Noun Compound and Named Entity Recognition and their Usability in Keyphrase Extraction

Proceedings of Recent Advances in Natural Language Processing, pages 162–169, Hissar, Bulgaria, 12-14 September 2011. Noun Compound and Named Entity Recognition and their Usability in Keyphrase Extraction Istv´an Nagy T.1, G´abor Berend1 and Veronika Vincze2 1Department of Informatics, University of Szeged {nistvan,berendg}@inf.u-szeged.hu 2Hungarian Academy of Sciences, Research Group on Artificial Intelligence vinczev@inf.u-szeged.hu Abstract We investigate how the automatic identi- fication of noun compounds and named entities can contribute to keyphrase ex- traction and we also show how previously identified noun compounds affect named entity recognition and vice versa, how noun compound detection is supported by identified named entities. Our exper- iments demonstrate that already known noun compounds yield better performance in named entity recognition and already known named entities enhance noun com- pound detection. The integration of noun compound and named entity related features into a keyphrase extractor also proves to be more effective than the model not including them. Our results indicate that the above features tend to be benefi- cial in several NLP-related tasks. 1 Introduction In natural language processing, the proper treat- ment of multiword expressions (MWEs) is essen- tial for many higher-level applications (e.g. infor- mation extraction or machine translation). Multi- word expressions are lexical items that can be de- composed into single words and display idiosyn- cratic features (Sag et al., 2002), in other words, they are lexical items that contain space. They are frequent in language use and usually exhibit unique and idiosyncratic behavior, thus, they of- ten pose a problem to NLP systems. Named enti- ties (NEs) are another class of linguistic elements that require special treatment in many NLP sys- tems ranging from information retrieval to ma- chine translation. In this paper, we demonstrate how the automatic identification of noun compounds and named en- tities can contribute to keyphrase extraction and we also investigate how previously identified noun compounds affect named entity recognition (NER) and vice versa, how noun compound detection is supported by identified named entities. We briefly describe our methods, then discuss our results in detail. We argue that previous knowledge of noun compounds can enhance keyphrase extraction and NER while previously identified NEs can con- tribute to noun compound identification. We be- lieve that employing NE- and noun compound- related features in other higher-level applications will also enhance performance. 2 Noun compounds and named entities in NLP applications A compound is a lexical unit that consists of two or more elements that exist on their own. Com- pounds can be classified as follows (Sag et al., 2002; Kim, 2008): nominal compounds (bass player), adjectival compounds (dark skinned), ad- verbial compounds (all in all), prepositional com- pounds (in front of), and multiword conjunctions (in order that). Named entity recognition is another widely re- searched topic in NLP. There are several meth- ods developed for many languages and domains (Grishman and Sundheim, 1995; Chinchor, 1998; Tjong Kim Sang, 2002; Tjong Kim Sang and De Meulder, 2003). Multiword named entities can be composed of any words or even characters and their meaning …

Content truncated to 500 words