For a training data source, what additional text is permitted apart from the Extract Text field?

In the context of a training data source, apart from the Extract Text field, it is permissible to include OCR (Optical Character Recognition) text or translation if applicable. This is crucial because OCR text refers to the text that has been extracted from images or scanned documents, allowing for textual analysis and training of models on a broader range of data formats. Additionally, if the documents contain content in multiple languages, translations can be included to enhance understanding and improve model accuracy.

Incorporating these forms of additional text expands the dataset's utility, enabling the model to learn from diverse sources of input. It enhances the training process by providing more comprehensive information that reflects various scenarios and textual presentations that the model may encounter in real-world applications.