BibRank: Automatic Keyphrase Extraction Platform Using Metadata

Abstract

Automatic Keyphrase extraction is the process of automatically identifying the essential phrases from a document. Keyphrases are used in crucial tasks such as document classification, clustering, recommendation, indexing, searching, and summarization. This thesis introduces BibRank, a new semi-supervised automatic keyphrase extraction method that exploits an information-rich dataset collected by parsing bibliographic data in BibTeX format. BibRank combines a novel weighting technique of the bibliographic data with positional, statistical, and word co-occurrence information. We have benchmarked BibRank and state-of-the-art techniques against the dataset. The evaluation indicates that BibRank is more stable and has a better performance than state-of-the-art methods.