Entity based co-reference resolution with name entity recognition using hierarchical classification

Abstract

Co-reference resolution is a process of finding relation of mentions in the discourse that refers to the same entity. Name entity recognition (NER) is the process of determining and classifying the names of people, organizations, locations and other entities within text. Classification of entity is essential to identify correct co-referent. Hence, NER is one of the key components in co-reference resolution. A novel approach to Co-reference resolution and NER is proposed in this paper. Many systems are restricted to produce labels from to a small set of entity classes, e.g., person, organization, location or miscellaneous. In order to intelligently understand text and extract a fine range of information, it is useful to more precisely determine the conceptual classes of entities mentioned in unstructured text. This paper presents 20 fine grain classes using concept hierarchy for classification of 5 classes. Wikipedia is used to recognise fine grain entity. Huffman encoding mechanism is used to encode according to classes in the annotated text file. The process of quotation attribute handling, the mention detection model and pronoun resolution is proposed which makes use of Web scale N-Grams corpus and precise syntactic features set. Our techniques are evaluated on text dataset resulting improvement over the state of the art methods.