Resume Screening–Testing For Data Stability

Resume parsing and evaluation are critical tasks for human resources teams, and with the surplus of candidates applying to jobs in today’s time, the process of resume evaluation is both manual and time consuming. The process of resume parsing has been traditionally done manually, which is time-consuming and prone to errors. However, with the advancements in Natural Language Processing (NLP), automated resume parsing has become a popular research area. In addition to parsing, the advent of machine learning has allowed resume screening to be automated, through the use of key words and named entities. In this technical paper, we provide a comprehensive analysis of pre-existing research on resume parsing and evaluation frameworks, and provide a detailed overview of the different NLP techniques used for resume parsing, such as Named Entity Recognition, along with the various machine learning algorithms used for screening resumes, such as cosine similarity. We then implement our own models using resume datasets provided by a third-party firm, and analyze the stability and consistency of output data for each model using frequency distribution graphs, statistical averages, etc. in order to determine the best model for our given dataset. This paper provides a number of valuable graphs and observations that can be helpful for researchers and practitioners in the field of recruitment, machine learning.