If you have a data stream that ingests on a recurring basis or wish to use an existing taxonomy on a new data set, you may be interested in making edits to your taxonomy logic to improve your analysis. However, you may find that even after you make your edits, the output from your Taxonomy Analysis is not reflecting your edits. In this article, we will walk through three ways to troubleshoot this issue.
Ensure that you have saved your taxonomy edits before running the model.
A common cause for issues with new taxonomy logic not being reflected results is an edit not being saved prior to rerunning a taxonomy model. This can be particularly confusing because a user can make edits to their taxonomy, leave the Taxonomy Editor, and return to the Editor to find that their edits have not been lost.
A quick way to see whether you have saved your taxonomy edits to check whether any of your labels are orange or green colored while working in the Taxonomy Editor. Labels colored green indicate that it is a new label, whereas labels colored orange indicate that it is an existing label that has been edited.
If your taxonomy edits have not been saved, you will also see a blue save icon in the upper right-hand corner of the Taxonomy Editor. Clicking this button will save your edits and present you with a popup window to reprocess models that are using your taxonomy.
2. Ensure your model is running on the correct field.
When deploying a new taxonomy model in a workspace, the platform will automatically select a text field for your taxonomy to run on. However, the preselected field may not be the field you wish to analyze. You can check to see if you are running your model on the correct field by going to the Settings Tab of your workspace, clicking on the taxonomy model in question, and checking to see if the Text Field option displays the field you wish to analyze.
3. Double-check that your logic captures words or phrases that appear in your data.
A final way to troubleshoot is to check to see whether your logic is capturing words or phrases available in your data. If no records analyzed by your taxonomy model meet a label’s logic, then the label will not appear in your output.
A quick way to check whether your data records that meet your taxonomy logic is to search for a specific word or phrase targeted by your logic by filtering on your data. If you have a simple widget that displays the total number of records in your dataset, you can apply a global filter that targets the field being analyzed by your model. Using the “contains” parameter, you can search a word or phrase being used by your taxonomy logic. If your data contains the word or phrase, your Number of Records widget should be greater than 0. You can do further digging by moving over to the Data Tab of your workspace to look at all the records that meet your filter’s requirements.
In the example below, we are looking at Kickstarter Projects. Suppose I want to categorize the records in a taxonomy that groups the records by project category (Fashion, Technology, etc.)
Please sign in to leave a comment.