Hello,
I am looking for pointers how I can create a recommendation engine using the language models and training them on my product data. For instance, I have a database of products with descriptions; when a user prompts “I want a TV less than 50 inches and costs around $500 with 3 hdmi ports”, then I can return the matching products in the database.
I have been watching youtube videos around this topic and I could not find an exact use case. All I could find are tutorials feeding a custom knowledge-base to the model, and generating “facts” from those models. For example this video talks about feeding a financial report and asks details about that report. This video on the other hand trains the model how to write midjourney prompts so the model learns how to talk that way.
I am looking for a way to feed tens of thousands of product data into the model and associate responses to product ids.
Initially I was thinking of creating one huge wall of text that looks like “product id 1 is a tv with such and such features and costs xx dollars. product id 2 is a bed with such and such features and costs xx dollars” but that feels very inefficient and I’m not sure if that will achieve the objective.
Could somebody point me in the right direction how I can do this? Thanks!