Hello, I have a question. help me plz T.T

I am creating a GPT4 API.

After giving GPT4 limited categories and criteria, we ask it to answer which category the example product name best fits.

example)
{
“model”: “gpt-4”,
“temperature” : 0,
“messages”: [
{
“role”: “system”,
“content”: “Category [Underwear, household appliances, jewelry, health food, kitchen appliances, household goods, kitchenware, miscellaneous goods, general food, furniture/interior, bedding, beauty, leisure sports, travel, clothing, rental, insurance, other, mobile products ] Among them, ["Kitchen appliances"] for products used in the kitchen and use electricity, ["Kitchen appliances"] for products used in the kitchen and do not use electricity, ["Kitchen appliances"] for products used in the kitchen and use electricity In the case of products, "household appliances"], [in the case of daily life-related products that do not use electricity, "household goods"], [in the case of bags, shoes, and sunglasses, "miscellaneous"], [automobile-related , In the case of medical devices, camping, leisure, massage guns, and massagers, please answer only one category, referring to "Leisports"], [If the word consultation is present but not related to insurance, "Rental"]”
},
{
“role”: “user”,
“content”: “Lizaday Camella Tote Bag”
}
]
}

Currently, I am sending category information to the system side of the API, but is there a better way?

Also, I would appreciate it if you could give me a better method and example, excluding fine-tuning.

Translated using a translator.
The question may be strange.
I’m sorry.
have a good day

Welcome to the community!

Categorizing things is often done by employing embedding model vector comparison.

You can try to write out category descriptions, embed them, and then see what matches most closely to your search query :slight_smile:

2 Likes

Oh hello teacher!!

When comparing embedding model vectors, is it okay to not continue to include criteria and categories in “system” like in the written example?
They say the token is too big and need to be reduced.

I’m talking about this:

https://platform.openai.com/docs/guides/embeddings

The decision to include examples when comparing items to category descriptions will depend on a number of factors, but I’d opt for not doing that because it might confuse things.

Thank you for your reply.

You are saying that it is not a good idea to provide standards because it can cause confusion…

The link you provided does not work.

OMG~ T.T.

Apologies: here’s the link

https://platform.openai.com/docs/guides/embeddings

—-

Yes: you can try to either match to descriptions or to examples. It’s sometimes a balance that you need to find.

Thank you for your kind explanation.

I’ll try again by referring to the URL and if it doesn’t work, I’ll write a post.

thank you
have a good day!

2 Likes