New Applications for Information Extraction

We will present ongoing research at UMBC on extracting and exploiting information from novel sources. The first problem we will discuss is extracting and exploiting semantic information from structured forms like spreadsheets, database relations, and Web tables. We've developed an approach that uses Semantic Web knowledge to interpret such tables and associate their entities, values and relations with nodes in a reference linked data collection. The resulting interpretation can be used to annotate tables, confirm existing facts in the linked data collection, and propose new facts to be added. The second problem focuses on systems for detecting emerging events that can have a significant impact on society. With an increasingly "instrumented and interconnected" world, we can try to detect such events, and be predictive and proactive in our responses so as to mitigate their adverse impact. Situational Awareness is a key component of this process, enabling both the commanders thatdirect responses, and responders that carry them out, to be more efficient and effective. We will touch upon two related elements: using social media data streams (human-as-sensor) to supplement the more traditional physical sensors in event detection and using agent oriented approaches to help take data from a variety of sources and produce a common operational picture.