Semantic Name Matching

Names are studied in different fields, and, among the issues they present,name variations (e.g., translations, misspellings, etc...) and name variants (e.g., pseudonyms) pose a challenge to name matching, i.e., discovering instances that differ typographically but represent the same entity. Our scenario for name matching is a P2P, entity-based network of users divided in local level (the users), community level (groups of users), and global level(all the entities). Entities at local level are a partial view of the real word entity, represented at the global level. In this framework, name variations and name variants change the orthography of names because of linguistic and social factors, and their presence depends on the scenario level considered. Thus, they are hard to tackle by an automatic approach such as name matching. Our proposed solutions is to use a taxonomy we created to understand and predict the variations and variants of different entity names, and divide the entity name in different entries to accommodate the original name plus variations and variants. Our approach is novel because we take advantage of a multidisciplinary method, drawing from various fields (i.e., philosophy, sociology and geography) importing terms and views not found in computer science. We also draw from areas close to name matching, building from their findings and expanding them.