Semantic Driven Triplet Extraction From Unstructured Text

The increasing complexity of the web arises the need of tools able to extract knowledge from this huge amount of information. The tool proposed in this work allows automatic extraction of knowledge from Reuters agency news stream. The approach maps relevant terms in a semantic space and selects from the corpus a set of triplets of words in a subject-verb-object (s-v-o) form. The mapping into a semantic space allows for the identification of meaningful relations among the words, giving also a measure of the strength of these relations.