Estimating Similarity Among Entities Aided by the Web when Only the Entity Name is Available

• Priscila Sad de Sousa IF Sudeste MG
• Anderson A. Ferreira UFOP

Resumo

Estimating the similarity between entity names plays an important role in several tasks, such as entity resolution and recommendation tasks. Identifying the similarity between entity names, such as between titles of scientific articles, may not be feasible from direct comparison or using knowledge-based similarity approaches. Being an immeasurable source of data, Web can aid in this similarity check. In this work, we propose a method to calculate the similarity between two values of textual names, based on features inferred from data obtained from the Web and with the aid of genre terms. Experiments show that the method is able to check the similarity between names even those names no share terms in common.

Palavras-chave: Data Integration, Entity Resolution, Similarity among Entities, Web Text Analysis