Supervised and Semi-supervised Methods based Organization Name Disambiguity

Supervised and Semi-supervised Methods based Organization Name Disambiguity Shu Zhang a, Jianwei Wu b, Dequan Zheng b, Yao Meng a, Yingju Xia a, and Hao Yu a aFujitsu Research and Development Center Dong Si Huan Zhong Rd, Chaoyang District, Beijing and 0086, China {zhangshu, mengyao, yjxia, yu}@cn.fujitsu.com bSchool 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 a widespread social media, which rapidly gained worldwide popularity. Pursuing on the problem of finding related tweets to a given organization, we propose supervised and semi-supervised based methods. This is a challenging task due to the potential organization name ambiguity. The tweets and organization contain little information. The organizations in training data are different with those in test data, which leads that we could not train a classifier to a certain organization. Therefore, we induce external resources to enrich the information of organization. Supervised and semi- supervised methods are adopted in two stages to classify the tweets. This is a try to utilize both training and test data for this specific task. Our experimental results on WePS-3 are primary and encouraging, they prove the proposed techniques are effective in performing the task. Keywords: Twitter, name ambiguity, online reputation management. 1 Introduction Twitter is an online social networking and microblogging service, which rapidly gained worldwide popularity, with 200 million users as of 20111, generating over 200 million tweets and handling over 1.6 billion search queries per day2. How to manage this information to grasp the response of people to governmental policies, the feedback and comment of people on commercial products have received considerable attention in research community. There are some researches such as opinion mining, online reputation management, which focus on monitoring user generated media. 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 paper focuses on finding related tweets to a given organization. This is a challenging task due to the potential organization name ambiguity. For example, the name of company “Apple” which has a separate meaning fruit apple. The word “Amazon” could be used to refer river or company. Filtering spurious name matches is important to effectively detect and analyze relevant contents that people say about the organization. To overcome the problem that the tweets and organizations contain little information, we induce external resources to enrich the information of organization. The organizations in training data are different with those in test data, which leads that we could not train a classifier to a certain organization. Therefore, supervised and semi-supervised methods are adopted in 1 http://www.bbc.co.uk/news/business-12889048 2 http://blog.twitter.com/2011/08/your-world-more-connected.html 25th Pacific Asia Conference on Language, Information and Computation, pages 615–621 615 two stages to classify the tweets. This is a try to utilize both training and test data for this specific task. The remainder of the paper is organized as follows: Section 2 describes the related work on name disambiguity. Section 3 gives …

Content truncated to 500 words