Final undergrad year, 2002.
Tibianna helps you search the web when you aren’t exactly sure how to formulate what you want in a query.
By ranking results, you can give Tibianna clues as to whether you want more of pages like this, but less of pages like that. Tibianna also allows you to narrow down your query to a particular meaning of the word. For example, it tells you that the word java can be used for a programming language, coffee and so on. With a click, you can dig down into one particular meaning, or exclude a particular meaning, for example searching for “java help” without reference to the programming language.
Here is the abstract from the thesis:
While web search engine technology has improved over time, there is often a fundamental reliance on keyword matching for searches. What happens however, when the user does not know what keywords to use or when keywords have multiple, disparate meanings?
This paper presents preliminary learning results of a prototype learning search engine that attempts to address this problem. Tibianna allows a user to manually rank a set of results based on their own relevancy function. Once a required number of results are ranked, the set is downloaded, processed and presented to support vector machines (SVMs) for learning. Once trained, Tibianna can actively reorder or discard search engine results based on the model it has learned. This provides a way of improving search results without requiring query reﬁnement. Learning outcomes from experimental trials with Tibianna are presented, demonstrating the implications of using different preprocessing techniques, and corpus sizes.
Results can be ranked via the drop-down box next to each result:
After training, Tibianna takes over, automatically excluding and re-ranking results based on a machine learning algorithm:
Tibianna was a fully-functional prototype written in C#
- “Tibianna: A Learning-Based Search Engine with Query Reﬁnement.” Heyer, C. and Diederich, J. (2002) In Proceedings of the Seventh Australasian Document Computing Symposium (ADCS′02), Sydney, Australia. University of Sydney, pp. 105-108.
- “MyNewsWave: User-centered Web search and news delivery.” Heyer, C., Madden, J., Hollingsworth, K., Heydon, P., Bartlett, K. and J. Diederich, J. (2002) In Proceedings of the Seventh Australasian Document Computing Symposium (ADCS′02), Sydney, Australia. University of Sydney, Australia. pp. 121-124