AI-Tutor

AI Tutor Bot is an interactive web application designed to classify user questions into categories such as factual, conceptual, or other. Once a question is classified, the app generates an appropriate response using GPT-4. This solution is built on Streamlit for easy UI interaction but also integrates FastAPI for backend flexibility in future expansions.

You can try out the app here.

Features

1. Question Classification:

   • Uses an SVM (Support Vector Machine) model to classify questions into predefined categories:
     • Factual: For questions requiring a direct and specific answer.
     • Conceptual: For questions requiring an explanation of a concept.
     • Other: For questions that don't fit into the other categories (e.g., open-ended, opinion-based).

2. GPT-4 Integration:

   • After classification, the app generates responses using GPT-4 based on the type of question.
   • Factual questions receive short, direct answers.
   • Conceptual questions receive detailed explanations.
   • Other questions are answered by analyzing the tone and intent behind the question.

3. Preprocessing Pipeline:

   • User input is preprocessed using tokenization and lemmatization before being fed into the classification model. This ensures the model accurately interprets the core meaning of the question.

4. Streamlit UI:

   • Provides an intuitive and easy-to-use interface where users can input questions and get real-time classified answers.
   • Currently deployed and hosted on Streamlit at gpt-tutors.streamlit.app.

5. FastAPI Integration:

   • Although the current version runs on Streamlit, there’s also a FastAPI backend built-in, allowing for future flexibility and potential integration with more complex services.

---

How the SVM Model was Trained

Data Preparation:

We prepared a dataset consisting of factual, conceptual, and other questions. Each question was labeled according to its type:

• Factual: Objective questions that require specific information.
• Conceptual: Questions that require a deeper explanation of a concept.
• Other: Questions that do not fit into the factual or conceptual categories, including opinion-based or open-ended questions.

Preprocessing:

The dataset was preprocessed using NLP techniques such as tokenization and lemmatization to clean the text before feeding it into the model. This preprocessing step ensures that only meaningful words are passed to the model, improving classification accuracy.

Steps involved:

1. Tokenization: Breaking down the text into individual words (tokens).
2. Lemmatization: Reducing words to their base form (e.g., "running" becomes "run") to ensure consistency.

Training the SVM:

We trained the Support Vector Machine (SVM) classifier using the preprocessed dataset:

• Vectorization: We used TF-IDF (Term Frequency-Inverse Document Frequency) to convert the text into numerical vectors.
• Training: The SVM model was trained on labeled data to differentiate between factual, conceptual, and other questions. The model was tuned to handle questions of varying complexities.
• Evaluation: The model was evaluated on a test set, and fine-tuned to optimize its performance across different question types.

Model Files:

• SVM Model: The trained SVM model is saved as svm_model.pkl.
• TF-IDF Vectorizer: The vectorizer used to convert text into vectors is saved as tfidf_vectorizer.pkl.

---

How the Application Works

1. User Input: Users enter a question in the Streamlit app's input field.
2. Classification: The SVM model classifies the question into one of three categories: factual, conceptual, or other.
3. GPT-4 Response: Based on the question type, a prompt is created and passed to GPT-4, which generates the appropriate response.
4. Answer Display: The app displays the classified category and the GPT-4 response on the screen.

---

Running Locally

Prerequisites

• Python 3.7 or higher
• Install required packages via requirements.txt

pip install -r requirements.txt

Running the Streamlit App To run the Streamlit app locally, use the following command:

streamlit run streamlit_app.py

This will launch the app in your default browser.

Running the FastAPI Backend If you'd like to experiment with the FastAPI backend, use the following command to start the server:

uvicorn fastapi_app:app --reload