An Adaptive Method for Organization Name Disambiguation with Feature Reinforcing

Copyright 2012 by Shu Zhang, Jianwei Wu, Dequan Zheng, Yao Meng, and Hao Yu 26th Pacific Asia Conference on Language,Information and Computation pages 237–245 An Adaptive Method for Organization Name Disambiguation with Feature Reinforcing Shu Zhang1, Jianwei Wu2, Dequan Zheng2, Yao Meng1 and Hao Yu1 1 Fujitsu Research and Development Center Dong Si Huan Zhong Rd, Chaoyang District, Beijing and 0086, China {zhangshu, mengyao, yu}@cn.fujitsu.com School of Computer Science and Technology, Harbin Institute of Technology No.92, Xidazhi Street, Harbin 150001, China {jwwu, dqzheng}@mtlab.hit.edu.cn Abstract Twitter is an online social networking, which has become an important source of information for marketing strategies and online reputation management. In this paper, we probe the problem of organization name disambiguation on twitter messages. This task is challenging due to the fact of lacking sufficient information both from organization and the tweets. We mine organization information from web sources to train a general classifier. Further, we mine tweets information. We train an adaptive classifier for a given organization name with more features derived from twitter messages labeled by the general classifier. The experiments on WePS-3 show mining web sources to enrich organization are effective. The adaptive classifier trained for a given organization is promising. 1 Introduction Twitter is an online social networking and microblogging service, which rapidly gained worldwide popularity, with 140 million active users as of 2012 1, generating over 340 million tweets and handling over 1.6 billion search queries 1 http://blog.twitter.com/2012/03/twitter-turns-six.html per day2. People share their opinions on almost anything on Twitter, such as news, governmental policies, products and companies. Therefore, Twitter becomes an important information resource for the purpose of marketing strategies and online reputation management. How to retrieval, analyze and monitor Twitter information has been receiving a lot of attention in natural language processing and information retrieval research community (Kwak, et al., 2010; Boyd, et al.,2010; Tsagkias, et al., 2011). One of the essential things of these researches is first to get the information which is related to the studied entity, such as product, company, or certain event. This work is caused by the ambiguity of entities. For example, the name of company “Apple” has a separate meaning referring to one kind of fruit. The word “Amazon” could be used to refer river or company. Therefore, when the entity name is ambiguous, filtering spurious name matches is important to accurate detection and analysis of contents that people say about the given entity. This paper focuses on finding related tweets to a given organization. Assuming that tweets are retrieved by the query of organization name, such as “apple”, the task is to identify whether a tweet is relevant to the target organization (“Apple Inc.”) or not. Yerva et al. (2010) adopt support vector machines (SVM) classifier to classify tweets with external resources. Yoshida et al. (2010) classify 2 http://engineering.twitter.com/2011/05/engineering- behind-twitters-new-search.html 237 Copyright 2012 by Shu Zhang, Jianwei Wu, Dequan Zheng, Yao Meng, and Hao Yu 26 th Pacific Asia Conference on Language, Information and Computation pages 237-245 organization names into “organization-like names” …

Content truncated to 500 words