The NICT concept dictionary

Summary form only given. In this demonstration we present a system that guides a user´s information search (or knowledge discovery) by displaying, in a coordinated manner, many valuable keywords having important semantic relations to the user´s topic of interest. For example, the system contains large-scale databases of typical and not-so-typical semantic relations like “object/troubles”, “troubles/preventions”, “cause/effects” or “foods/health effects”, and so on. Clicking these keywords issues a usual Web search or leads to further recursive exploration of the semantic relation space, enabling the discovery of valuable “unknown unknowns” for the user. The semantic relations in the system are automatically acquired from a large collection of Web documents using state-of-the-art knowledge acquisition methods. These methods require only minimal human intervention, so the system can be easily customized to obtain many different kinds of relations and adapted to new target domains. Additionally, we demonstrate the system´s usefulnes for innovation support based on analogy and lexical word similarity. Finally, we demonstrate the system´s recently developed speech interface, which enables the user to perform on-the-fly information extraction from 600 million Web pages using natural language questions, in the form of an speech-enabled question answering (QA) system.