Project Members: Ning Yu, Kiduk Yang
This project is an extension of the HARD (High Accuracy Retrieval from Documents) track we worked on during summer 2004. It aims to develop a metadata, interactive search system which can produce highly personalized results. Users for this system will be those who are willing to spend more time and effort than what's spent on common search to get better search result.
HARD04 corpus will serve as the testing data for this system.
We choose geography and subject out of total 4 metadata assigned for HARD04 to generate a metadata search.. For subject metadata, we are going to discard some elements since the original 13 elements are not very well designed. Both subject and geography metadata algorithm will be refined based on what we got for HARD track. The metadata search will leverage the traditional search by filtering and re-ranking the base-line results.
Passage retrieval technique and clarification form will also be introduced from HARD track to get relevant feedbacks from user. User are able to choose the best sentence or related text, relevant synonym sets, relevant definition for rare words and relevant phrases which will contribute to later query expansion and refinement.
©2004