Cosema: Content-based Semantic Annotator release_rev_b01f9c8d-3d5e-4074-ad58-2c24ecb123a6

Abstract

In this paper, we present a library for creating automatic annotations for entities and concepts inside any textual content. The tool is based on DBpedia. In particular, the annotations are generated using the DBpedia link structure as a source of knowledge for Word Sense Disambiguation. DBpedia is used as a reference to obtain information on lexicographic relationships. By using such information in combination with statistical information extraction techniques, it is possible to deduce concepts related to the terms extracted from a corpus. Moreover, by combining statistical information extraction with named entity recognition and the use of the OKKAM ENS infrastructure, it is also possible to obtain unique annotations for entities in the content. The advantage of this approach, in addition of improving information retrieval and categorization capabilities, consists in the fact that the generate concept and entity annotations can be referred to with unique identifiers around the Web. For this reason different description for the same entity or concept can be semantically aggregated from the Web.
In text/plain format