Abstract. This paper reports on preliminary research aiming at discovering and exploiting structure in real-life, dynamic text collections. The text collection is a technical library catalog in the domain of aerospace test and evaluation. The vigilance parameter of Adaptive Resonance Theory (ART) neural networks proved to be a useful structure detector, allowing for the determination of clustering tendency as well as discovery of useful hierarchical structures in a text collection. 1.

Abstract- Filtering the immense amount of data available electronically over the World Wide Web is an important task of search engines in data mining applications. Users when performing search often formulate hypotheses that they want to find supporting data for. The initial hypothesis reflects their preliminary knowledge of the subject. The final hypotheses at the end of the search reflect what they learned about a given subject and reflect the supporting information they found during search. We propose an evolutionary computation-based method that automatically generates queries and retrieves information to prove or disprove a given hypothesis. In case there is no supporting data, the system evolves another hypothesis for which it can find supporting data. We show preliminary results obtained for hypotheses related to plague where the data used is the Entrez PubMed data set.

There is a large and continually growing quantity of electronic text available, which contain essential human and organization knowledge. An important research endeavor is to study and develop better ways to access this knowledge. Text clustering is a popular approach to automatically organize textual document collections by topics to help users find the information they need. Adaptive Resonance Theory (ART) neural networks possess several interesting properties that make them appealing in the area of text clustering. Although ART has been used in several research works as a text clustering tool, the level of quality of the resulting document clusters has not been clearly established yet. In this paper, we present experimental results with binary ART that address this issue by determining how close clustering quality is to an upper bound on clustering quality.