Abstract

The goal of Question-Answering (QA) systems is to find short and correct answers to open-domain questions by searching a large collection of documents. The subject of this research is focused on finding patterns to formulate a "complete" and "natural" answer to questions, given the short answer. Finding such patterns is important as they can be used to enhance existing QA systems to find and provide answers to the user in a more "natural way" and providing a pattern to find the answer. Based on a number of patterns of type of answer formulation compiled from a survey carried out at the beginning of this project, the work of producing long and natural answers to specific type of questions is reduced to patterns matching, but additional research could be done to process and analyze the context of the question and the short answer given, in order to provide a more relevant, natural and correct answers. The first chapter of this major report gives a general description of the nature and scope of our project, the second chapter introduces the fields of Natural Language Generation (NLG) and Question-Answering systems (QA) as a background to our system, which we expect to be used as a resource to increase the potential of QA system, to answer questions of a wide variety of topics in different grammatical and natural ways. The third chapter presents the process, analysis and results of the survey applied to find the answer formulation patterns. The fourth chapter describes the system, its requirements, scope, analysis and results, and finally chapter five shows the evaluation of the system and gives the conclusions and future research avenues.