Upload 3 files

README.md

@@ -1,3 +1,88 @@
----
-license: mit
----
+
+# 🧠 Named Entity Recognition (NER) with BERT
+
+This is a web-based Named Entity Recognition (NER) tool powered by a fine-tuned BERT model, built using PyTorch, Hugging Face Transformers, and Streamlit.
+
+## 🚀 Demo
+
+Try the live app (if deployed):  
+👉 [Your Streamlit Cloud or Hugging Face Space link]
+
+---
+
+## 📌 Features
+
+- 🔎 Extracts and highlights entities in raw text (PER, LOC, ORG, etc.)
+- 💬 Clean and interactive Streamlit UI
+- 🧠 Fine-tuned on the CoNLL-2003 dataset
+- ☁️ Deployable via Streamlit Cloud or Hugging Face Spaces
+
+---
+
+## 🖥️ Run Locally
+
+1. Clone this repo:
+   ```bash
+   git clone https://github.com/prasanna-badiger-7/ner-bert-app.git
+   cd ner-bert-app
+   ```
+
+2. Install requirements:
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the app:
+   ```bash
+   streamlit run app.py
+   ```
+
+---
+
+## 📁 Project Structure
+
+```
+ner-bert-app/
+│
+├── app.py               # Streamlit frontend
+├── ner_model/           # Saved model & tokenizer
+├── requirements.txt     # Dependencies
+└── README.md
+```
+
+---
+
+## 🧠 Model
+
+Fine-tuned `bert-base-cased` on the CoNLL-2003 dataset using Hugging Face Transformers and Datasets.
+
+---
+
+## 📦 Dependencies
+
+- `streamlit`
+- `transformers`
+- `torch`
+
+Install all with:
+```bash
+pip install -r requirements.txt
+```
+
+---
+
+## 📸 Example
+
+![Demo Screenshot](demo.png)
+
+---
+
+## 🧑‍💻 Author
+
+Made with ❤️ by [Prasanna](https://github.com/your-username)
+
+---
+
+## 📜 License
+
+MIT License

app.py

@@ -0,0 +1,40 @@
+from transformers import AutoTokenizer, AutoModelForTokenClassification
+import torch
+import streamlit as st
+
+model = AutoModelForTokenClassification.from_pretrained('./ner_model')
+tokenizer = AutoTokenizer.from_pretrained('./ner_model')
+label_list = model.config.id2label
+
+
+def predict_entities(text):
+    tokens = tokenizer(text, return_tensors='pt', truncation=True, is_split_into_words=False)
+    with torch.no_grad():
+        outputs = model(**tokens)
+
+    predictions = torch.argmax(outputs.logits, dim=2)
+    tokens = tokens['input_ids'][0]
+    prediction_ids = predictions[0]
+
+    result = []
+    for token_id, pred_id in zip(tokens, prediction_ids):
+        token = tokenizer.decode([token_id])
+        label = label_list[pred_id.item()]
+        if token not in ['[CLS]', '[SEP]', '[PAD]']:
+            result.append((token, label))
+    return result
+
+st.set_page_config(page_title="NER App", layout="wide")
+
+st.title("🧠 Named Entity Recognition (NER) with BERT")
+text_input = st.text_area("Enter text to analyze:")
+
+if st.button("Analyze"):
+    if text_input:
+        results = predict_entities(text_input)
+        st.markdown("### 🔍 Extracted Entities:")
+        for word, label in results:
+            if label != "O":
+                st.markdown(f"**{word}** → `{label}`")
+    else:
+        st.warning("Please enter some text.")

requirements.txt

@@ -0,0 +1,3 @@
+torch
+transformers
+streamlit