Creating a Knowledge Base model

This document teaches how to create machine learning models using Knowledge Base algorithms so you can integrate Knowledge Base models in their automation.

The Machine Learning Model Builder tool, available in IBM RPA Studio, transforms knowledge base files into machine learning models called Knowledge Base models. The Builder, then, automatically imports the created model into the Tenant. The Machine Learning Model Builder also creates text classification models.

See details about creating a knowledge base file in Creating a knowledge base file.

Follow these steps to import and publish a machine learning model using Knowledge Base algorithms:

  1. Log in to IBM RPA Studio, selecting the desired Tenant.
  2. In the menu bar on the header, select the Tools tab.
  3. Click the Machine Learning Model Builder button.
  4. In the Name field, type in a unique name.
  5. In the Culture field, select the model analysis culture.
Important:The model supports the following languages: Brazilian Portuguese, Portuguese, and English.
  1. In the Algorithm field, select the desired algorithm:
  2. Knowledge Base V2
    Recommended algorithm. This algorithm can use an internal IBM Robotic Process Automation database storing synonyms for words and expressions, and the algorithm has improvements over the legacy algorithm.
    Knowledge Base
    Legacy algorithm. This algorithm does not use words and expressions synonyms.
  3. Optional: In the Options field, select Use synonyms if the selected algorithm is Knowledge Base V2 and you want the model to use words and expressions synonyms.
  4. In the File field, select the knowledge base file in XLSX format.
  5. Enable the Set as production option if you want to set this model version as the production version.
Note:The local IBM RPA service requests the production version if the user does not specify a version number.

IBM RPA Studio can train published Knowledge Base models. See Training a Knowledge Base model for more information. Bots can use published models. See Using a Knowledge Base model for details.

See also