ADOR — Augmented Document Reader

PoC for financial document Named Entity Recognition (NER).

Project Structure

ADOR/
├── architecture/
│   └── ADOR_Architecture(GAD).docx      # Global Architecture Document
├── parser/
│   └── docx_parser.py                   # Rule-based parser for DOCX term sheets
├── ner/
│   ├── ner_extractor.py                 # Three NER model backends for chat messages
│   └── GMD_NER_Methodology.md           # Methodology: NER model selection
├── llm/
│   └── ADOR_NER_LLM_Pipeline (GMD).docx # Methodology: LLM + RAG pipeline for PDFs
├── streamlit_app.py                     # Streamlit UI
└── requirements.txt

Quick Start

pip install -r requirements.txt
python -m spacy download en_core_web_trf

Launch the UI

python -m streamlit run streamlit_app.py

Run the DOCX parser from the CLI

python parser/docx_parser.py "path/to/termsheet.docx"

Features

Tab 1 — DOCX Parser

Drag and drop any .docx term sheet. A rule-based parser scans tables and paragraphs to extract 9 financial entities.

Tab 2 — NER Chat

Type a financial chat message and choose one of three NER model backends.

Entities — NER Chat