hamelsmu/code_search

Implements semantic search by embedding code snippets and natural language queries into a shared vector space using sequence-to-sequence deep learning models, enabling cross-modal retrieval without explicit keyword matching. The approach trains neural networks to map both code and natural language descriptions to comparable representations, allowing queries to find semantically similar code regardless of exact syntax or terminology. Provides Docker containers (GPU/CPU variants) and Jupyter notebooks for reproducibility, though the author notes more refined techniques are available in the successor project CodeSearchNet.

490 stars. No commits in the last 6 months.