Architecture of a question-answering system for a specific repository of documents

Abstract

This paper describes the architecture for a Question-Answering System for a repository of documents that are labeled according to a standard. This architecture defines the basic structure into three modules to ensure a recovery process of the appropriate response. The first module performs analysis operations and representation of the question entered by the user in natural language. The second module allows for the operations concerning the search for documents; and finally the third module performs the extraction of the response. For the analysis of a question, Natural Language Processing (NLP) techniques are used in order for the question to be represented in a logical manner. Also, other components are used such as ontology, taxonomy, and knowledge bases that could process the question, place it in a context, and determine the type of question. An information retrieval module that interacts with the document repository, it is used to obtain documents. These documents are analyzed to determine the paragraphs that could be considered as a response. The aim of the model allows for an answer to a question that is asked in natural language of a document repository. This proposal will establish an appropriate mechanism to enable components of the representation of information based on a natural language text and later retrieval. © 2010 IEEE.