Creating keyword-based topic detection models

You can create a topic detection model for keyword-based analysis in the Analytics Center. Keyword-based categorization models can act as substitutes or supplements for machine learning-based categorization models in cases in which machine learning models are undeveloped or do not produce satisfactory results, for example, they have low prediction accuracy.
  • Make sure that you can access the Analytics Center. You can do this by starting the pyDecisionAnalytics portal. Add this portal to the list of portals in your access group. For more information see, Access Group form - Completing the Definition tab.
  • Make sure that the system locale language settings are set to UTF-8.
  • Create a CSV, XLS, or XLSX file that contains a taxonomy, which is a subject-specific collection of categories that you want to assign to the analyzed content. Each category in a taxonomy is identified by a set of keywords ( should words, must words, and and words ).

For more information about taxonomy file definition, see Requirements and best practices for creating a taxonomy for rule-based classification analysis on the PDN.

  1. In Designer Studio, click Launch > Analytics Center.
  2. In Analytics Center, click Create, and then click Text categorization.
  3. Specify the name of the categorization model.
  4. In the Detection type field, select Topic.
  5. In the Creation section, select Use category keywords.
  6. In the Language section, expand the drop-down list and select a language for the model.
  7. Click Choose file to select and upload a taxonomy file from your directory.
  8. If any problems with the file that you selected occur, in the Errors section, perform the following actions: or improperly formatted columns.
    1. Based on the provided error descriptions, correct the problematic records.
    2. Re-upload the file.
  9. In the Save model section, finalize the creation of a keyword-based categorization model by providing its application context:
    • To use the default rule context for decision data rules that contain sentiment analysis models, select Use default context.
    • To manually specify the Applies to class, ruleset, and ruleset version parameters of the new rule, select Specify context.
  10. Click Create.