Aquisição de Conhecimento de Mundo para Sistemas de Processamento de Linguagem Natural

Abstract

One of the challenges of research in Natural Language Processing(NLP) is to provide semantic and linguistic resources to express knowledge of the world to support tasks such as Information Extraction, Information Retrieval systems, Questions & Answering, Text Summarization, Annotation Semantics of texts, etc. For this challenge this work proposes strategies for acquiring knowledge of the world. We propose two methods. The first is a semi-automatic method that has main idea of using a semantic reasoning process on pre-existing knowledge base semantics. The second is an acquisition method that utilizes automatic Wikipedia for generating semantical content. Wikipedia was used as a source of knowledge because of the reliability, dynamism and scope of its content. In this work we propose a method for acquiring semantic relations between concepts from the texts of Wikipedia articles that makes use of an implicit knowledge that exists in Wikipedia and in hypermedia systems: links between articles. Throughout the descriptive text of a Wikipedia article appear links to other articles that are evidence that there is a relationship between the current article and another article referenced by the link. The proposed method aims to capture the semantic relationship expressed in the text between them (current article and link to another article), no regular expressions identifying similar relationships through a semantic similarity measure.