Named entity recognition and disambiguation using linked data and graph-based centrality scoring

Abstract

Named Entity Recognition (NER) is a subtask of information extraction and aims to identify atomic entities in text that fall into predefined categories such as person, location, organization, etc. Recent efforts in NER try to extract entities and link them to linked data entities. Linked data is a term used for data resources that are created using semantic web standards such as DBpedia. There are a number of online tools that try to identify named entities in text and link them to linked data resources. Although one can use these tools via their APIs and web interfaces, they use different data resources and different techniques to identify named entities and not all of them reveal this information. One of the major tasks in NER is disambiguation that is identifying the right entity among a number of entities with the same names; for example "apple" standing for both "Apple, Inc." the company and the fruit. We developed a similar tool called NERSO, short for Named Entity Recognition Using Semantic Open Data, to automatically extract named entities, disambiguating and linking them to DBpedia entities. Our disambiguation method is based on constructing a graph of linked data entities and scoring them using a graph-based centrality algorithm. We evaluate our system by comparing its performance with two publicly available NER tools. The results show that NERSO performs better. © 2012 ACM.