Relevance Search uses the Dynamics 365 instance's installation language for linguistic analysis / noise filtering / stop words. In a multi-language-environment or in case the default system language is different from the default user/KB-article language, this can lead to very poor results when searching for complete sentences or natural language constructs: Relevance search will find way too many completely irrelevant KB-articles, because noise (the , this, and, or,...) that is present in EVERY KB-article is not filtered out.
When using the integrated KB-search in a case - form, this is especially harmful, as it may result in a significant performance- and usability penalty.
Instead, Relevance Search should apply linguistic analysis algorithms that match a KB-article's individual configured language on creating the search index.
Under Review