Support Article
Text Analytics Model detects incorrect language
SA-60258
Summary
Pega 7.2 is updated to Pega 7.4.
The Text Analytics Model, imported from Pega 7.2 or built from the start in Pega 7.4, detects incorrect text language and the Topic Detection fails to detect any category.
Error Messages
Not Applicable
Steps to Reproduce
- Log in to Pega 7.4.
- Launch the Analytics Center portal.
- Click Create > Text Categorization to create a new model.
- Select the following options:
Detection Type - Topic
Set Creation - New Model
Set Language - English
- Click Start.
- Upload the Taxonomy file and the Training dataset provided by the user.
- Select the below options to generate the model.
- Uniform Sampling
- Support Vector Machine
- Open the generated model and click Test.
- Provide the sample text as below and click Test.
Topic Detection Preference - Always use rule based topic detection
Granularity - Sentence level
The model run result detects incorrect language and the Topic Detection returns no categories.
Root Cause
An issue in the custom application code or rules.
Taxonomy functionality was implemented post Pega 7.2 to provide more accuracy to Text Analysis. The issue occurred because the Taxonomy file was not defined for the model.
Resolution
Perform the following local-change if the Taxonomy file is not defined:
Upload a dummy Taxonomy file and run the model. For better accuracy from the model, refine the Taxonomy file.
Published March 30, 2019 - Updated December 2, 2021
Have a question? Get answers now.
Visit the Collaboration Center to ask questions, engage in discussions, share ideas, and help others.