Exploring the Efficiency of Text-Similarity Measures in Automated Resume Screening for Recruitment

With many online recruitment portals requesting job applicants to upload their resumes, the automated process of screening and shortlisting candidates can accelerate selection and decision-making. This study explores the use of text similarity measures as an alternative to experienced human hiring managers in processing resumes. Three text similarity measures: Cosine, Sqrt-Cosine, and Improved Sqrt-Cosine (ISC) similarity were utilized as computer programs in scanning resumes for the recruitment of a business development manager and a software engineer. The decisions of the algorithms were compared to those of an expert hiring manager within the same scenarios. The findings indicate that ISC and Sqrt-Cosine were closer to the expert-human decision than Cosine similarity. These specific text-similarity algorithms can also make acceptable decisions even when recruiting for high-level positions and can do so in seconds when executed as a program on a normal CPU processor. This study suggests that these algorithms can efficiently facilitate the process of decision-making in recruitment and shortlisting candidates and can be an effective alternative to expert hiring managers.