Learning to Match

- A Novel and Unified Framework for Machine Learning

Tasks in many applications can be formalized as matching between heterogeneous objects, including search, recommendation, image annotation, paraphrasing, and drug design. For example, search can be viewed as a problem of matching a query with a document. So far, separate machine learning techniques have been developed to address individual problems in different domains. Therefore, it is necessary and important to study the problems as well as the techniques in a unified framework to understand them better and to enhance the state-of-the-art for each of the tasks.

This is exactly the motivation behind the work of researchers at Noah’s Ark Lab who proposed to study the general problem of learning to match, which straddles the theory of machine learning and its applications. A number of methods have been developed under the theme of learning to match. For example, inspired by a model for matching in search, a model for matching in recommendation has been invented in the lab. It employs the gradient boosting tree techniques and it significantly improves upon existing methods in several benchmark datasets of recommendation. A workshop on matching in search has been organized at SIGIR 2014 and a survey paper on the topic has been published in the Foundations and Trends of Information Retrieval.

See our technical blog about Learning to Match.