WikiSense: Supersense Tagging of Wikipedia Named Entities Based WordNet

WikiSense: Supersense Tagging of Wikipedia Named Entities Based WordNet1 Joseph Chang a, Richard Tzong-Han Tsai a, and Jason S. Chang b aDepartment of Computer Science and Engineering, Yuan Ze University No. 135 Yuan-Tung Road, Chungli, Taoyuan, Taiwan 32003 {s951533, thtsai}@mail.yzu.edu.tw bDepartment of Computer Science, National Tsing Hua University No. 101, Section 2, Kuang-Fu Road, Hsinchu, Taiwan 30013 jschang@cs.nthu.edu.tw Abstract. In this paper, we introduce a minimally supervised method for learning to classify named-entity titles in a given encyclopedia into broad semantic categories in an existing ontology. Our main idea involves using overlapping entries in the encyclopedia and ontology and a small set of 30 handed tagged parenthetic explanations to automatically generate the training data. The proposed method involves automatically recognizing whether a title is a named entity, automatically generating two sets of training data, and automatically building a classification model for training a classification model based on textual and non-textual features. We present WikiSense, an implementation of the proposed method for extending the named entity coverage of WordNet by sense tagging Wikipedia titles. Experimental results show WikiSense achieves accuracy of over 95% and near 80% applicability for all NE titles in Wikipedia. WikiSense cleanly produces over 1.2 million of NEs tagged with broad categories, based on the lexicographers’ files of WordNet, effectively extending WordNet to form a very large scale semantic category, a potentially useful resource for many natural language related tasks. Keywords: semantic category, word sense disambiguation, WordNet, Wikipedia 1 Introduction Machine readable natural language resources, like ontologies or semantic categories, are crucial in natural language processing or information retrieval tasks that demands world knowledge. Such tasks include question classification and answering, knowledge mining and semantic search. In these tasks, lexical databases such as WordNet or Suggested Upper Merged Ontology (SUMO) are widely used to provide meaning representation and semantic relations. These handcrafted ontologies rely on manual compilation and maintenance of small groups of experts over long periods of time. New words and phrases are added to the vocabulary progressively over the years through each new release. With the scale close to that of a dictionary, these resources consists mostly of common nouns, verbs, adjectives, adverbs, and a small amount of named entities. Only highly well-known people, places, organizations, and events, are included. Although these resources are very useful for many natural language tasks, they sometimes suffer severely from the out-of-vocabulary problem due to limited coverage, most notably for lack of NEs. Furthermore, for NEs they cover, there are also issues with consistency and relevance. For example, Charles Dickens is found in WordNet, but none of the books he wrote (e.g., A Tale of Two Cities). The generally accepted lexicographer guidelines called for using the powerful tool of frequency to make informed linguistic decisions about creating and framing an entry. A typical Web search (e.g., Google, http://www.google.com) show Celine Dion, a contemporary Canadian pop singer, has slightly higher visibility on the Web than Johnny Cash, the late American country singer. (13,100,000 v.s. 11,600,000 page counts) However, Celine Dion is …

Content truncated to 500 words