A Natural Language Question Answering System for Exploring Online Conversations

Abstract

The proliferation of social media has resulted in the exponential growth of on- line conversations. Due to the volume and complexity of conversations, it is often extremely difficult to gain insights from such conversations. This dissertation hy-pothesizes that synergetic integration of natural language processing with informa-tion visualization techniques can help users to better fulfill their information needs. More specifically, we developed a question-answering method that allows the user to ask questions about a conversation and then automatically answers the question by highlighting results in a visual interface. The visual interface, named ConVisQA, was developed by extending ConVis which visually summarizes a conversation by providing an overview of topics and sentiment information. We demonstrate the effectiveness of our approach through a user study with blog readers. The dis-sertation concludes with a user study comparing our interface with a traditional interface for blog reading as well as considerations for future work.