| Popis: |
The motivation of OST to improve the disciplinary nomenclature was strong and the solution proposed by Milojevićseemed attractive. By keeping the historical classes without changing their contours too much, we can move to a classification by article that improve the quality of the classification. This quality will not be competitive with classifications produced with pure algorithms as the Leiden algorithmbut our classification does not upset the usual names of classes. We also revised the classification in a higher level of the nomenclature in maintaining a strict hierarchical structure between the two levels. However, the implementation requires a lot of care. The overall level of content similarity between the old and new classification is just over 50%. This was expected but the interpretation of the new classes must be careful. In addition, the volumes of the classes are sometimes widely modified: large categories in the initial classification have more references in the whole corpus and are more often cited. Therefore documents quoting them add to these categories. On the other hand, new fields with few specialized journals may disappear. The same is true for interface fields that are redistributed in their parent disciplines. Therefore, the new classification misses recent emerging fields introduced by Clarivate. The use of this revised nomenclature will now be tested at OST to describe disciplinary profile of actors and large-scale institutional comparisons. This experiment will be the subject of a review and further improvements. |
