The official help page has provided an example of German language.
https://help.libreoffice.org/6.1/he/text/shared/optionen/01010400.html
In language-dependent custom dictionaries, the field contains a known root word, as a model of affixation of the new word or its usage in compound words. For example, in a German custom dictionary, the new word “Litschi” (lychee) with the model word “Gummi” (gum) will result recognition of “Litschis” (lychees), “Litschibaum” (lychee tree), “Litschifrucht” (lychee fruit) etc.
Can someone provide an english example just to make it clear how it works?
Indeed (I am german). But you certainly know that this is to some extend a "german problem" because we have tons and tons of compound words. Everything is a compound word. Just some examples "door handle" is "Türklinke", "coffee machine" is "Kaffemaschine", "birthday cake" is "Geburtstagskuchen" and the infamous "Association for Subordinate Officials of the Main Maintenance Building of the Danube Steam Shipping Electrical Services" is "Donaudampfschiffahrtselektrizitätenhauptbetriebswerkbauunterbeamtengesellschaft". Really.
Now to your question.
I must say, after searching a bit, I think for English this is basically a useless feature. There are just so few and not very flexible real compound words. But here is an actual example. Take "sun" as new word, and specify "moon" as model word. Then the dictionary can deduce from "moonlight" that there also is a "sunlight". But I think in English this example almost exhausts the entire capabilities, while in a language like German you can generate a very high number of words, including all of their declination (which is really useful).