Question answering system focusing on sentiment analysis using social networks and public media in the context of market research

Abstract

The Internet has earned much credit for being an easily accessible resource and exchange of information. Many computer science researchers focus their attention to the efficiency of search engine queries, information retrieval and answer extraction. Due to the rise in Internet usage, the birth of social networks and public media has now become part of our daily lives. Social networks and public media have gained popularity through the recent years due to the ease of expressing oneself freely. Most of these expressions of opinion are available online for a vast amount of users to access and read. Market researchers have for years invested millions of hours and monetary figures into obtaining infonnation which will help their product grow. The usual technique has been manually creating questionnaires which are sent to many people in order to gather their opinions about a product. Market research is definitely not a new field and is usually a main trigger of higher profits. The problem with market research is that this task is often time-consuming which may cost them a lot of money in the long term. This work attempts to enrich the market research experience with more easily obtained results and reduce the amount of time used to gather this information. Market research will be achieved using the efforts in question answering and information retrieval techniques whilst using social networks and public media as the main source of information. Due to the nature of opinion-type questions, this work will take a special consideration in the fields of sentiment analysis. The proposed system is a Question Answering System which handles questions requiring a more opinionoriented result, often the selected choice for market research. The targeted audience is online users and thus, for this reason, a special focus will be placed on sentiment analysis obtained from public media and social networks where many human opinions are readily found. The main modules of a question answering system will be discussed briefly. These modules are: Question Processing, Information Retrieval, Answer extraction including Sentiment Analysis, and Answer Composition and Evaluation.