Resummizer: Customizable Resume Summarization using Natural Language Processing

Abstract

"With the increased demand for job opportunities, companies receive a large number of resumes whenever there is a job vacancy. Recruiters face many difficulties when choosing the best candidates for a particular job by going through each resume. Therefore, many companies have automated the recruitment process up to certain levels. The existing Resume Summarization tools tend to have their own limitations as summarizing a resume is a complex process with many challenges.

Amongst various limitations, the existing resume summarization systems do not show effort on summarizing resumes based on areas of interest entered by the user which are not predefined. As a solution to this problem, a multi-label text classification model has been used in this research to identify the categories of each sentence in the resume. The multi-label text classification algorithm used in this research is an extended Support Vector Machine (SVM) algorithm which uses the Binary Relevance technique by passing Support Vector Classifier (SVC) which supports multi-label classification. A new dataset specific for this approach was created by the author. Then using a novel algorithm, the author compares the identified categories with the areas of interest in order to provide a customized summary of the resume based on the areas of interest entered by the user.

The overall accuracy, precision, recall and f1-score of the multi-label text classification model are 85.54%, 97.72%, 96.31% and 96.55% respectively. The accuracy can be improved further by adding more records to the dataset for classes which have less instances."