Generative AI project in e-commerce domain

Hii so i am doing a project on product description generation using the user reviews and large language models.
I am having a doubt in the data that is to be finetuned with.
My dataset has only user reviews is there any need that my dataset should be labelled before finetuning or the dataset that i am using should be a labelled data

There must be some labelling to the fine-tuning dataset similar to how the original LLM model was trained with, or at least some label that instructionally the LLM can understand!

1 Like

Thank you for the reply
The as disscussed earlier the i am having the reviews data and it needs to be labelled what are the possible labelling i can do can it be the senitment fror each of the reviews and and the review with the sentiment can be used for the fiinetuning of the large langauge model for the task of product description generation that is more or like a summarization of the what customers say about the specific product

Hello @Aravindraj_DS
Have you also considered using prompt engineering and few-shot learning? If the model you’re using is a SOTA testing these approaches could also be worth it.