Recent advances in temporal mining techniques and the proliferation of text documents have opened up possibilities as well as the need for mining temporally significant information from text repositories. The goal of this project is to find news ways to extract meaningful temporal relationships from unstructured text. The approach is based on temporal segmentation of document collections to detect time-varying nature of term use in the collection (e.g., apple or window). This technique will allow the system to do topic tracking and topic detection, in addition to time-sensitive queries. A semantically rich temporal query language that uses keywords, temporal literals, propositional temporal operators as well as Boolean is developed and implemented in an experimental system Chronomine. A novel temporal indexing structure is developed to support efficient querying. The results of this project include the formal semantics and a language for temporal mining of text documents and a system that is capable of temporal decision support from text collections. The project provides educational as well as research opportunities for both graduate and undergraduate students. The research results are incorporated into a prototype that will be freely available on the Web (http://faculty.ist.unomaha.edu/pchundi) for anyone to analyze document sets by uploading them. The source code of the prototype will be also made available to the community at large by making it an open source.
