The Data-driven Intelligent System Research Center (DIRECT) develops natural language processing technology that makes it possible to automatically discover valuable combinations of information scattered in many places, and to automatically generate important hypothetical knowledge through the deep semantic analysis of large volumes of data, including the Web. Recent events like Japan's tsunami and the ensuing nuclear incident have demonstrated the potentially disruptive effects of big crisis events across geographically, socially and semantically distinct spheres of society. In such social crisis scenarios it is crucial to be able to analyze gigantic data streams on a semantic level, and to allow people to efficiently explore unexpected but possible implications of given events and information, and formulate relevant hypotheses and likely future scenarios. Our goal is to realize these objectives using information technology. To accomplish this goal we conduct research on natural language processing, text mining, various types of inference mechanisms and multimedia information processing. We focus particularly on deep semantics, on a level that past technologies could not handle.