Candidate Ranking optimization leveraging Candidate Resume and Job Description using Neural Networks
The Challenge
In 2021-22, an HRTech client sought a recruitment assistance solution designed to parse resumes, rank candidates according to job descriptions, and extract key insights from interview transcripts. Tatras delivered a comprehensive backend solution addressing all these needs, utilizing classical Computer Vision, Image Processing, and Natural Language Processing (NLP) techniques. Notably, this solution was developed prior to the advent of Large Language Models (LLMs).
Hypothesis
- The solution necessitates processing various text types to fulfill specific requirements, involving text extraction, processing, and analysis to derive meaningful insights. These insights will inform and enhance subsequent downstream tasks.
- The solution is designed to significantly reduce the HR workload by automating the processing of candidate information from multiple sources. Additionally, it enhances model transparency by providing explainable outputs. The system also allows users to generate new results based on varying selection parameters, offering greater flexibility in decision-making.
Execution
- Annotated resume data and trained vision models for semantic segmentation of resume.
- Annotated Job related text data from JD and resume for NER.
- Developed supervised and unsupervised ranking algorithm as per user preference.
- Developed unsupervised topic modelling and subject classifier for interview transcript data.
- Developed a hierarchical and clustered knowledge base for Job titles, skills and functions.
Outcomes
- Above 94% accurate resume parser.
- Efficient and explainable ranking for candidates given JD.
- Topic based filtering of interview conversation and relevancy analysis.
- Enriched and domain specific knowledge base to assist ranking and understand relevancy of topic in interview.
Project Highlights
70%
increase in matching efficiency
for candidate roles
for candidate roles