Named entity disambiguation using linked open data

Named entity recognition is the task of identifying named entities such as persons, organizations and locations in natural language texts. One of the problems in named entity recognition is that different entities having the same or similar names, that is the ambiguity problem. This problem requires a decision on choosing the right entity among a number of possible entities with the same or similar names. Human readers do this naturally when reading a text, because they are aware of the context. But, for the software to do this, a ?named entity disambiguation? task needs to be executed to decide about the right entities. There are many techniques developed for named entity recognition and disambiguation tasks in the literature. A recent approach in these tasks is using open knowledge bases such as Wikipedia, and more recently the structured linked data counterparts that are derived from Wikipedia and similar knowledge bases, such as DBpedia, YAGO, and Freebase. In this thesis, we present a named entity disambiguation technique that is based on using Linked Open Data and a graph centrality algorithm for disambiguation. The system we developed is called NERSO. We evaluated our system using publicly available data sets, and compare its performance with other named entity recognition and disambiguation (or annotation) tools, and present the results in this thesis. Our results show that NERSO performs either similar or better than the other tools.