Note: this article relies on a basic knowledge of how Data X-Ray Classes and Classifiers function. If you have not done so yet, we recommend reading this article before proceeding.
What are Anti-Classes?
Anti-classes are essentially a special type of Class that is used to filter out false positives. If you think about a vending machine, it is the coin return slot when a coin has not fit into other holes.
The Data X-Ray calls its anti-class "Unclassifiable Text". The class consists of data that might "look like" data that you actually want to classify, but that are not actually that data so you want to filter it out of consideration in the context of a particular Classifier.
An example might be if you were a car parts manufacturer with a parts catalog number of
B4244-4432-1193 . This number looks very similar to other types of numbers such as United States Social Security Numbers (
123-45-6789 ) or even phone numbers (
07-1234-56789 ). Therefore where these nuances present themselves, you as the user may need to teach this nuance to the machine learning model you are building.
Build Your Own an Anti-Classes
The Ohalo team has already pre-loaded a fairly comprehensive anti-class for you called "Unclassifiable Text". If you simply check the box to include this default class into any homegrown classifier that you have built, you will get fairly good results out of the box. However, when nuance presents itself, you will need to build another Class that reflects that represents the nuance, categorize it as a "non-sensitive" (or other category, as appropriate) Class, and add it to the Classifier.
If you have any questions, do not hesitate to email us at [email protected] or simply click the Intercom button to start a chat with our support team.