Interview Bot with Automatic Question Generation and Answer Evaluation

Abstract

Resume screening and interview chatbots are the basic requirements for today’s hiring processes for a given job post. There are a large number of candidates who apply and manually sorting these resumes as per the job requirement is a tedious task for shortlisting the candidates. The motivation for the project lies in solving the challenge of automating the process of recruitment from resume screening to the first technical test of the candidate. The challenge for screening candidates’ resumes would be to get structured data from differently structured resumes using Natural Language Processing and get all the keywords consisting of programming languages. The other challenge lies in the generation of questions and evaluation of answers given by the candidates using Natural Language Processing techniques. The challenge of proctoring the candidate during the test is the detection of eye movements of candidates during the interview and lip reading to check for any malpractice. The idea here is that the candidate’s resume would be scanned for all the keywords consisting of programming languages to the concept words extracted from the resume. Then based on that, the candidate would be assigned a job role that suits his/her skill set which reduces the time required to find the relevant candidate for a given Job Description. The candidate’s interview would be conducted to know the candidate’s technical skills to some extent. Multiple Choice Questions based tests could be used for the initial screening. Still, they don’t assess the in-depth knowledge of the candidate as much as the short answer questions can. Hence generation of short answer questions and evaluation is important. The question generated for the candidate would be either based on sentence-level information or paragraph-level information; the sentences and paragraphs would be of different technical questions on the programming languages and concepts. The questions would be selected both randomly and based on candidates’ responses. During the interview, the camera would be used to proctor the exam so that there should be no more than one person in front of the camera and the candidate should not use any other device during the exam. The final score generated for the candidate would be based on the match with the job description and candidate interview.