Material Detail

This video was recorded at MUSCLE Conference joint with VITALAS Conference. Information retrieval is a very wide domain which can involve various types of activities and tasks. Many complex factors are participating in a search for information and many systems have been experimented. Nowadays a general consensus has been established around a keyword/document matching process which appears to be efficient on large scale and have enough reliability to satisfy a significant part of the users. Btu this claim has to be limited and for some subjects, search is still a difficult task. Many reasons can be proposed to explain these phenomena, but the most salient ones are the difficulty for users to express their needs while searching for information and the limitation of shared knowledge between users and information retrieval systems, meaning that both users and machines don't really understand the information and knowledge space used as references by the other. This presentation try to provide an overview of one way to resolve those gaps: using feedback learning. The aim is to make the system learning on user behaviour in order to better define its current needs. Machine learning algorithms applied on signal coming from user while performing a search can lead to the understanding of what is really relevant to the users and then can be exploited to help him during its tasks. The work, engaged through the VITALAS1 project, is presented: study of users search logs and definition of a feedback learning framework. Then research on implicit relevance feedback and query optimisation is presented as a first attempt to exploit the feedback learning framework. Finally an overview of the next steps within those studies is presented and especially their impact on the VITALAS project.

Disciplines:

• Science and Technology  / Computer Science