alirezaaminzadeh/enterprise-rag-studio

OrgMind RAG — Hybrid TF-IDF Retrieval Index

Aria AI Engineering Team · Space · Dataset

Model Description

Char_wb n-gram TF-IDF embedding index for bilingual (EN/FA) enterprise policy retrieval. Powers OrgMind RAG Studio hybrid search (cosine similarity + keyword reranking).

Intended Use

• Internal knowledge base Q&A with citations
• HR / HSE / Procurement / IT policy lookup
• Not a replacement for legal, HR, or safety professional advice

Training

python scripts/build_dataset.py
python scripts/train_index.py

Metrics

See dataset/eval_results.json and dataset/benchmark_report.json after python scripts/benchmark.py.

Limitations

• Extractive answers only in default pipeline
• Coverage limited to indexed documents
• TF-IDF may underperform dense embeddings on paraphrased queries