ChatGPT-Prompt-Engineering

☰ Table of Contents

• Goal
• Key concepts in the project
  • Prompting principles
  • Prompt development lifecycle
  • Capabilities of prompting
• References

🎯 Goal

Optimize the use and capabilities of prompting

Quick access to notebooks

• Summarizing_Prompt.ipynb

• Inferring_Prompt.ipynb

• Transforming_Prompt.ipynb

• Expanding_Prompt.ipynb

• Building_Chatbot.ipynb

• Prompting_Guidelines.ipynb

• Prompt_Development_Lifecycle.ipynb

💡 Key concepts in the project

Prompting principles

• Principle 1 - write clear and specific instructions

  • Use delimiters to quote the task and avoid prompt injection to ensure to execute content as a whole, i.e. '', "", ``, <>
  • Ask for structured output formats to respond, i.e. JSON, HTML
  • Check if the conditions are satisfied
  • Few-shot prompting

• Principle 2 - give the model time to 'think'

  • Specify steps to complete a task
  • Instruct model to think before conclude and complete a task

Prompt development lifecycle

• Process
  • Idea/ Problem
  • Implementation
  • Experimental results
  • Error analysis

Capabilities of prompting

• Summarizing

  • Summarize a task content based on prompt
  • Can be single/ multiple tasks
  • For multiple tasks
    • Define a List of all required task contents
    • Define for-loop of List
    • Print( i, response, "\n")

• Extracting

  • Extracting is part of natural language processing (NLP)
  • Can be included in the process of summarizing and inferring

• Inferring

  • Help provide basic understanding of a task content
  • Classify sentiment of a task content (Yes/ No)
  • Identify specific sentiments of a task content (What kinds of sentiments)
  • Extract keywords as an understanding of a task content
  • Can be single/ multiple tasks

• Transforming

  • Identify language
  • Translate from language A to language B
  • Tranform to formal and informal tone
  • Tranform from format A to format B, i.e. HTML, JSON
  • Transform from text A to text B, i.e. re-writing, shortening
  • Check gramma and spelling
  • Can be single/ multiple tasks

• Expanding

  • Create a response including all the other functions, i.e. extract and summarize the task content for creating a response

📚 References

• OpenAI API: OpenAI API access
  • if there is rate and region limit, check out Limit guide
• Best practices for prompt engineering Prompting FAQ
• Prompting with OpenAI examples: Examples
• Reference task contents for prompting: DeepLearning.AI

🛠️ This project is supported by DeepLearning.AI and OpenAI.