Have you ever asked AI to write some text or analyse some code, only for it to spout a load of clichés? It’s not the neural network’s fault – it’s down to how you’ve phrased your prompt!
Our computational linguists have shared some basic rules that’ll save you hours:
Set the role and context:
Instead of ‘Explain what Python is’, write: ‘You’re a Senior NLP Developer.
Explain the basic syntax of Python to a first-year student using simple examples from text analysis’.
Provide negative constraints (Negative Prompts):
Specify what NOT to do.
For example: ‘No introductory phrases, no clichés like “in today’s world”, be concise, stick to the facts’.
Use a step-by-step approach (Chain-of-Thought):
Don’t lump all the tasks together.
Write: ‘Step 1: find the error in the code. Step 2: explain why it occurred. Step 3: provide the corrected version.’
Bookmark this so your lab work and coursework get done twice as fast! 📌
