Spontaneous Dialogues
Authors:	Mark Seligman, Jan Alexandersson, and Kristiina Jokinen
Series:	Linkping Electronic Articles in Computer and Information Science ISSN 1401-9841
Issue:	Vol. 4 (1999), No. 031
URL:	http://www.ep.liu.se/ea/cis/1999/031/

Abstract:	In the processing of spontaneous language, information concerning discourse-level co-occurrences of words or morphemes - relatively long-term predictions on the scale of several utterances - may help to reduce perplexity in speech recognition, facilitate lexical disambiguation, and contribute to topic tracking. This working paper describes a new set of facilities for tracking lexical co-occurrences. The major innovation is the use of semantic smoothing: we track co-occurrences of semantic tokens associated with words or morphs in addition to co-occurrences of the words or morphs themselves. Such smoothing offers an approach to the problem of data sparseness: it is possible to retrieve reasonable semantically-mediated associations for morphs not in the training corpus. We report on preliminary experiments with a corpus of morphologically-tagged transcripts of 16 spontaneous Japanese dialogues concerning direction-finding and hotel arrangements. We close with discussion of lexical disambiguation and topic tracking as they relate to co-occurrence networks.

Abstract:

In the processing of spontaneous language, information concerning discourse-level co-occurrences of words or morphemes - relatively long-term predictions on the scale of several utterances - may help to reduce perplexity in speech recognition, facilitate lexical disambiguation, and contribute to topic tracking. This working paper describes a new set of facilities for tracking lexical co-occurrences. The major innovation is the use of semantic smoothing: we track co-occurrences of semantic tokens associated with words or morphs in addition to co-occurrences of the words or morphs themselves. Such smoothing offers an approach to the problem of data sparseness: it is possible to retrieve reasonable semantically-mediated associations for morphs not in the training corpus. We report on preliminary experiments with a corpus of morphologically-tagged transcripts of 16 spontaneous Japanese dialogues concerning direction-finding and hotel arrangements. We close with discussion of lexical disambiguation and topic tracking as they relate to co-occurrence networks.

Original publication 1999-12-30	Postscript part I -- Checksum Checksum (old) Information about recalculation of checksum Postscript part II -- Checksum II Checksum II (old) Information about recalculation of checksum