By using more than 3000 labelled text parts of patents from the 3GPP telecommunication sector, the dwh GmbH designed intelligent classifiers that are able to assign hundreds of thousands of telecommunication patents to their respective classes.
Various bag of words models transform the strings to a suitable input for the classifiers, which can assign the patents to either one or multiple classes according to the requirements. Additionally, keyword lists were used to achieve a more finely granulated class distinction.