Exploiting Crowdsourcing to Seed Organization Profile Scouting

We propose a new approach to the automatic mining and validation of organizational data from available sources like the Web called organization profiling that will enable the generation of a persistent and up-to-date database of organizational attributes and relationships of interest to an analyst. Building on our existing core technologies in context-aware information retrieval and ontology-based information extraction, the Scout system will automatically retrieve source documents likely to contain desirable attribute information and extract those attributes from unstructured and semi-structured text. We will utilize an innovative crowdsourcing approach to quickly and efficiently seed Scout with a baseline snapshot-in-time database for system training, monitoring, and evolution. Cross-source validation techniques will determine the interpretation of extracted values relative to an evolving, consensus understanding of each attribute including a summary of its values through time. Both user-targeted and automatically detected attributes will be collected and packaged in a new data structure called an Organization Profile. Streamlined user workflows that minimize user involvement and maximize system utility will be developed to permit the targeting of attributes for extraction as well as the efficient review and validation of extracted data. Phase I will prove the feasibility of our overall approach via an end-to-end demonstration prototype. BENEFIT: The proposed Scout system will have a wide range of applications in areas where analysts face serious challenges in collecting and fusing data hidden in a sea of electronic text. Market Research and Business Intelligence are particularly attractive applications where we see the opportunity for Scout to dramatically improve the ability to plan product development and marketing strategies and monitor competitors.