Project settings
the NLU core of CAILA is a plug-in function for the platform. If the use of the NLU core is enabled for the account, then additional configuration parameters are displayed in the project settings.
More about creating and editing a project
When a project is created using the NLU core of CAILA, you need to fill out the following fields:
-
Template — select the Project for CailaPub template and the project will be configured for working with CAILA. Or, select any other project template from the list and configure the configuration file
chatbot.yamlfor working with CAILA. -
NLU language — select the language in which the bot and the clients will communicate. The language selected will affect preprocessing, i. e. the definition of intents and entities. You can select the interface language in the user profile settings.
-
Classification algorithm — select the classifier algorithm:
- STS — the intent recognition algorithm that uses the
Semantic Text Similarityfunction on text objects to find the best matching intent. The similarity function takes into account variants of words, normal word forms, synonyms and other linguistic information. This algorithm is recommended for use if the training dataset for the bot is small. - Classic ML — the classical machine learning algorithm for intent recognition. Its operation is based on normalized words and word combinations. The algorithm does not take semantic information into account. We recommend using it with selections that include 5–7 training phrases per intent where the number of intents is more than 7–10.
- Deep Learning — the algorithm based on deep neural networks that takes semantic information of words into account. It is recommended for use on large datasets: over 30 intents and at least 10 training phrases per intent.
- STS — the intent recognition algorithm that uses the
-
Spelling correction — when this function is on, intents and entities will be recognized while taking into consideration the variants of the respective words.
-
Timezone — specify the time zone, this parameter affects the recognition of entities related to time intervals: "now", "in 5 minutes", etc. If no information about the time zone has been provided in the client request, then the information from the project settings will be used.
-
NLU Settings — here you can set new parameters for configuring NLU as JSON object in text view, and also connect an external NLU service to your project.
Learn more about advanced NLU settings
Caila API key
The API key allows you to use a trained classifier in third-party applications.
Go to Project Editing. On the Classifier tab in front of the API-key Caila field, click Generate.
Project Import
To import a Caila project go to project editing. Go to the Classifier tab, attach or drag a file to the Import Caila project field.