Which term refers to the estimated ability of the classifier to correctly categorize a document?

The term that refers to the estimated ability of the classifier to correctly categorize a document is classification confidence. This measure indicates how confident the model is in its prediction regarding which category or class a particular document belongs to. It is typically expressed as a percentage, showing the likelihood that the classifier's output is accurate.

In the context of machine learning and artificial intelligence, particularly when dealing with document classification tasks, understanding classification confidence is crucial. It helps in determining not just the predicted category but also how much trust can be placed in that prediction. For example, a classifier that outputs a high confidence level (close to 100%) in its classification is seen as making a very strong assertion about the document’s content.

Other options such as extraction confidence and OCR confidence relate to different aspects of document processing. Extraction confidence pertains to the accuracy of data extracted from documents, while OCR confidence is related to the accuracy of Optical Character Recognition technology in recognizing text from images. Stream accuracy might refer to the reliability of information over a continuous data stream but does not specifically address the classifier's categorization ability.