Improving term networks through the detection of semantic perspectives
Improving term networks through the detection of semantic perspectives
Samenvatting
We present an industry use case in which we discuss the practical use of a formal list of definitions. The list of definitions will
be used as a basis for a knowledge graph that will serve knowledge panels when interacting with information in the company
and that will help improve question answering using large language models. We evaluate whether the list is complete, whether
the terms are relevant and whether there might be ambiguity in these terms. For this purpose we have analyzed a large body
of company documents that represent how employees use these terms in practice, as well as a body of formal documents that
represent how these terms are used in the industry in general. This has lead to the finding that some of the terms may not
be ambiguous, but can be interpreted differently. There is a constant balance in a term list between completeness, level of
abstraction and relevance. Determining which terms may lead to confusion because of multiple interpretations is a relevant
step forward in creating usable knowledge graphs.
Organisatie | HAN University of Applied Sciences |
Afdeling | Academie IT en Mediadesign |
Lectoraten | |
Lectoraat | Data & Knowledge Engineering |
Partners | Alliander |
Datum | 2024-09-17 |
Type | Conferentiebijdrage |
Taal | Onbekend |