Skip to content

Latest commit

 

History

History
79 lines (55 loc) · 3.12 KB

README.md

File metadata and controls

79 lines (55 loc) · 3.12 KB
Raw

Anatomy Chatbot

This chatbot, implemented using the Gemini API, is designed to train on anatomy-based textbooks in PDF format. It processes the content of the textbooks and allows users to query the material. The chatbot leverages advanced NLP techniques to provide accurate, context-aware responses to any anatomy-related questions, making it a valuable tool for students, researchers, and medical professionals seeking quick and reliable information from the text.

Installation

create the project directory

mkdir anatomy-chatbot

move to the project directory

cd anatomy-chatbot

clone the repository

git clone https://github.com/lohithgsk/anatomy-chatbot.git

Usage

Get the GEMINI API KEY from the https://ai.google.dev/gemini-api/docs/api-key.

Add the API KEY to the .env file.

API_KEY = ''

Install the dependencies from the requirements.txt file

pip install -r requirements.txt

Run the Streamlit application

streamlit run app.py

Architecture

Architecture

Phase 1: Core Chatbot Functionalities

  • Data Extraction and Cleaning: Extracted relevant text, images, and tables from anatomy PDFs using the Gemini API. Data was cleaned by removing irrelevant sections and stored in a database for fast access.

  • Increasing Text Retrieval Accuracy: The Gemini API was used to enhance the chatbot's ability to retrieve accurate and relevant text based on user queries. Text was indexed for fast retrieval.

  • Support for Probing Questions: When the chatbot is unsure of a query, it asks clarifying questions to improve accuracy. This was done by setting confidence thresholds in the response.

  • Retrieval of Relevant Image and Table Content: Images and tables were tagged during data extraction. Relevant visual data is fetched and presented along with the text for user queries.

  • Latency Optimization: Caching and efficient processing were implemented to reduce response times. Cloud services were used to ensure scalability and quick responses.

Phase 2: Advanced Chatbot Functionalities

  • Image-Based Prompts: The chatbot processes image inputs, using the Gemini API to interpret and respond to visual data such as anatomical diagrams.

  • Flowchart Generation for Relevant Queries: The chatbot generates flowcharts to simplify complex anatomical processes, enhancing user understanding of the queried concepts.

  • Table Reconstruction: The chatbot reconstructs tables using data extracted by the Gemini API, ensuring accurate and structured presentation for the user.

  • Answering Reference Questions at the End of Each Chapter: The chatbot accurately answers reference questions by linking the query to specific sections of the anatomy textbooks, using indexed content.

  • Final Latency Optimization: Asynchronous processing and additional caching were implemented to handle higher loads and further reduce response times.

Feedback

If you have any issues, reach out to us.