Job information retrieval based on document similarity

Job information retrieval (IR) exhibits unique characteristics compared to common IR task. First, searching precision on job posting full text is low because job descriptions cannot be properly used in common IR methods. Second, job names semantically similar to the one mentioned in the searching query cannot be detected by common IR methods. In this paper, job descriptions are handled under a two-step job IR framework to find job postings semantically similar to seeds job posting retrieved by the common IR methods. Preliminary experiments prove that this method is effective.