Custom Text Classification

This is a tutorial on Custom text classification


Custom text classification is a feature that allows you to classify texts in different category. The custom text classification is based on the user's data (texts and categories associated). The engine learns from the user's data in order to be able to classify texts.

Use Case

Imagine that you receive tons of messages from your customers (via your website for example). Everyday, you need to have someone in the company classifying the messages in order to redirect the customer to the right guy and to provide the right response. This represent a long and tedious work and it would be great if an automated system could categorize them so that the right person sees the right comments.

Custom text classification is here for you! You will give to the text classification engine many samples of messages that were classified manually, the engine will learn from those examples and be able to predict the category of all the next messages!


Why is AI text analysis the right tool for this problem?

Text analysis is also called NLP (Natural Language Processing). This regroups two types of technologies in AI as a Service:

  • Pre-trained engines: detection of syntax / entities / keywords / sentiment, topic classification, text summarization, question answering, etc. All these features are very powerful most of the time. But sometimes the data is too specific and results are not good, so you can use:
  • Custom text engines: custom text classification / entity recognition / sentiment analysis / etc. These features use data from the user to train a custom model. Once trained, this model can be requested to perform prediction.
    The providers are not particularly clear about all the technical bacnkground (mathematic and data science methods used). We know that it usest two principal concepts:
    • Neural architecture search, which automates the design of neural networks. This helps custom text classification engines discover new architectures for problems that require them.
    • Transfer learning, in which pretrained models apply what they've learned to new data sets. Transfer learning helps custom text engines apply existing architectures to new problems that require it.

Do I have to use pre-trained Text / NLP APIs OR Custom text classification?

The NLP / Text APIs allows you to detect syntax, entities, keywords, sentiment in text, and classifies text into a predefined set of categories. If your text consists of news articles or other content you want to categorize, or if you want to discover the sentiment of your examples, the Natural Language API is worth trying. But if your text examples don't fit neatly into general topic-based classification API, and you want to use your own labels instead, it's worth experimenting with a custom text classification to see if it fits your needs. The custom text classification model will be much more adapted to your use case.

What’s Next

Let's have a look at the Workflow feature OR see complete tutorial for Custom text classification

Did this page help you?