Interpretação automatizada de textos : processamento de anáforas

• Autor(a) principal: Freitas, Sérgio Antônio Andrade de
• Data de Publicação: 2005
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal do Espírito Santo (riUfes)
• Texto Completo: http://repositorio.ufes.br/handle/10/4114
• topic: Eletrônica Industrial, Sistemas e Controles Eletrônicos; Processamento de textos (Computação); Algoritmos de computador; Anáfora (Linguística); Processamento da linguagem natural (Computação); 621.3

This thesis presents a solution to the interpretation of de nite descriptions in Portuguese. For example, consider the following text: (1) a. Mariana bought a new car. b. The engine was damaged. The sentence (1a) introduces two entities: Mariana and a car which is new. The sentence (1b) introduces only one entity the engine. In a human or computer interpretation process, the use of the de nite article the preceeding a noun indicates that the introduced entity was already present at the discourse, i.e., it is an anaphoric entity. The resolution of an anaphora is a reference problem, but in the example (1) there is another problem: although the car is the entity that gives context to the engine, we can not say that the engine is the car (as for a pronominal anaphora). It also must be determined how the engine is related to the car. This is a de nite description problem. The interpretation of any kind of anaphora can be represented by the following equation: R(A, T ) (2) where A denotes an entity introduced by the context interpretation of a pronoun, an ellipsis or a de nite noun phrase, T denotes its antecedent and R is the relation between A and T . The equation's resolution process is summarized as: given A nd T and R. This thesis proposes a methodology to the de nite description interpretation that the relation R is of: part of, member of, subcategorized by and corefers. These relations are obtained by a set of pragmatic rules [Freitas, Lopes e Menezes 2004, Filho e Freitas 2003], which are here de ned (chapter 3). Also if A is not anaphoric then it is acommodated in the discourse context. The computational methodology is implemented in a logic programming system [Damásio, Nejdl e Pereira 1994] that permits an abductive reasoning [Kakas, Kowalski e Toni 1992] at the semantic representation of the discourse [Kamp e Reyle 1993]. The interpretation of the entities is the basis to the Discourse Nominal Structure [Lopes e Freitas 1994] (chapter 4), which allows: (1) to track the most salient entities at each sentence [Freitas e Lopes 1994], (2) to limit the number of possible antecedents [Freitas e Lopes 1996] and (3) to give a discourse entities summary. The result is an integrated metodology to solve anaphors and ellipses. Finally, the Nominal Structure of the Discourse can help the search/index of digital documents.