Need to find positive Portuguese customer reviews? A prompt is the question and context sent to a generative AI to guide its response. Unlike a search query made mainly of keywords, it can also specify the task, audience, background, and desired form of the answer. Its wording matters: a small change in phrasing can change the response. section 5.5.3 section 1.1
Make the request useful, then check what comes back:
- State the problem clearly and concisely. Include relevant examples or background, and specify a format when the result must feed another step. section 1.3.6 section 8.4.2
- Inspect the answer and refine the prompt if necessary. Keep notes on prompts whose outputs you have verified, including errors to avoid. section 1.3.6
- Check consequential results against records or other reference points. A plausible answer can still be wrong, and a model may agree with a challenge even when its first answer was right. section 1.1 section 6.2.2
From a request to a review workflow
Ask: “Generate a list of 20 keywords indicating positive sentiment to be used for searching customer reviews in Portuguese.” A Python API call can return a list for searching reviews, but inspect its format before processing it. The requested 20 keywords are not the same as a 20-token output limit: a low max_tokens setting can truncate the list. For a different approach, ask for the sentiment of supplied Portuguese reviews, then compare the answers with review scores before relying on a larger run. section 5.5.3 section 6.2.2
Listing 5.5 Generating keywords via ChatGPT API
from openai import OpenAI
def generate_keywords(temperature=0.5, max_tokens=150):
client = OpenAI(
api_key="your_api_key", # Make sure to use your actual API key here
)
prompt = """Generate a list of 20 keywords indicating positive
sentiment to be used for searching customer reviews in Portuguese."""
try:
response = client.chat.completions.create(
messages=[
{
"role": "user",
"content": prompt,
}
],
model="gpt-4-0125-preview", # Replace with the model you have
access to
temperature=temperature,
max_tokens=max_tokens,
)
return(response.choices[0].message.content.split("\n"))
except Exception as e:
print(f"An error occurred: {str(e)}")
# Example usage
print("Generated Keywords:")
print(generate_keywords())
The Python listing places the keyword request in a user message to the ChatGPT API and splits the returned text into lines. Treat that split as something to verify, not a guaranteed output format; check current API documentation if the call needs updating. section 5.5.3
More detail is not automatically better. Supply context that changes how the task should be interpreted—for instance, known sarcasm in text being classified—and ask for the explanation or format you need. Test revised prompts rather than trusting a polished answer at first sight. section 8.4.2 section 1.3.6
