Named Entity Disambiguation Based on Classified and Structural Semantic Relatedness

Named entity disambiguation is presented to solve the problem of name ambiguity. Traditional disambiguation methods merely use the word frequency to calculate the weights of attributes, yet ignore the important information from low frequency words. We propose a named entity disambiguation method based on classified and structural semantic relatedness. Structural semantic relatedness is computed by capturing the explicit semantic relatedness and the implicit structural semantic knowledge. Classified semantic relatedness is computed by main attributes which can determine the domain entity identity. The experimental results show our method can significantly improve the disambiguation performance and achieve 90.5% accuracy of disambiguation.