To review and improve the classification results of problematic documents, complete the following steps.
- Switch to Definition Mode.
- Identify the engine that causes the problem.
- Determine whether classification fails because the confidence is too low or whether the distance is too low.
- If classification assigns documents to the wrong class, modify the learnset or the parameters for the engine that returns the wrong result.
- If too many documents are not classified at all, consider adding another classification method, or provide more samples for the learnset.
- If the previous steps did not help, you may have to change the settings for classification evaluation.