Likey: Unsupervised Language-Independent Keyphrase Extraction

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 162–165, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics Likey: Unsupervised Language-independent Keyphrase Extraction Mari-Sanna Paukkeri and Timo Honkela Adaptive Informatics Research Centre Aalto University School of Science and Technology P.O. Box 15400, FI-00076 AALTO, Finland mari-sanna.paukkeri@tkk.fi Abstract Likey is an unsupervised statistical ap- proach for keyphrase extraction. The method is language-independent and the only language-dependent component is the reference corpus with which the doc- uments to be analyzed are compared. In this study, we have also used an- other language-dependent component: an English-specific Porter stemmer as a pre- processing step. In our experiments of keyphrase extraction from scientific articles, the Likey method outperforms both supervised and unsupervised baseline methods. 1 Introduction Keyphrase extraction is a natural language pro- cessing task for collecting the main topics of a document into a list of phrases. Keyphrases are supposed to be available in the processed docu- ments themselves, and the aim is to extract these most meaningful words and phrases from the doc- uments. Keyphrase extraction summarises the content of a document as few phrases and thus provides a quick way to find out what the docu- ment is about. Keyphrase extraction is a basic text mining procedure that can be used as a ground for other, more sophisticated text analysis meth- ods. Automatically extracted keyphrases may be used to improve the performance of information retrieval, automatic user model generation, docu- ment collection clustering and visualisation, sum- marisation and question-answering, among others. This article describes the participation of the Likey method in the Task 5 of the SemEval 2010 challenge, automatic keyphrase extraction from scientific articles (Kim et al., 2010). 1.1 Related work In statistical keyphrase extraction, many variations for term frequency counts have been proposed in the literature including relative frequencies (Dam- erau, 1993), collection frequency (Hulth, 2003), term frequency–inverse document frequency (tf- idf) (Salton and Buckley, 1988), among others. Additional features to frequency that have been experimented are e.g., relative position of the first occurrence of the term (Frank et al., 1999), im- portance of the sentence in which the term oc- curs (HaCohen-Kerner, 2003), and widely stud- ied part-of-speech tag patterns, e.g. Hulth (2003). Matsuo and Ishizuka (2004) present keyword ex- traction method using word co-occurrence statis- tics. An unsupervised keyphrase extraction method by Liu et al. (2009) uses clustering to find exemplar terms that are then used for keyphrase extraction. Most of the presented methods require a reference corpus or a training corpus to produce keyphrases. Statistical keyphrase extraction meth- ods without reference corpora have also been pro- posed, e.g. (Matsuo and Ishizuka, 2004; Bracewell et al., 2005). The later study is carried out for bilingual corpus. 2 Data The data used in this work are from the SemEval 2010 challenge Task 5, automatic keyphrase ex- traction from scientific articles. The data consist of train, trial, and test data sets. The number of scientific articles and the total number of word to- kens in …

Content truncated to 500 words