SHEF: Semantic Tagging and Summarization Techniques Applied to Cross-document Coreference

Proceedings of the 4th International Workshop on Semantic Evaluations (SemEval-2007), pages 292–295, Prague, June 2007. c⃝2007 Association for Computational Linguistics SHEF: Semantic Tagging and Summarization Techniques Applied to Cross-document Coreference Horacio Saggion Department of Computer Science University of Sheffield 211 Portobello Street - Sheffield, England, UK, S1 4DP Tel: +44-114-222-1947 Fax: +44-114-222-1810 saggion@dcs.shef.ac.uk Abstract We describe experiments for the cross- document coreference task in SemEval 2007. Our cross-document coreference sys- tem uses an in-house agglomerative clus- tering implementation to group documents referring to the same entity. Clustering uses vector representations created by sum- marization and semantic tagging analysis components. We present evaluation results for four system configurations demonstrat- ing the potential of the applied techniques. 1 Introduction Cross-document coreference resolution is the task of identifying if two mentions of the same (or similar) name in different sources refer to the same individ- ual. Deciding if two documents refer to the same individual is a difficult problem because names are highly ambiguous. Automatic techniques for solv- ing this problem are required not only for better ac- cess to information but also in natural language pro- cessing applications such as multidocument summa- rization and information extraction. Here, we con- centrate on the following SemEval 2007 Web Peo- ple Search Task (Artiles et al., 2007): a search en- gine user types in a person name as a query. Instead of ranking web pages, an ideal system should orga- nize search results in as many clusters as there are different people sharing the same name in the doc- uments returned by the search engine. The input is, therefore, the results given by a web search engine using a person name as query. The output is a num- ber of sets, each containing documents referring to the same individual. As past and recent research (Bagga and Baldwin, 1998; Phan et al., 2006), we have addressed the problem as a document clustering problem. For our first participation in SemEval 2007, we use two ap- proaches: a lexical or bag-of-words approach and a semantic based approach. We have implemented our own clustering algorithms but rely on available ex- traction and summarization technology developed in our laboratory to produce document representations used as input for the clustering procedure. 2 Clustering Algorithm We have implemented an agglomerative clustering algorithm. The input to the algorithm is a set of document representations implemented as vectors of terms and weights. Initially, there are as many clus- ters as input documents; as the algorithm proceeds clusters are merged until a certain termination condi- tion is reached. The algorithm computes the similar- ity between vector representations in order to decide whether or not to merge two clusters. The similar- ity metric we use is the cosine of the angle between two vectors. This metric gives value one for identi- cal vectors and zero for vectors which are orthogo- nal (non related). Various options have been imple- mented in order to measure how close two clusters are, but for the experiments reported here we have …

Content truncated to 500 words