Question Answering: Technology For Intelligence Analysis

This chapter presents a former analyst’s perspective on Question Answering (QA) systems. Although the author’s background is in intelligence analysis, he is currently engaged in sponsoring basic research in linguistic pragmatics. Given this background, the author believes that today’s technology has the potential to help fulfil the vision of providing a highly capable tool enabling an analyst – any analyst, be they journalists or financial experts on Wall Street – the ability to interact with data in a meaningful way, much as one would conduct a dialogue with another person, in order to answer questions or aggregate relevant information in the attempt. Currently, analysts derive most of the information they need from a large corpora of documents. The preferred means of doing so, for the most part, is a threestep process of collect, filter, and then process manually or semi-automatically. The predominant paradigm employed usually involves the Web, an information retrieval (IR) engine such as Google, and intense reading. Their real need, however, is to obtain specific answers to specific questions. The loftiness of the goal notwithstanding, the author believes that technology is progressing to the point where QA systems can address this need. Although this chapter is more a vision statement and less a research roadmap per se, the author does suggest areas of research that are pertinent to the vision, and discusses the possible impact a dialogue-based QA system might have on analysts in general. To this end, two scenarios are presented in the Appendix.