Unsupervised Approaches for Automatic Keyword Extraction Using Meeting Transcripts

This paper explores several unsupervised approaches to automatic keyword extraction using meeting transcripts. In the TFIDF (term frequency, inverse document frequency) weighting framework, we incorporated part-of-speech (POS) information, word clustering, and sentence salience score. We also evaluated a graph-based approach that measures the importance of a word based on its connection with other sentences or words. The system performance is evaluated in different ways, including comparison to human annotated keywords using F-measure and a weighted score relative to the oracle system performance, as well as a novel alternative human evaluation. Our results have shown that the simple unsupervised TFIDF approach performs reasonably well, and the additional information from POS and sentence score helps keyword extraction. However, the graph method is less effective for this domain. Experiments were also performed using speech recognition output and we observed degradation and different patterns compared to human transcripts.