Bipolar Person Name Identification of Topic Documents Using Principal Component Analysis

Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 170–178, Beijing, August 2010 Bipolar Person Name Identification of Topic Documents Using Principal Component Analysis Chein Chin Chen Department of Information Management National Taiwan University paton@im.ntu.edu.tw Chen-Yuan Wu Department of Information Management National Taiwan University r97725035@ntu.edu.tw Abstract In this paper, we propose an unsuper- vised approach for identifying bipolar person names in a set of topic documents. We employ principal component analysis (PCA) to discover bipolar word usage patterns of person names in the docu- ments and show that the signs of the en- tries in the principal eigenvector of PCA partition the person names into bipolar groups spontaneously. Empirical evalua- tions demonstrate the efficacy of the proposed approach in identifying bipolar person names of topics. 1 Introduction With the advent of Web2.0, many online colla- borative tools, e.g., weblogs and discussion fo- rums are being developed to allow Internet users to express their perspectives on a wide variety of topics via Web documents. One benefit is that the Web has become an invaluable knowledge base for Internet users to learn about a topic comprehensively. Since the essence of Web2.0 is knowledge sharing, collaborative tools are generally designed with few constraints so that users will be motivated to contribute their know- ledge. As a result, the number of topic docu- ments on the Internet is growing exponentially. Research subjects, such as topic threading and timeline mining (Nallapati et al., 2004; Feng and Allan, 2007; Chen and Chen, 2008), are thus being studied to help Internet users comprehend numerous topic documents efficiently. A topic consists of a sequence of related events associated with a specific time, place, and person(s) (Nallapati et al., 2004). Topics that involve bipolar (or competitive) viewpoints are often attention-getting and attract a large number of topic documents. For such topics, identifying the polarity of the named entities, especially per- son names, in the topic documents would help readers learn the topic efficiently. For instance, for the 2008 American presidential election, In- ternet users can find numerous Web documents about the Democrat and Republican parties. Identifying important people in the competing parties would help readers form a balanced view of the campaign. Existing works on topic content mining focus on extracting important themes in topics. In this paper, we propose an unsupervised approach that identifies bipolar person names in a set of topic documents automatically. We employ principal component analysis (PCA) (Smith, 2002) to dis- cover bipolar word usage patterns of important person names in a set of topic documents, and show that the signs of the entries in the principal eigenvector of PCA partition the person names in bipolar groups spontaneously. In addition, we present two techniques, called off-topic block elimination and weighted correlation coefficient, to reduce the effect of data sparseness on person name bipolarization. The results of experiments based on two topic document sets written in English and Chinese respectively demonstrate that the proposed PCA-based approach is effec- tive in identifying bipolar person …

Content truncated to 500 words