|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Lilyana
Mihalkova |
|
Machine learning has progressed remarkably in the past decades, but most effort has focused on learning from data in which each entity is independent of the rest. In contrast, in many applications, entities of varying types are connected by a rich set of relations, such as collaborations and shared interests in a social network. A fundamental challenge in learning from such relational data is discovering the structure, or the dependencies and regularities present among the relations in the data. In the first part of the talk, I will use Markov logic, a general and expressive representation, to show how to learn structure accurately and efficiently by transferring a source model that was previously acquired in a different but related domain. I will describe in detail an algorithm that revises the source model in the case when a significant amount of data from the target domain is available. I will then briefly address transfer learning in the challenging case when target-domain data is severely limited, as well as structure learning from scratch. In the second part of the talk, I will describe our recent progress on applying Markov logic to the problem of resolving ambiguities in Web searches. In contrast to previous research on this topic, our work does not assume the availability of a long history of each user's interactions with the search engine. Instead, our system bases its predictions on a short glimpse of user search activity, captured in a session of 4-6 previous searches on average, by relating the current session to previous similarly short sessions of other users. |
|
|
|
|
|
|
|
Official
inquiries about AIIS should be directed to Alexandre Klementiev
(klementi AT uiuc DOT edu) |
||