If the graphs at the Training tab contains a lot of documents with probabilities close to 50%, the model needs to be retrained. For this, the user should mark documents as relevant or irrelevant. One needs to mark up documents for which the probability value is close to 50%, or if the document is clearly marked by mistake (the calculation shows a probability of 80%, and the document is clearly not of class 1, or vice versa).
- In the Predictive coding section click the name of a model.
- Click the Training tab.
- Click the document name.
- Click the Relevant button if the document is relevant.
- Otherwise, click the Not relevant button.
You can also mark documents as relevant or irrelevant in the list.
- Select several documents.
- Click the ⋮ icon in the upper right corner of the table.
- Select Relevant or Not relevant.