How keralia/nlp-resume-matching-system works
TalentMatch is a resume-to-job matching product with a browser-based role discovery and ATS audit experience, plus a Python matching and academic evaluation suite. The user-facing flow initializes the browser engine with bundled resume, job, and skill data; users select a role or provide resume text and receive match diagnostics. The Python path loads the same kinds of inputs, prepares and scores candidates, and evaluates rankings. Browser component interactions are limited to the shown excerpts; unsampled integrations are not inferred.
Main entry points: run_experiments.py Experiment runner · hybrid_engine.py Hybrid ranking · data_loader.py Data loader · app.js TalentMatch UI
Components 16
Browser experience
- TalentMatch UI
js/app.js - Interactive demo
demo.py
Matching inputs
- Browser data
js/data.js - Resume corpus
dataset/resumes_corpus.json - Job descriptions
dataset/job_descriptions.json - Skills taxonomy
dataset/skills_taxonomy.json
Matching engine
- Browser matcher
js/engine.js - Data loader
src/data_loader.py - Resume preprocessing
src/preprocessor.py - Skill diagnostics
src/skill_extractor.py - TF-IDF and BM25
src/baseline_matcher.py - Semantic matching
src/semantic_matcher.py - Hybrid ranking
src/hybrid_engine.py
Python evaluation
- Experiment runner
run_experiments.py - Ranking metrics
src/evaluate.py
Other components
- Recruiter
Connections 23
- Recruiter → TalentMatch UI: selects role
js/app.js - TalentMatch UI → Browser data: reads data
js/app.js - TalentMatch UI → Browser matcher: initializes
js/app.js - Browser matcher → TalentMatch UI: returns diagnostics
js/engine.js - Data loader → Resume corpus: loads
src/data_loader.py - Data loader → Job descriptions: loads
src/data_loader.py - Data loader → Skills taxonomy: loads
src/data_loader.py - Interactive demo → Data loader: loads corpus
demo.py - Interactive demo → Hybrid ranking: ranks candidates
demo.py - Experiment runner → Data loader: loads corpus
run_experiments.py - Experiment runner → Resume preprocessing: extracts sections
run_experiments.py - Experiment runner → Skill diagnostics: computes skill scores
run_experiments.py - Experiment runner → TF-IDF and BM25: scores lexical
run_experiments.py - Experiment runner → Semantic matching: scores semantics
run_experiments.py - Experiment runner → Hybrid ranking: ranks candidates
run_experiments.py - Experiment runner → Ranking metrics: evaluates rankings
run_experiments.py - Hybrid ranking → Resume preprocessing: extracts sections
src/hybrid_engine.py - Hybrid ranking → TF-IDF and BM25: computes lexical scores
src/hybrid_engine.py - Hybrid ranking → Semantic matching: computes semantic score
src/hybrid_engine.py - Hybrid ranking → Skill diagnostics: computes skill metrics
src/hybrid_engine.py - TF-IDF and BM25 → Resume preprocessing: tokenizes text
src/baseline_matcher.py - Semantic matching → Resume preprocessing: tokenizes text
src/semantic_matcher.py - Browser matcher → Skills taxonomy: uses skill taxonomy