Hi!
I would like to know if ChatGPT or another appropriate AI tool is suitable to identify suitable category paths for a list of offer titles/product titles from a predefined list of category paths. The most suitable category path is to be identified for each offer title.
Simplified example:
Given offer title: “Dining Room table mizzo extendable white”
List of category paths:
Furniture & Living > Furniture & Fittings > Sofas
Furniture & living > Furniture & furnishings > Living room sofas
Furniture & Living > Furniture & Fittings > Tables
Furniture & Living > Furniture & Fittings > Chairs
Furniture & living > Furniture & furnishings > Other furniture
Garden & Plants > Garden furniture > Garden chair cushions
Garden & Plants > Garden furniture > Garden tables
Garden & Plants > Garden furniture > Garden benches
As a result I expect “Furniture & Living > Furniture & Fittings > Tables”, because the offer title clearly refers to a table and it is not a garden table.
Can this use case be realised with the help of ChatGPT 3 or 4 or another AI tool? The Excel list with several thousand category paths is very extensive.
Thank you in advance!
Markus
i’m looking to implement something like this as well - did you find an optimal solution? any suggestions?
So far, no. I think ChatGPT would be intelligent enough for that, but it lacks a way to provide the AI with the category list.
@chadmoon
I just asked ChatGPT itself. Here is the answer:
Yes, ChatGPT or another AI tool can potentially be used to identify suitable category paths for a list of offer titles/product titles from a predefined large list of category paths. This is a classic example of a text classification problem, where the task is to classify a given text (in this case, the offer title) into one of several predefined categories (in this case, the category paths).
One approach to solve this problem using ChatGPT would be to fine-tune the pre-trained language model on a labeled dataset consisting of offer titles and their corresponding category paths. During fine-tuning, the model would learn to map each offer title to the most appropriate category path. Once the model is trained, it can be used to predict the category path for new offer titles.
Alternatively, you could use a pre-trained text classification model such as BERT, which is specifically designed for text classification tasks. You would need to train the model on your labeled dataset, and then use it to predict the category path for new offer titles.
In either case, the quality of the results would depend on the quality and size of the labeled dataset used for training, as well as the quality of the category paths themselves. The Excel list with several thousand category paths is extensive, and you would need to ensure that it is properly formatted and cleaned before using it for training or prediction.
There are several ways in which you can provide the large list of category paths to ChatGPT. One way would be to convert the list into a suitable format, e.g. a CSV or Excel file, and then import this file into the training programme for ChatGPT.
1 Like