Chatbot Documentation Task
Task 1: Identify Existing Online Chatbot Platforms That Support PDF
Upload and Q&A
1. ChatGPT (with File Upload Support)
Platform: chat.openai.com
Features:
Users can upload PDF documents directly into the chat (available with ChatGPT Plus and
Pro accounts using GPT-4).
GPT-4 scans and reads the file content.
Users can ask questions in natural language, and ChatGPT responds based on the
document content.
Handles long-form documents and multiple files.
Pros:
Highly accurate answers.
Good contextual understanding of documents.
Ability to reference specific sections, tables, or figures.
Cons:
Not free (requires paid subscription).
Sometimes has limitations with scanned or image-based PDFs.
2. ChatPDF
Website: https://www.chatpdf.com
Features:
Upload a PDF, and the system creates a chatbot interface to interact with the document.
Extracts and summarizes content.
Allows natural language Q&A.
Works well for research papers, textbooks, and reports.
Pros:
Free tier available.
Simple and fast user interface.
Good for academic documents.
Cons:
File size limit on free plan.
May struggle with complex formatting or visual data like charts.
Limited memory/context window.
3. Humata.ai
Website: https://www.humata.ai
Features:
Upload PDFs and chat with them instantly.
Designed for deep understanding of documents—especially research papers, contracts,
and legal docs.
Can generate summaries and explain complex sections.
Pros:
Attractive, modern interface.
Great for technical papers and scientific content.
Offers citation-based answers.
Cons:
Requires signup.
Free plan has restrictions on usage and file size.
4. AskYourPDF
Website: https://askyourpdf.com
Features:
Drag and drop PDFs to create a chat interface.
Uses embeddings and LLMs to find and answer based on relevant sections.
Chrome extension also available.
Pros:
Easy to use and fast setup.
Can integrate with ChatGPT API.
Document history is saved for future access.
Cons:
Some limitations in free version.
May misinterpret layout-heavy PDFs.
5. SciSummary
Website: https://www.scisummary.com
Features:
Tailored for summarizing scientific papers.
Supports PDF upload, summarization, and question answering.
Generates concise outputs for academic content.
Pros:
Highly specialized for research papers.
Useful for students, researchers, and academics.
Cons:
Not general-purpose (best for scientific PDFs).
Limited interactivity compared to other tools.
Task 2: Building a Custom Chatbot System
Objective:
Design and implement a chatbot capable of reading PDF documents, extracting key information,
and answering user questions interactively.
System Components:
1. PDF Parser:
o Use Python libraries like PyMuPDF, pdfplumber, or pdfminer.six to extract clean
text from PDFs.
o Handle multi-column layouts and images with OCR if necessary.
2. Text Chunking and Embedding:
o Split the extracted text into manageable chunks (e.g., using sentence or
paragraph breaks).
o Convert text into embeddings using models like OpenAI's text-embedding-ada-
002, Sentence-BERT, or HuggingFace Transformers.
3. Vector Store:
o Store embeddings in a vector database like FAISS or ChromaDB for similarity
search.
4. User Query Processing:
o When a user asks a question, convert the query into an embedding.
o Perform semantic search in the vector database to retrieve the most relevant
text chunks.
5. Answer Generation:
o Feed the retrieved chunks and question into an LLM (e.g., OpenAI GPT-4 or
HuggingFace model) to generate a contextual answer.
6. Chat Interface:
o Create a frontend using Streamlit, Gradio, or a web-based chatbot UI.
o Allow users to upload PDFs and interact via chat.
Tools and Libraries Used:
Python
PyMuPDF / pdfplumber
LangChain or LlamaIndex
OpenAI API or HuggingFace Transformers
FAISS / Chroma
Streamlit / Gradio
Outcome
A functioning chatbot that accurately reads PDFs and answers questions in real time.
Easy-to-use interface with upload, chat, and response features.
Potential applications in education, legal, healthcare, and research industries.
