Once we have obtained a list of topics by using the entities recognised in a document from their Wikidata classes and superclasses, as well as the topics obtained from the topic modelling methods, we should implement a phase of combining these topics to produce a final list of topics as an output. An initial approach will be the following:
Topics that appear both in the NER and topic modeling results will have the maximum degree of confidence.
Topics that are returned by NER and not by topic modelling will be next.
Topics returned by the Topic Modeling methods and not by NER, but with mappings to an ontology will also be considered.
Finally, if a topic is returned by the Topic Modelling methods and not by NER, and it has no mapping to an ontology, it will be discarded,
Once we have obtained a list of topics by using the entities recognised in a document from their Wikidata classes and superclasses, as well as the topics obtained from the topic modelling methods, we should implement a phase of combining these topics to produce a final list of topics as an output. An initial approach will be the following: