How do Deep categorization models work?

Deep categorization models enable you to define powerful and flexible rules that detect key pieces of information and use them to assign relevant categories. Their functionality provides for advanced categorization capabilities that are able to process text considering morphosyntactic, semantic and contextual information by using an expanded set of operators and parameters to do so.

These models are perfect for those scenarios where the a high linguistic accuracy is needed, or where the use of lemmatization or semantic information simplifies the categorization task enormously.

For example, a possible application would be to classify documents according to their degree of confidentiality, so that if sensitive information is detected in the text such as IDs or full names, accessibility will be restricted. Likewise, another example could be Voice of the Customer detection, where various models are used to analyze different aspects of customer feedback: the quality of customer service, the product the feedback relates to, the channel used, etc.

If you are already familiar with MeaningCloud, you'll probably know that there is another resource that focuses on text classification: classification models. The difference between these two resources is the scenarios in which they are applied. Text Classification is designed for large classification models where it's not necessary to define rules with great detail in order to successfully classify a text. Deep Categorization, on the other hand, is designed for categorization scenarios where there are not as many categories, but the casuistic for each one of them is complex enough that advanced morphosyntactic and semantic rules are needed for classification with an acceptable degree of precision/recall.

Deep categorization models are one of the analytics services available in MeaningCloud's APIs. In the following sections, we'll learn how to create and use models. You can try it out now.