One of important tasks in E-commerce, marketing and other fields is categorization of products.

In this post, we will show you many examples of product categorizations, with a machine learning based approach, which will also tell you why a given product was categorized into specific category. With a nice visual explanation for each product categorization.

For this purpose, we will be using a leading, highly accurate product categorization service that we developed and that is available on https://www.productcategorization.com/.

It is used by E-commerce companies, Saas platforms and many other users.

We invite you to check it out, it has an option of free categorizations as well.

Before we turn to product categories examples let us say a few words about machine learning models and AI explainability.

AI explainability (XAI)

To offer product categorization at scale, i.e. ability to classify millions of products in some reasonable amount of time, it is necessary to resort to some sort of automation.

This is why most of product categorization services rely on machine learning models.

One of the specific features of machine learning models is that they give you results, but mostly do not offer an explanation for it. E.g. they may same that some product is from category Automotive, but will not say why this is so.

This non-explainability of decisions made by machine learning models can be even more annoying when real-life decisions of great importance are being made, e.g. you go to the bank for a loan and application is rejected by an agent, which is really relying on some automated, machine learning based program in the background. But you do not exactly why this decision was made, was it the credit score or some other factor.

This “black box” nature of machine learning models has led to the desire to make their results explainable, in language that us humans can understand.

This AI explainability or XAI is becoming an important part of AI application as more and more regulators and regulations demand it, e.g. GDPR. Especially in the decisions where people are affected by the machine learning decisions.

Let us now turn to product categories.

Product categories examples

For generating product descriptions we will use https://beta.openai.com/playground. It is another way of using AI, this time for quickly generating texts that we want.

Example 1 – bird house

One just has to enter prompt or command, e.g. “write description of bird house” and one gets result back from openAI solution:

A bird house is a small wooden or plastic structure designed to attract birds so that they will nest in it. The entrance hole is usually round or oval, and just large enough for the adult bird to enter. The bird house may have a perch on which the bird can rest, and a roof to keep the bird dry.

Now let us use our service at productcategorization.com and try to classify this specific text. The service offers 3 tiers of classifications, each higher Tier has larger number of categories supported. The highest, Tier 3 classifier supports 1113 possible categories.

Let us try the middle option, Tier 2 categorization of our bird house text.

This is the screenshot from the results:

The resulting categorization -> Home & Garden / Decor means that Tier 1 category is Home & Garden and Tier 2 category is Decor. This “path” is also known as full taxonomy path and you can often see it on the product pages as they allow for easier searching.

The words that most contributed to resulting classification are entrance, house, bird, birds and rest.

Example 2 – Dish Washer

Let us continue with another product, dish washer. We will be asking OpenAI to “write a long description of benefits of dish washer.

We get:

The dishwasher is a great invention that has made life easier for billions of people.

Here are just some of the many benefits of dishwashers:

1. They save time

Dishwashers can importantly reduce the amount of time you spend on cleaning dishes. Rather than spending hours scrubbing pots by hand, you simply load them in the dishwasher and it will do all the work for you.

2. They save water

Dishwashers use far less water than washing dishes by hand. Actually, some dishwashers use as little as 3 gallons of water per washing cycle. This can make a large difference if you’re trying to conserve water in your home.

3. They save energy

Dishwashers are energy efficient. They use less water and less soap, which then means that they use less energy overall. This can be a great way to save money on your energy bill

4. They’re convenient

Dishwashers are extremely convenient. They can be used at any time, day or night. And, they can be loaded and unloaded without any hassle.

Sending this to our categorizer at productcategorization.com, this time let us try Tier 3 version, we get the following categorization, together with explainability:

The category: /Home & Garden/Kitchen & Dining/Kitchen Appliances is correct for our dishwasher text, with three words most contributing to it: dishwashers, dishwasher and dishes. Note that with Tier 3 we get all three Tiers back – Home & Garden (Tier 1), Kitchen & Dining (Tier 2) and Kitchen Appliances (Tier 3).

Example 3 – Bracelets

Next, we are interested in bracelets:

A bracelet is an article of clothing or jewelry. It is made of various materials, including metal, leather, cloth, beads, or plastic, that is worn around the wrist.

Bracelets can have multiple meanings and purposes, they can be personal, aesthetic, or religious/spiritual.

In many cultures, bracelets are believed to have healing  with the history of bracelets being as old as the history of humanity.

In ancient Egypt, bracelets were worn as a symbol of wealth and power. In the Middle Ages, knights wore bracelets as a sign of courage. In the Renaissance, bracelets were a sign of social status, which is kind of what is often the case today.

Sending this text to classifier gives us the following categorization:

with resulting category being /Apparel & Accessories/Jewelry/Bracelets.

Note how detailed the Tier 3 categorizer is, it even has a category specifically for Bracelets.

Words that most contributed to resulting categorization is bracelets, beads, wealth and bracelet.

Conclusion

We have demonstrated example of product categories for a set of texts from E-commerce domain, with product categorization obtained with machine learning model from https://www.productcategorization.com/demo_dashboard/. 

We invite you to try it out on your own, with your own examples.

We are also adding below the list of product categories that are possible for Tier 1 and Tier 2 classifiers of the service. Tier 3 list is too large to be included in the post, but can be obtained on the website in results.

Appendix – Product categories list

Tier 1 categorizes has the following possible categories:

Tier 2 categorizer has the following categories: