SBIR/STTR Research Award
Centering Resonance Analysis: A superior data mining algorithm for textual data streams
Award amount
$100,000Program and phase
STTR · Phase IAward year
2003Award details
- Awarding agency
- Department of Defense
- Branch
- Air Force
- Recipient company
- Crawdad Technologies, L.L.C.
- Proposal award date
- Not reported
- Contract end date
- Not reported
- UEI
- Not reported
- Solicitation number
- Not reported
- Topic code
- Not reported
Research abstract
Crawdad Technologies, L.L.C. is an R&D firm located in Chandler, Arizona that specializes in quantitative communication analysis. Their first product Crawdad 1.0 is based on Centering Resonance Analysis (CRA; patents pending), a radical new way to modeland analyze text. CRA models texts as networks, and uses sophisticated methods to perform a variety of analyses. CRA is the only text analysis approach that creates a rich, high-precision model of text, while eliminating the need for complicated semanticrules, training sets, or corpora. Further research is needed to develop Crawdad capabilities to deal with large amounts of rapidly streaming textual data, and further commercialization efforts are required to migrate to an enterprise level solution. Thepurpose of Phase I is to investigate CRA's ability to perform data mining of streaming news articles, under storage constraints. The final product of Phase I will be a feasibility report that will contain (a) an architectural design for Crawdad 2.0, thatwill operate in a environment of streaming texts (e.g. news articles, emails), (b) the storage and computational requirements associated with CRA, and (c) measures of the ability of CRA to determine text relevancy against a query criterion. The proposedresearch will significantly advance our knowledge of how the discourse-based approach can facilitate effective data mining and filtering in a real-time, storage-constrained environment. Phase II activities will draw from the results of Phase I research todevelop Crawdad 2.0. Phase I focus on news articles will not detract from expanding the focus of Phase II, as Crawdad works equally well on different types of unstructured textual data. Phase II research will focus on issues involved with multiplelanguages and document types, trend analysis, statistical process control, and extension into areas such as article summarization and topic detection and tracking. In parallel with Phase II, our commercialization efforts will focus on system integrationissues and scalability. Commercialization plans include collaborating with a large systems integration firm in co-marketing and beta testing via an existing agreement already in place.
Technology keywords
Not reported