Document Query Chatbot

A Streamlit-based chatbot that processes text files, clusters sentences by topic using K-means, and answers queries with relevant sentences using the all-MiniLM-L6-v2 sentence transformer model. Built in Google Colab with ngrok for public access, this project showcases NLP, topic clustering, and web app development.

Features

• File Upload: Upload .txt files (100–1000 words) with single or multiple topics.
• Query Processing: Enter queries (e.g., "How is AI used in healthcare?") with abbreviation support (e.g., "AI" → "artificial intelligence").
• Topic Clustering: Groups sentences into topics (2–5 clusters) using K-means on sentence embeddings.
• Top-K Results: Returns 1–5 most relevant sentences with similarity scores and explanations.
• Robustness: Handles edge cases (empty files, vague queries) with clear error messages.
• Web Interface: Streamlit UI with file uploader, query input, and sliders for customization.

Tech Stack

• Python: Core programming language.
• Streamlit: Web app framework for the UI.
• sentence-transformers: all-MiniLM-L6-v2 for sentence embeddings.
• scikit-learn: K-means clustering for topic grouping.
• NLTK: Text tokenization and query preprocessing (lemmatization).
• PyTorch: Backend for model computations.
• ngrok: Public URL tunneling for Colab.
• Google Colab: Development and testing environment.

Setup (Google Colab)

1. Open the Notebook:
   • Download document_query_chatbot.ipynb from this repository.
   • Upload to Open in Colab (set to "Anyone with the link can view").
   • Use Python 3 runtime (GPU optional for faster processing).
2. Prepare Text Files:
   • Place sample.txt, sample2.txt, mixed_sample.txt, and complex_sample.txt in /content/drive/MyDrive/.
   • Alternatively, upload files via Colab’s sidebar (folder icon > Upload).
3. Run Cells Sequentially:
   • Cell 1: Installs libraries and ngrok.
   • Cell 2: Mounts Google Drive and downloads NLTK resources.
   • Cell 3: Configures ngrok with an authtoken.
   • Cell 4: Saves the Streamlit app (app.py).
   • Cell 5: Runs the app and provides a public ngrok URL.
   • Cell 6: Fallback for testing without Streamlit/ngrok.
4. Access the App:
   • Click the ngrok URL from Cell 5 (e.g., https://abc123.ngrok.io).
   • Upload a text file, enter a query, set clusters/top-k, and view results.

Usage

1. Upload a File:
   • Use complex_sample.txt (~800 words, covers AI, renewable energy, quantum computing, biotech).
2. Enter a Query:
   • Examples: "How is AI used in healthcare?", "What’s RE’s role in climate change?", "How does quantum computing help AI?".
3. Set Parameters:
   • Number of clusters (2–5): Try 4 for multi-topic files.
   • Number of sentences (1–5): Try 3 for detailed results.
4. View Results:
   • Top Matching Sentences: Relevant sentences with topics, scores, and explanations.
   • Expander: All sentences with their clusters and scores.
5. Test Edge Cases:
   • Empty query: Should show "Error: Please enter a query!".
   • Single-sentence file: Should show "Error: File must contain at least two sentences.".

Sample Files

• sample.txt: AI-focused (~100 words).
• sample2.txt: Renewable energy (~100 words).
• mixed_sample.txt: AI and renewable energy (~200 words).
• complex_sample.txt: AI, renewable energy, quantum computing, biotechnology (~800 words).

Demo

• Colab Demo: Run document_query_chatbot.ipynb in Colab to test with ngrok.
• Live Demo (Planned): Deployed on Streamlit Community Cloud (link TBD).

Installation (Local Setup, Optional)

1. Clone the repository:

   git clone https://github.com/your-username/document-query-chatbot.git
   cd document-query-chatbot