Evaluating co-occurrence order for automatic thesaurus construction

Identifying the best semantically similar terms in an automatic thesaurus construction task is still an open problem in natural language processing. Many methods have been proposed to solve this problem. In this work we present a comparison between three corpus based methods for automatically build thesaurus. These methods look for related terms using the relations between terms and they differ among themselves in the co-occurrences order of these relations. The evaluation process was carried out by domain specialists who evaluated the related terms generated by each method in an experiment applied to the data privacy domain.