Records in Dimensions will not be classified if there is not enough information for the classification process to be carried out accurately. At least 200 characters in the abstract/summary section are required to perform a sufficiently reliable categorization. Any less than this and the document may not be categorised, and may not appear in search results using categories as filters. (With Fields of Research classification codes, we now use the journal-level FOR classification code, if it is available, to label articles from that journal. Therefore, when using Fields of Research codes you may find there are codes assigned, although there is no abstract, or not enough text to normally assign a classification).
Is there a certain threshold required for a category assignment?
Yes - these are dependent on the training information to produce the original classification. Category assignment thresholds are used by the classifying algorithm, and these are based on information that has been learned by the machine learning from manual coding carried out by subject matter experts. Therefore the “thresholds” that are commonly used for manual category allocation will be reproduced by the Dimensions emulation.