Exploratory Relation Extraction in Large Text Corpora

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 2087–2096, Dublin, Ireland, August 23-29 2014. Exploratory Relation Extraction in Large Text Corpora Alan Akbik Thilo Michael Database Systems and Information Management Group Einsteinufer 17, 10587 Berlin, Germany {firstname.lastname}@tu-berlin.de Christoph Boden Abstract In this paper, we propose and demonstrate Exploratory Relation Extraction (ERE), a novel approach to identifying and extracting relations from large text corpora based on user-driven and data-guided incremental exploration. We draw upon ideas from the information seeking paradigm of Exploratory Search (ES) to enable an exploration process in which users begin with a vaguely defined information need and progressively sharpen their definition of extraction tasks as they identify relations of interest in the underlying data. This process extends the application of Relation Extraction to use cases characterized by imprecise information needs and uncertainty regarding the information content of available data. We present an interactive workflow that allows users to build extractors based on entity types and human-readable extraction patterns derived from subtrees in dependency trees. In order to evaluate the viability of our approach on large text corpora, we conduct experiments on a dataset of over 160 million sentences with mentions of over 6 million FREEBASE entities extracted from the CLUEWEB09 corpus. Our experiments indicate that even non-expert users can intuitively use our approach to identify relations and create high precision extractors with minimal effort. 1 Introduction 1.1 Motivation and Problem Statement Relation Extraction (RE) is the task of creating extractors that automatically find instances of semantic relations in unstructured data such as natural language text (Riloff, 1996). An example extraction task might be to find instances of the EDUCATEDAT relation, which relates persons to their educational in- stitution and may include the entity pair <Sigmund Freud, University of Vienna> as relation instance. Motivated by an explosion of readily available sources of text data such as the Web, RE offers intriguing possibilities for querying and analyzing data as well as extracting and organizing the contained informa- tion (Sarawagi, 2008). As scalable computing architectures capable of processing ever larger amounts of data are being developed (Dean and Ghemawat, 2004) and dependency parsers are becoming more accurate and more robust (Petrov and McDonald, 2012), so rises the potential of developing means to directly access the structured information contained in natural language text. In spite of such positive trends however, currently established methods of creating relation extractors suffer from a number of limitations. The first is one of cost; the process of creating extractors requires either labeled data to be produced at sufficient quality and quantity in order to train a supervised machine learning algorithm (Culotta and Sorensen, 2004; Mintz et al., 2009), or the manual creation of a complex set of extraction rules (Str¨otgen and Gertz, 2010; Reiss et al., 2008). In either case, the process is tedious and time-consuming and requires trained specialists with an extensive background in NLP, rule-writing or machine learning (Chiticariu et al., 2013). Worse, this process needs to be repeated …

Content truncated to 500 words