Currently, everything is operating in a fully automated manner, which may occasionally result in the presence of unusual categories.
The rationale behind this approach is as follows: (1) Initially, a minimum of 50-100 "auto-generated" categories are gathered. (2) a concise selection is made, focusing on the most prevalent and pertinent categories. (3) Ultimately, the aim is to refine the algorithm to ensure that all news items are placed within the shortlist identified in Step 2.