/Chatbot

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

DWRChatbot

Welcome to Our Drinking Water Regulations Chatbot Repo!

This project aims to build a chatbot that is capable of answering user questions about California Drinking Water Regulations. The project will be documented in a paper that may be published in an industry journal if conditions are sufficient.

Project team members include talented students participating in the UCB Data Science Discovery Programwater as well as industry professionals at the California Data Collaborative and the California Water Resource Control Boards.

A pdf copy and a word copy of the regulations can be found in the Regulations folder. Alternatively the regulations can be scraped from the following website: https://govt.westlaw.com/calregs/Browse/Home/California/CaliforniaCodeofRegulations?guid=I717E98405B6111EC9451000D3A7C4BC3&originationContext=documenttoc&transitionType=Default&contextData=(sc.Default)&bhcp=1

The examples folder includes the following code examples:

  1. Chatbot Tutorial from PyTorch: https://pytorch.org/tutorials/beginner/chatbot_tutorial.html
  2. How to Use OpenAI API for Q&A and Chatbot Apps: https://help.openai.com/en/articles/6643167-how-to-use-openai-api-for-q-a-and-chatbot-apps
  3. How to build an AI that can answer questions about your website: https://platform.openai.com/docs/tutorials/web-qa-embeddings
  4. Question answering example notebooks: https://platform.openai.com/docs/guides/fine-tuning/example-notebooks
  5. ChatGPT @ Home: Large Language Model (LLM) chatbot application, written by ChatGPT: https://youtu.be/QumfkMQr47M
  6. Fine-tune a GPT3 model (Bloom) with the Amazon Customer Reviews Dataset: https://github.com/data-science-on-aws/data-science-on-aws/blob/gpt3/00_quickstart/06_Train_and_Fine_Tune_GPT3_PyTorch.ipynbv

Student team members are encouraged to review the examples, select one code example, and test run the codes. Alternatively, students may find and test other code examples elsewhere that are even more suitable to this project. The team will discussed results of this exercise and may select the most promising code example to customize for this project or develop new codes.