Information re-retrieval: repeat queries in Yahoo's logs
SIGIR '07 Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval
The re:search engine: simultaneous support for finding and re-finding
Proceedings of the 20th annual ACM symposium on User interface software and technology
Web History Tools and Revisitation Support: A Survey of Existing Approaches and Directions
Foundations and Trends in Human-Computer Interaction
Keeping Found Things Found: The Study and Practice of Personal Information Management: The Study and Practice of Personal Information Management
| Hi-index | 0.00 |
Although one of the most common uses for the Internet to search for information, Web search tools often fail to connect people with what they are looking for. This is because search tools are designed to satisfy people in general, not the searcher in particular. Different individuals with different information needs often type the same search terms into a search box and expect different results. For example, the query "breast cancer" may be used by a student to find information on the disease for a fifth grade science report, and by a cancer patient to find treatment options. This thesis explores how Web search personalization can help individuals take advantage of their unique past information interactions when searching. Several studies of search behavior are presented and used to inform the design of a personalized search system that significantly improves result quality. Without requiring any extra effort from the user, the system is able to return simple breast cancer tutorials for the fifth grader's "breast cancer" query, and lists of treatment options for the patient's. While personalization can help identify relevant new information, new information can create problems re-finding when presented in a way that does not account for previous information interactions. Consider the cancer patient who repeats a search for breast cancer treatments: she may want to learn about new treatments while reviewing the information she found earlier about her current treatment. To not interfere with refinding, repeat search results should be personalized not by ranking the most relevant results first, but rather by ranking them where the user most expects them to be. This thesis presents a model of what people remember about search results, and shows that it is possible to invisibly merge new information into previously viewed search result lists where information has been forgotten. Personalizing repeat search results in this way enables people to effectively find both new and old information using the same search result list. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)