In session personalisation for recommender systems

Previous eorts in recommendation of candidates for talent search followed the general pattern of receiving an initial search criteria and generating a set of candidates utilizing a pre-trained model. Traditionally, the generated recom- mendations are nal, that is, the list of potential candidates is not modied unless the user specically changes his/her search input. In this paper, we are proposing a candidate recommendation model which takes into account the im- mediate feedback of the user, and updates the candidate recommendations at each step. This setting also allows for very uninformative initial search queries, since we pinpoint the user's intent due to the feedback during the search ses- sion. To achieve our goal, we employ an intent clustering method based on topic modeling which separates the can- didate space into meaningful, possibly overlapping, subsets (which we call intent clusters) for each position. On top of the candidate segments, we apply a multi-armed bandit approach to choose which intent cluster is more appropriate for the current session. We also present an online learning scheme which updates the intent clusters within the session due to user feedback, therefore provides even more relevant results to the users. Our initial experiments demonstrate the potential benets of our proposed methodology.

Vijay Dialani is a Software Engineer(TLM) at Google, working on natural language understanding and conversational AI. Prior to that he was at LinkedIn. At LinkedIn, he was the technical lead for development of a number of products such as Next Generation Recruiter and Intelligent Matches within the LinkedIn Talent Solutions. He currently leads efforts in online learning and preference elicitation. Previously, he was an academic and was director for the Infolab at Boise State University, with active collaborations with Micron and Simplot. He had also worked at eBay, GE Research, Microsoft Live Labs and IBM Research. He has been a regular reviewer for journals such as Concurrency and Computation. His research has been published at WWW, CIKM and ICDE. He has published 20+ papers, with 650+ citations and filed 15+ patents. While in academia he created graduate level courses in Data Science, Machine Learning and Cloud Computing. He has advised several graduate students, been on graduate committee of several masters and PhD students.