The entry point will be what we call a “Java rule”. Here you have a simple demo rule.
In a Java rule you can loop over the tokens of a sentence. Each token has a word form, and one or more lemmas and POS tags. If the token has been properly disambiguated, there is only one lemma and one POS tag, but that is not always possible or desirable. I guess you plan to use only word forms. But the other parameters could be useful. I will write more about this.
As a result, the Java rule returns an array of matches (possible errors). See here.
You will probably need an abstract Java rule plus another rule for each language (English, German, etc.) with the language data. See some abstract rules here and some English implementations here.