Massive Scale Text Processing Using Phrases

Sehda plans to separate relevant data from the irrelevant through a technique in which every document being processed would be marked automatically with contextual meta-data to characterize the genre, author, and information in the article. Application of Sehda's technology to the results of this categorization scheme should permit the automatic extraction of information responsive to any inquiry in a natural, flexible, manner far more effective than current key word searches at identifying and grading relevant responsive material. Sehda will also build a browser which will allow the user to navigate this data seamlessly, by examining different cuts through the data in order to narrow in on relevant associations the viewer wishes to focus on. Finally, we have devised potential ways to do "Rare Event Detection" using statistical analysis of the content of unusual responses, communications or word associations, although this idea is only at the formative stage, and would require additional research and testing.