Explore Person Specific Evidence in Web Person Name Disambiguation

Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 832–842, Jeju Island, Korea, 12–14 July 2012. c⃝2012 Association for Computational Linguistics Explore Person Specific Evidence in Web Person Name Disambiguation Liwei Chen, Yansong Feng, Lei Zou, Dongyan Zhao Institute of Computer Science and Technology Peking University Beijing {clwclw88,fengyansong,zoulei,zhaodongyan}@pku.edu.cn Abstract In this paper, we investigate different usages of feature representations in the web person name disambiguation task which has been suf- fering from the mismatch of vocabulary and lack of clues in web environments. In litera- ture, the latter receives less attention and re- mains more challenging. We explore the fea- ture space in this task and argue that collecting person specific evidences from a corpus level can provide a more reasonable and robust es- timation for evaluating a feature’s importance in a given web page. This can alleviate the lack of clues where discriminative features can be reasonably weighted by taking their corpus level importance into account, not just relying on the current local context. We therefore pro- pose a topic-based model to exploit the person specific global importance and embed it into the person name similarity. The experimen- tal results show that the corpus level topic in- formation provides more stable evidences for discriminative features and our method out- performs the state-of-the-art systems on three WePS datasets. 1 Introduction Resolving ambiguity associated with person names found on the Web is a key challenge in many Internet applications, such as information retrieval, question answering, open information extraction, automatic knowledge acquisition(Wu and Weld, 2008) and so on. For example, if you want to know more about a guy named George Foster and feed Yahoo! with his name, the results are not satisfactory where you get more than 40 different persons named George Fos- ter scattering in the top 100 returned pages. None of the dominant search engines currently helps users group those returned pages into clusters according to whether they refer to the same person. Users thus have to either read those pages carefully or adjust their queries by adding extra modifiers. This moti- vates an intensive study in automatically resolving person name ambiguity in various web applications. However, resolving web person name ambiguity is not a trivial task. Due to the difficulties in fig- uring out or predicting the number of namesakes in the returned pages, the task has been investi- gated in an unsupervised learning fashion in the lit- erature, which is apparently different from the tra- ditional word sense disambiguation or entity link- ing/disambiguation tasks, where the inventories of candidate word senses or entities are usually known given the target word or entity mention. A general framework for this task can be formu- lated as first extracting various features from the web pages, and then grouping these pages into several clusters each of which is assumed to represent one specific person. Despite of the inevitably noisy na- ture of web data, a key challenge is how to handle …

Content truncated to 500 words