Estimating Skill Proficiency from Resumes

Anindita Sinha Banerjee, Sachin Pawar, Girish K. Palshikar, Devavrat Thosar, Jyoti Bhat, Payodhi Mandloi

Abstract: This paper attempts to solve a real-life problem of estimating proficiency levels of skills held by a person from her resume. This is a challenging problem because no other source of information than resumes is available and skill proficiency is a complex function of various aspects of a person’s experience. A resume often mentions various skills held by a person and her experience in applying these skills as part of work or project experience. We extract skills and other relevant information automatically from resumes and capture skill related information in terms of a feature vector. We propose two techniques to automatically learn a skill proficiency estimation function using these feature vectors – (i) supervised neural network based technique where we introduce a novel loss function to combine label information with domain-specific constraints and (ii) weakly supervised clustering based technique. We evaluate these techniques along with two competent baselines on a dataset of Information Technology (IT) resumes focusing on 5 major skills – Java, Python, Databases, Embedded Systems, and Machine Learning.