Generating unified candidate skill graph for career path recommendation

Abstract

'How should I progress in my career?' is an important question that every working professional seeks answer multiple times during her career. Given the amount of career position data of individuals available online, personalized career path recommendation systems that could mine and recommend the most relevant career paths for a user are on the rise. However, such recommendation systems typically are only effective within a single organization where there are standardized job roles. At an industry sector level such as Information Technology or across such different industry sectors (such as retail, insurance, health care), mining and recommending the most relevant career paths for a user is still an unsolved research challenge. Towards addressing this problem, we propose a system that leverages the notion of skills to construct skill graphs that can form the basis for career path recommendations. We perceive skills are more amenable for career path standardizations across the organizations. Our proposed system ingests a users profile (in a pdf, word format or other public and shared data sources) and leverages an Open IE pipeline to extract education and experiences. Subsequently, the extracted entities are mapped as specific skills that are expressed in the form of a novel unified skill graph. We believe that such skill graphs which capture both spatial and temporal relationships aid in generating precise career path recommendations. An evaluation of our current skill extraction model with an industrial scale dataset yielded a precision and recall of 80.54% and 86.44% respectively.