Ontologising relational triples into a portuguese thesaurus

Authors

Abstract

Having in mind the automatic acquisition and integration of knowledge from different heterogeneous resources, this paper proposes several automatic methods for attaching term-based relational triples to the synsets of a thesaurus, without exploiting the extraction context for disambiguation. After using the proposed methods to attach triples, extracted from a Portuguese dictionary, to the synsets of a Portuguese thesaurus, two perspectives on their performance are given. The resulting synset-based triples were automatically validated based on support found in corpus and were then evaluated based on a handcrafted gold resource.

Keywords

ontologising, wordnet, semantic relations

Subject

Natural Language Processing

Related Project

Onto.PT

Conference

15th Portuguese Conference on Artificial Intelligence (EPIA 2011), Lisbon, Portugal 2011