Update README.md

README.md

@@ -1,90 +1,20 @@
-# SmartDoc NER Pro 🔍
-
-**Advanced Named Entity Recognition system** that extracts Persons, Organizations, Locations, and Miscellaneous entities from text and documents with high accuracy and a rich feature set.
-
-Built by **Nafees Ahmad** | PAF-IAST, Pakistan
-
----
-
-## What's New vs Basic NER
-
-| Feature | Basic Version | SmartDoc NER Pro |
-|---|---|---|
-| Model | spaCy en_core_web_sm (F1 ~85%) | BERT-Large-NER (F1 **92.8%**) |
-| Input | Text only | Text + PDF + DOCX + TXT |
-| Processing | Single input | Single + **Batch processing** |
-| Output | Labels only | Labels + confidence scores + inline highlights |
-| Analytics | None | Frequency, co-occurrence, confidence stats |
-| Search | None | Keyword search + type/confidence filter |
-| Export | None | CSV + JSON + **Annotated DOCX** |
-
 ---
-
-## Features
-
-1. **Text NER** — Paste any text and extract entities with inline colour highlighting
-2. **Document NER** — Upload PDF, DOCX, or TXT files for extraction
-3. **Batch Processing** — Process multiple documents simultaneously
-4. **Entity Analytics** — Frequency distribution, Person–Organization co-occurrence, confidence statistics
-5. **Search & Filter** — Search entities by keyword, type, or confidence threshold
-6. **Export** — Download results as CSV, JSON, or annotated Word document
-
+title: SmartDoc NER Pro
+emoji: 🔍
+colorFrom: blue
+colorTo: indigo
+sdk: gradio
+sdk_version: "4.0.0"
+python_version: "3.10"
+app_file: app.py
+pinned: false
 ---
 
-## Model Accuracy
-
-| Model | F1 Score (CoNLL-2003) |
-|---|---|
-| spaCy en_core_web_sm | ~85% |
-| BERT-base-NER | ~91.3% |
-| **BERT-large-NER (this project)** | **~92.8%** |
-
-Model source: [dslim/bert-large-NER](https://huggingface.co/dslim/bert-large-NER)
-
-The model uses `aggregation_strategy="max"` which merges sub-word tokens intelligently, significantly improving entity boundary detection compared to raw token-level output.
-
----
-
-## Project Structure
-
-```
-smartdoc_ner/
-├── app.py                    # Main Gradio application
-├── requirements.txt
-├── README.md
-├── modules/
-│   ├── ner_engine.py         # BERT model + highlighting
-│   ├── doc_reader.py         # PDF/DOCX/TXT extraction + chunking
-│   ├── analytics.py          # Frequency, co-occurrence, dedup
-│   ├── exporter.py           # CSV/JSON/annotated DOCX export
-│   ├── entity_search.py      # Search & filter
-│   └── batch_processor.py    # Multi-document processing
-└── exports/                  # Exported files saved here
-```
-
----
-
-## Setup & Run
-
-```bash
-# Clone the repo
-git clone https://github.com/Nafees369/Named-Entity-recognition
-cd smartdoc_ner
-
-# Install dependencies
-pip install -r requirements.txt
-
-# Run the app
-python app.py
-```
-
-App opens at: `http://127.0.0.1:7860`
+# SmartDoc NER Pro 🔍
 
----
+**Advanced Named Entity Recognition** — Persons · Organizations · Locations · Miscellaneous
 
-## Technical Highlights
+**Model:** `dslim/bert-large-NER` (F1 ≈ 92.8% on CoNLL-2003)
+**Supports:** Text · PDF · DOCX · TXT · Batch Processing · Export (CSV / JSON / DOCX)
 
-- **Chunked processing** — Long documents are split into sentence-aware 400-character chunks before passing to BERT (which has a 512-token limit), preventing silent truncation errors
-- **Offset tracking** — Character offsets are tracked across chunks so inline highlighting works correctly on the full document
-- **Confidence filtering** — All outputs include token-level confidence scores; user-adjustable minimum threshold removes uncertain predictions
-- **Annotated DOCX export** — Highlighted entity output in a downloadable Word document, colour-coded by entity type
+*Built by Nafees Ahmad*