next up previous
Next: Experiment Up: WordSieve: A Method for Extraction Previous: Context Profiles


We have evaluated the performance of the algorithm for the ability to index documents according to a search task given to the user, where the context is defined as a topic given to the users, but unavailable to WordSieve. In our tests, WordSieve is able to more strongly match documents to task context than TFIDF. The software used for data collection in our experiment is CALVIN, an Intelligent Agent for research assistance [13]. This system is designed to observe a person using a computer and suggest resources that have been helpful in similar contexts in the past. CALVIN uses WordSieve to index documents.[*]


Travis Bauer