Design and development of a system based on short questions for retrieving relevant documents that express opinion.

Abstract

Opinion mining is an interesting area of research because of its applications in various fields. Collecting opinions of people about products or social events and problems through the Web is becoming increasingly popular every day. The views of users are beneficial regarding both the public and stakeholders in terms of decision-making. Opinion mining is a way to retrieve information through search engines, Web blogs, and social networks. Summarizing information manually constitutes a time-consuming process due to the huge number of reviews in the form of unstructured text. Hence, effective and efficient computational methods in terms of mining and summarizing reviews from corpuses and web documents, are of critical importance. This study aims at designing and implementing a system that will consist of an ensemble of different documents that express positive, negative, or even a neutral view about an issue, a product, or a person. Thus, the above-mentioned system will generate results, ranked by certain criteria (e.g relevance, date, reliability) and corresponding to users’ query. The latter will be in the form of filters, keywords, or even a combination of them. More specifically, many different methods of pre-processing will be used for the summarizing of reviews and user’s query. Also, extraction of aspects and, in case of their detection to the user’s query, expansion of them based on thesauri will be two main procedures of the system. Finally, the conduction of a variety of experiments while the use of the statistical measure Term Frequency-Inverse Document Frequency were essential for the evaluation of the system.