Feature Proposal] Advanced SkillsScorer with Jaccard & Semantic Similarity #535
Description
Is your feature request related to a problem? Please describe.
The current skill-matching logic (if based only on keywords) can be limited. It might miss strong candidates who use different but semantically related terms on their resume. For example, a job asking for "Python Web Development" might not match a resume that lists "Flask" and "Django," even though they are directly related.
Describe the solution you'd like
I want to contribute an advanced SkillsScorer module that uses a hybrid approach for more accurate scoring:
Jaccard Similarity: To find and score exact keyword matches.
SentenceTransformer Semantic Similarity: To use an AI model (all-MiniLM-L6-v2) to find and score conceptual matches (like "Flask" vs. "Python Web Development").
This will provide a much more intelligent and accurate skill-fit score.
I've read the CONTRIBUTING.md and I'm ready to start. My main question is: Should I be adding this feature to the older Streamlit app or the new Next.js/FastAPI webapp? I'm happy to work on whichever you prefer.
Describe alternatives you've considered
The alternative is to keep using only exact keyword matching, which is less accurate and can miss good matches.
Additional context
I have the Python classes for the SkillsScorer and SemanticMatcher already built and tested. I'm ready to integrate them once I know which part of the project to add them to.