Question

• The text seems to match "elimination" - there is "eliminated", is the problem in the form of the word?
• If you enter an incomplete word, for example without the ending, the selection will be empty.
• At the same time, with one incomplete word ("backup"), it was possible to get a selection.
• The search is strictly tied to declension.

Answer

For search in the KCS widget, we use a 3rd-party full-text PostgreSQL engine. One of its features is that words are reduced to their base form (i.e., lemmatization is performed): "homes" - "home"

Lemmatization occurs according to the connected language dictionaries.

In this regard, search by word parts is not supported. This explains why "backup" yields 0 results, as well as why "Bin Setting" is not found (bin and backup are incomplete words that are not sought in dictionaries).

The same situation will occur when searching for, for example, some commands that are written incompletely. In such cases, it is better to use quotes, then the article is found:

As for the different number of search results, there are several possible reasons:

Lemmatization works differently, i.e., the form of the word can be interpreted differently, for example, singular and plural (backupUYu and backupYVe).

Absence of a word in the dictionary, i.e., PostgreSQL does not recognize all forms.

In the second case, when a word is not found in the dictionary, it searches for occurrences as is, that is, only articles with the ending "backupUYu".

Judging by the tests, in this case, this is precisely the reason, the word is absent from the dictionary: search results for "backup", "backup", "backups" and "backupUYu" - they are all different. The search uses a method that does not work with logical operators and prefix search.

In general, the engine offers various search methods; the one used in Swarmica is similar to regular user search, which correlates with the KCS principle of describing the problem in the client's words.

Alternative

The Premium plan also offers neurosearch using AI. This search can already take into account large phrases and chunks of text, substitute synonyms, declensions, various endings, summarize text, etc.