... standard, thatis, the manual keywords are treated as the oneand only truth. The evaluation measures are preci-sion (how many of the automatically assigned key-words that are also manually assigned ... that of supervised machine learning,and the prediction models were trained on man-ually annotated data. No new training was done on the text categorization documents, but modelstrained on other ... cat-egorization task. One reason for this is that the keywords have been evaluated using manually as-signed keywords as the gold standard, meaningthat paraphrasing and synonyms are severely...