The invention of artificial intelligence and natural language processing has revolutionised
human-machine interaction, and OpenAI's ChatGPT models are at the forefront of this.
GPT-3 and GPT-4 models generate human-like text, answer questions and perform all
kinds of language-related tasks. ChatGPT relies heavily on the quality of the requests
it receives. This paper addresses the issue of efficient prompting in order to maximise
its potential. Different prompting
techniques are described in the paper: prompt template design, continuous prompts,
few-shot learning, chain of thought prompting and metadata expansion. The conclusions
show that learning how to write good prompts is crucial for realizing the full potential
of ChatGPT in various natural language processing tasks.
