Long Tail Query Enrichment for Semantic Job Search

Online job boards are used by millions of job seekers, who browse through the postings for jobs that match their interest. Queries are crafted using terminology generated by the users, which may not match the language used in the job postings. Semantic enrichment methods attempt to fill such a lexical gap by re-writing the queries based on richer terms, which are mined using behavioral logs. However, one challenge of such approaches is dealing with noisy and sparse logs of queries that appear at the tail of query frequency distribution. Although, tail queries are individually uncommon, they make up a large portion of queries collectively. Ignoring such heavy long tails in the job search domain may result in turning away job seekers who are not able to find relevant jobs. Hence, in this paper, we propose a novel method which exploits the user click-skip logs and the content of job postings to generate semantic enrichments for tail queries. We evaluate our techniques using data from CareerBuilder, which is one of the largest job boards in the world. Our experiments demonstrate effective enrichment in terms of search results size and relevancy.