Text this: Evaluation of selected language models for the generation of patient assessment summaries from uncorrected transcriptions