🤖💬 ProServe AI – Customer Support Chatbot

An intelligent ML-powered customer support chatbot using NLP & semantic similarity

🚀 Project Overview

Modern businesses rely on AI-driven customer support systems to deliver fast and accurate responses at scale. ProServe AI replicates this real-world customer service automation using Machine Learning and Natural Language Processing.

ProServe AI is a retrieval-based customer support chatbot that understands user queries, converts them into numerical vectors, and retrieves the most relevant response from a structured knowledge base using TF-IDF vectorization and cosine similarity.

The system performs:

• Query preprocessing & cleaning
• TF-IDF feature extraction
• Semantic similarity computation
• Confidence-based response selection
• Human-agent fallback handling

📂 Dataset Used

This project uses a Customer Support Q&A Dataset, containing:

• customer_support_QA_dataset.csv - https://drive.google.com/file/d/1EEUd736xVvEhMtH7yseNn9Bt7UxvOl7V/view?usp=drive_link

Each record includes:

• ❓ Customer query
• 💬 Predefined support response

The dataset simulates real-world customer service conversations across multiple support scenarios.

🏗️ Project Workflow

1️⃣ Importing Libraries Used for data handling, NLP processing, ML modeling, and UI development.

• import pandas as pd
• import re
• from sklearn.feature_extraction.text import TfidfVectorizer
• from sklearn.metrics.pairwise import cosine_similarity
• import streamlit as st

2️⃣ Loading the Dataset

• df = pd.read_csv("customer_support_QA_dataset.csv")

3️⃣ Text Preprocessing

• Lowercasing text
• Removing special characters
• Normalizing extra spaces

def preprocess(text):
    text = text.lower()
    text = re.sub(r"[^a-z0-9\s]", " ", text)
    return re.sub(r"\s+", " ", text).strip()

4️⃣ TF-IDF Vectorization

• Converts cleaned queries into numerical vectors.

vectorizer = TfidfVectorizer(max_features=6000, ngram_range=(1,2))
X = vectorizer.fit_transform(df['clean_query'])

5️⃣ Cosine Similarity Computation

• Measures semantic similarity between user query and knowledge base.

similarity = cosine_similarity(user_vector, X)

6️⃣ Response Retrieval Logic

• Highest similarity score determines the response.
• Low-confidence queries trigger fallback to human support.

🧠 Chatbot Intelligence

The system includes:

• Confidence threshold handling
• Contextual chat memory
• Session-based conversation history
• Typing animation for realistic UX

🎨 User Interface

• ChatGPT-style dark theme
• Real-time chat interaction
• Sidebar memory panel
• Smooth typing animation

Built entirely using Streamlit + custom HTML/CSS.

📁 Project Structure

ProServe-AI/
│
├── customer_support_QA_dataset.csv   # Support knowledge base
├── ProServeAI.py                     # Streamlit chatbot app
├── ProServeAI-v2                     # Streamlit v2 chatbot app 
├── ProServe-AI-Main.ipynb            # Model development notebook
├── requirements.txt                 # Dependencies
└── README.md                         # Documentation

▶️ How to Run

pip install -r requirements.txt
streamlit run ProServeAI.py

📌 Use Cases

• Customer Support Automation
• FAQ & Helpdesk Systems
• SaaS Support Bots
• AI Knowledge Base Assistants
• Business Service Chatbots

📈 Future Enhancements

• Sentence-BERT embeddings
• FAISS vector database
• Multilingual support
• Intent classification
• Cloud deployment (AWS / GCP)

🏁 Conclusion

• ProServe AI demonstrates how Machine Learning and NLP can automate customer support effectively.
• By leveraging TF-IDF and cosine similarity, the system delivers fast, accurate, and scalable responses.
• The project serves as a strong foundation for production-grade AI-powered support systems.

👨‍💻 Developed By

Ayush

---

• 💼LinkedIn: https://linkedin.com/in/ayush130
• 💻GitHub: https://github.com/ayush13-0
• ✉️Email- bhanuseenu914@gmail.com

📜 License

• This project is licensed under the MIT License.