Looking to test creating a language model for a retail use case. Would I be correct that I could use a multi-classification model to determine what someone is ordering?
For example, similar to Amazon Alexa, there are various utterances (ways of saying the same thing) that can map to an intent. Could we train a model to take all of those and map to a category?
Example:
- I want a red shirt
- Can I have a red shirt
- Let’s go with the reddish polo
Could all map to “add red shirt” to the cart. Therefore the utterances would be the inputs and something like “add red shirt” would be the category. We would need a category for each product.
Is this the right way to think about it?
Thanks!