tree 9034841d0aba3b8bd878e6812f5060fc8f71c0d6
parent e406514c455b08c3aad5b4961eb93c324b2dbaa4
author Marc Kupietz <kupietz@ids-mannheim.de> 1788623445 +0200
committer Marc Kupietz <kupietz@ids-mannheim.de> 1788630109 +0200

Readme: show how to analyse the collocates of a lemma

The passage introducing collocationAnalysis() showed only how to pass a
word form, and left the plain lemma case to be inferred from the light
verb construction example, where the lemma layer appears inside a focus
query. lemmatizeNodeQuery was not mentioned at all, so that a model could
only arrive at it from prior knowledge of the package rather than from
the documentation - although the doc prompting tests accept it as a
correct answer.

Asked for the collocates of a lemma with this passage as it was, a model
reached for the word form in two of three runs; with the two ways spelled
out, in none of three.

The word form example was "setzts", which is a spoken contraction rather
than a form one would look for in a corpus, and reads oddly next to the
lemma it is meant to contrast with. It is now "setzte".

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Change-Id: I4a40b33d6c7d77538f54529be1eee331fec97e05