How abhi9-git/amazonml_challenge_2026 works
This repository is a partial business-entity-resolution challenge solution. Its visible workflow normalizes train/test entity records, then generates candidate matches using token, character n-gram MinHash/LSH, and embedding-based search. A separate validator checks submission TSVs against test entities and warns when final matches fall outside the candidate set. The tree and excerpts do not show a final matching/scoring implementation, so that stage is represented only as an external artifact; the README provides no further workflow detail.
Main entry points: validate_submission.py Submission checks · blocking.py Candidate generation · normalize.py Record normalization
Components 12
Entity Resolution
- Record normalization
code/business_entity_resolution/src/normalize.py - Candidate generation
code/business_entity_resolution/src/blocking.py
Submission Validation
- Submission checks
utils/validate_submission.py
Data Artifacts
- Train/test sources
- Normalized records
- Candidate pairs
- Matching results
- Test source files
Other components
- Challenge participant
- Sentence-transformer model
- Final match producer
- Validated submission
Connections 14
- Challenge participant → Record normalization: runs
- Train/test sources → Record normalization: reads
- Record normalization → Normalized records: writes
- Challenge participant → Candidate generation: runs
- Normalized records → Candidate generation: reads
- Candidate generation → Sentence-transformer model: encodes names
- Candidate generation → Candidate pairs: writes
- Candidate pairs → Final match producer: candidate input
- Final match producer → Matching results: produces
- Challenge participant → Submission checks: runs
- Submission checks → Test source files: reads
- Submission checks → Matching results: validates
- Submission checks → Candidate pairs: cross-checks
- Submission checks → Validated submission: reports status