Real-Time Resume Classification System Using LinkedIn Profile Descriptions

Abstract

In the domain of online job recruitment, accurate job and resume classification is vital for both the seeker and the recruiter. We have built an automatic text classification system that utilizes various techniques like Term frequency-inverse document frequency with Machine Learning and Convolution Neural network for training the model with texts and classifying them into labels and finally to compare their results. Using resume data of applicants, we have categorized them into different categories. Due to the sensitive nature of resume data, we have used domain adaptation. A classifier is trained on a large dataset of job description snippet, which is then used to classify resume data. Despite having a small dataset, consistent classification performance is seen. The primary filter for this type of work is the efficiency the system can provide. We aim to compare the results obtained by various algorithms that are generated using the same data so that the efficiency of each algorithm can be evaluated. From the result, it is evident that character-level CNN gives a better F1 score compared to other models.