Mining coordinated intent representation for entity search and recommendation

We study the problem of learning query intent representation for an entity search task such as product retrieval, where a user would use a keyword query to retrieve entities based on their structured attribute value descriptions. Existing intent representation has been mostly based on the query space. These methods overlook the critical information from the entity space and the connection in between. Consequently, when such representation methods are used in intent mining from user engagement logs in entity search, they cannot fully discover the comprehensive knowledge of user preference, which is essential for improving the effectiveness of entity search and recommendation, as well as many applications such as business intelligence. To address this problem, we propose a novel Coordinated Intent Representation, where each user intent is represented collectively in both the query space and the entity space. Specifically, a coordinated intent representation consists of a language model to capture typical query terms used for search and a series of probabilistic distributions on entity attributes and attribute values to characterize the preferred features of entities for the corresponding intent. We propose a novel generative model to discover coordinated intent representations from the entity search logs. Evaluation in the domain of product search shows that the proposed model is effective for discovering meaningful coordinated shopping intents, and the discovered intent representation can be directly used for improving the accuracy of product search and recommendation.