Semantic Retrieval for the Accurate Identification of Relational Concepts in Massive Textbases

Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 1017–1024, Sydney, July 2006. c⃝2006 Association for Computational Linguistics Semantic Retrieval for the Accurate Identification of Relational Concepts in Massive Textbases Yusuke Miyao∗ Tomoko Ohta∗ Katsuya Masuda∗ Yoshimasa Tsuruoka† Kazuhiro Yoshida∗ Takashi Ninomiya‡ Jun’ichi Tsujii∗† ∗Department of Computer Science, University of Tokyo †School of Informatics, University of Manchester ‡Information Technology Center, University of Tokyo Hongo 7-3-1, Bunkyo-ku, Tokyo 113-0033 JAPAN {yusuke,okap,kmasuda,tsuruoka,kyoshida,ninomi,tsujii}@is.s.u-tokyo.ac.jp Abstract This paper introduces a novel framework for the accurate retrieval of relational con- cepts from huge texts. Prior to retrieval, all sentences are annotated with predicate argument structures and ontological iden- tifiers by applying a deep parser and a term recognizer. During the run time, user re- quests are converted into queries of region algebra on these annotations. Structural matching with pre-computed semantic an- notations establishes the accurate and effi- cient retrieval of relational concepts. This framework was applied to a text retrieval system for MEDLINE. Experiments on the retrieval of biomedical correlations re- vealed that the cost is sufficiently small for real-time applications and that the retrieval precision is significantly improved. 1 Introduction Rapid expansion of text information has motivated the development of efficient methods of access- ing information in huge texts. Furthermore, user demand has shifted toward the retrieval of more precise and complex information, including re- lational concepts. For example, biomedical re- searchers deal with a massive quantity of publica- tions; MEDLINE contains approximately 15 mil- lion references to journal articles in life sciences, and its size is rapidly increasing, at a rate of more than 10% yearly (National Library of Medicine, 2005). Researchers would like to be able to search this huge textbase for biomedical correla- tions such as protein-protein or gene-disease asso- ciations (Blaschke and Valencia, 2002; Hao et al., 2005; Chun et al., 2006). However, the framework of traditional information retrieval (IR) has diffi- culty with the accurate retrieval of such relational concepts because relational concepts are essen- tially determined by semantic relations between words, and keyword-based IR techniques are in- sufficient to describe such relations precisely. The present paper demonstrates a framework for the accurate real-time retrieval of relational concepts from huge texts. Prior to retrieval, we prepare a semantically annotated textbase by ap- plying NLP tools including deep parsers and term recognizers. That is, all sentences are annotated in advance with semantic structures and are stored in a structured database. User requests are con- verted on the fly into patterns of these semantic annotations, and texts are retrieved by matching these patterns with the pre-computed semantic an- notations. The accurate retrieval of relational con- cepts is attained because we can precisely describe relational concepts using semantic annotations. In addition, real-time retrieval is possible because se- mantic annotations are computed in advance. This framework has been implemented for a text retrieval system for MEDLINE. We first ap- ply a deep parser (Miyao and Tsujii, 2005) and a dictionary-based term recognizer (Tsuruoka and Tsujii, 2004) …

Content truncated to 500 words