NLP Chatbot Suite — NullClass Data Science Internship

Apr 2025 – Jun 2025 · NullClass Edtech Pvt. Ltd.

---

Overview

Two NLP systems built end-to-end during a 2-month Data Science internship focused on real-world chatbot architecture, dynamic knowledge management, and interface development.

---

Project 1 — Extractive Summarization Chatbot

What it does

Takes long-form text input and extracts the most important sentences to produce a concise summary. Built with traditional NLP + a TensorFlow classification model.

How it works

1. Text preprocessing — tokenization, stopword removal, lemmatization (NLTK)
2. Feature extraction — Bag of Words representation
3. Sentence scoring — TensorFlow model ranks sentences by importance
4. GUI — Tkinter desktop interface for input/output

Results

87% summarization accuracy on held-out test data

Stack

Python TensorFlow NLTK Bag of Words Tkinter

---

Project 2 — Self-Updating Knowledge Chatbot

What it does

A Q&A chatbot that can ingest new documents and update its knowledge base dynamically — no retraining required. Uses vector search for retrieval and basic entity recognition to keep responses grounded.

How it works

1. Document ingestion — XML/CSV files parsed and chunked
2. Vector indexing — FAISS stores and retrieves embeddings
3. Dynamic updates — new documents added to the index at runtime via LangChain
4. Entity recognition — basic NER to improve response grounding
5. UI — Streamlit interface for chat interaction

Results

90% correct response retrieval on unseen queries

Stack

Python FAISS LangChain NLTK Streamlit

---

Repo Structure

Tasks/
├── task1/
│   ├── SummaryTask1.ipynb     # Extractive summarization model (TF + BoW)
│   └── requirements.txt
├── taskno3/
│   ├── app.py                 # Streamlit UI
│   ├── gui.py                 # Tkinter interface
│   ├── intentsMed.json        # Knowledge base / intents data
│   ├── requirement.txt
│   └── updateknowledge.py     # Dynamic FAISS knowledge base updates
└── README.md

Update folder names above to match your actual structure

---

Run locally

git clone https://github.com/Debangana13/Tasks
cd Tasks

# Project 1 — Summarization
pip install -r task1/requirements.txt
jupyter notebook task1/SummaryTask1.ipynb

# Project 2 — Self-updating chatbot
pip install -r taskno3/requirement.txt
streamlit run taskno3/app.py

---

Part of NullClass Edtech Pvt. Ltd. Data Science Internship · Certificate issued Jun 2025
Built by Debangana Ghosh