Product Classifier
In this example we demonstrate how we can use a model on Mixlayer to classify products according to a specific category taxonomy. We provide the model with a list of categories to select from and use constrained generation to limit what it can pick. To increase the reliability of the output a bit, we ask the model to provide some reasoning ahead of selecting its choice.
product-classifier.js
let categories = ["Shoes", "Shirts", "Pants", "Hats", "Socks"];
let product = "Nike AF1s";
// let product = "Levi's 501 Original Fit";
// let product = "Starter Flat Bill Trucker ";
prompt("You are an expert at categorizing apparel\n", {role:'system', hidden: true});
prompt("Categories are " + categories.join(", ") + ".\n", { role: 'user' });
prompt(`What can you tell about the product '${product}' by just its name?`, { role: 'user' });
gen({limit: 1000, role: 'assistant'});
prompt(`What category does '${product}' belong to? `,{ role: 'user' });
prompt(`Please just say the category, nothing else.\n`, { hidden: true, role: 'user' });
let cat_re = categories.join("|");
gen({ regex: cat_re, role: 'assistant' });output.txt
Categories are Shoes, Shirts, Pants, Hats, Socks.
What can you tell about the product 'Nike AF1s' by just its name?
Based on the name 'Nike AF1s', I can categorize it as 'Shoes'.
What category does 'Nike AF1s' belong to?
Shoes