Together Is Better: Knowledge-aware Model with Resume Fusion for Online Job Recommendation

Widespread adoption of online recruitment platforms has led to explosive growth in employment information, resulting in an ever-increasing demand from job seekers for accurate and effective job recommendations. Existing studies on the Person-Job Fit models focus on the correlation between resumes and job descriptions, with rare consideration given to user historical behavior such as click and application. On the contrary, job recommendation methods always ignore the crucial information lurking in the resume text. In addition, the continuous influx of a vast amount of job data poses challenges to the updating of online recommendation results. To this end, we propose a novel O nline J ob R ecommendation model via R esume F usion (OJRRF) in this article, aimed at making accurate and efficient online job recommendations with the merits of addressing job cold start and long tail problems. The key contribution lies in two facets: (1) incorporating resume text information into the knowledge graph attention framework to enhance job seekers’ vector representations jointly; (2) designing a hybrid recommender strategy by combining the knowledge-aware offline model with the content-based online model. Finally, we conducted extensive comparison experiments and online A/B test on the recruitment platform of JiuYeJie big data company to validate the effectiveness and real-time capability of OJRRF. The release code can be found in https://github.com/urnotada/OJRRF .