BERT-Based Semantic Resume Analysis and Markov Chain-Driven Career Roadmap Recommendation System

An ATS can narrow down the resume to find a match of words, and this can exclude those who describe the same job using different terminology. Also they do not provide a fit score, but rather display a person’s potential to fit a desired role. Career Chain AI is a web based product that demonstrates the combination of semantic resume job matching with a likelihood modelling based career path suggestion, shown in this paper. This tool extracts text from the PDF, DOCX and TXT resumes, splits parts of the resume, normalizes the skill names against a set of skills, and converts the resume and job text into the vectors using the allMiniLML6v2 Sentence BERT model. A weighted score combines semantic similarity, necessary and desired skill coverage, experience, education, project relevance and certificate relevance. A simple 1st order Markov model is then used to estimate the probabilities of role transitioning from the counts of transitions. Resume specific skill. Semantic fit reranked the possible transitions giving a loop free roadmap of up to three steps. Looking at the built prototype and its seed data shows support for 25 roles, 32 directed transitions and 75 skill records. See real case studies of transition to see how Career Chain AI can relate recommendations to probabilities and skill gaps. The final design would rather have guidance, than a hidden ranking, but would still want to keep a light Django framework which is usable for multiple academic tests.