Skip to main content

Project settings

To configure NLU for a project:

  1. Click  in the project card → select Project settings.

  2. Fill in the fields on Classifier tab:

    • Classification algorithm — select the classifier algorithm:

      • STS (Semantic Textual Similarity) is an algorithm that compares the semantic distance between words. It takes into account inversion, dictionary forms of words, their synonyms, and other information. It is recommended to use this algorithm when training the bot on a small dataset: 5–7 training phrases per intent, but no more than 1,000 phrases in the whole dataset.

      • Classic ML is a standard machine-learning algorithm for logistic regression-based intent recognition. It classifies data by source words, dictionary forms, and word stems. Semantic information is not taken into account. It is recommended to use this algorithm with a dataset of at least 20 training phrases per intent.

      • Deep Learning is an algorithm based on convolutional neural networks. It takes into account the semantics of words when forming hypotheses. It is recommended to use this algorithm with a significantly large dataset: at least 50 training phrases per intent.

      • Transformer is a multilingual algorithm. It evaluates the semantic similarity of the user request with all the training phrases from the intent. It is recommended to use this algorithm with a dataset of at least 10 training phrases per intent.


        The Transformer classifier runs on the Caila platform. If you have selected this classifier type for your project, you can apply advanced settings to replace the built-in Transformer with any other Caila classifier.

    • Spelling correction — enable this function to correct spelling errors in user requests. Only Russian and Ukrainian languages are supported.

    • Search for matches — enable this function to search matches:

    • Timezone — specify the default time zone. If there is no time zone information in the user request, the information from the project settings will be used. This parameter may be important when working with time recognition entities.

On the NLU settings tab, you can set parameters as a JSON object to configure NLU and connect an external NLU service to your project.

NLP Direct API key

The NLP Direct API key allows you to use a trained classifier in third-party applications.

Go to Project settingsClassifier → save NLP Direct API key.

NLU data and settings import

To import data and settings:

  1. Go to Project settingsClassifier.
  2. Attach or drag the file to the Import NLU data and settings field.