Name Disambiguation Using Semi-supervised Topic Model

Name ambiguity is increasingly attracting more attention. With the development of information available on the Web, name disambiguation is becoming one of the most challenging tasks. For example, some persons may share the same personal name. In order to address this problem, topic coherence principle is used to eliminate ambiguity of the name entity. A semi-supervised topic model (STM) is proposed. When we search online, many irrelevant documents always return to users. Wikipedia hierarchical structure information enrich the semantics of the name entity. Information extracted from Wikipedia is sorted out and put in the knowledge base. It is used to match the query entity. By utilizing the context of the given query entity, we attempt to disambiguate various meanings with the proposed model. Experiments on two real-life datasets, show that STM is more superior than baselines (ETM and WPAM) with accuracy 84.75 %. The result shows that our method is promising in name disambiguation as well. Our work can provide invaluable insights into entity disambiguation.