Job Title Prediction as a Dual Task of Expertise Prediction in Open Source Software

Xin Liu, Yu Wang, Qiwen Dong, Xuesong Lu

Abstract: Career path prediction is an important task in computational jobs marketplace. Recent advances in data science and artificial intelligence have imposed a huge recruitment demand on talents in the IT field. Previous studies predict a talent’s next job title solely based on her past experience in the resume, which can lead to errors if the resume contains fake information. With the popularity of open-source software, we argue that the next job title can be predicted based on a candidate’s past expertise in the open-source community. On the other hand, the career path can also affect the development of a talent’s expertise. Motivated by the observation, we propose to predict the job titles of IT talents as a dual task of forecasting their expertise development in open-source software. To solve the task, we design a dual learning model DualJE that leverages both the data-level and model-level duality. Experimental results show that DualJE is effective and performs much better than comparative models. A replication package for this work is available at https://github.com/DaSESmartEdu/DualJE.