The Influence of Context during the Categorization and Discrimination of Spanish and Portuguese Person Names

This paper presents a method for fine-grained categorization and discrimination of person names on the basis of the semantic similarity information. We employ latent semantic analysis which establishes the semantic relations between the words of the context in which the named entities appear. We carry out several experimental studies in which we observe the influence of the context and the robustness of our approach with different number of examples. Our approach is evaluated with Spanish and Portuguese. The experimental results are encouraging, reaching 90% for the Spanish and 82% for the Portuguese person name categorization, and 80% for the Spanish and 65% for the Portuguese NE discrimination of six conflated names.