What (more recent) model or translation strategies do you recommend for investigating and analyzing the translation of lexical collocations used in a literary novel?
Collocations are best learnt and analysed by consulting corpora. In your case, parallel corpora would be mostly helpful. However, I don't whether there are any for the language pair you are investigating. Alternatively, you can create parallel corpora on your own (i.e., from scratch) or comparable corpora. To consult parallel corpora, you may consider AntpConc, developed by Prof. Lawrence: https://www.laurenceanthony.net/software/antpconc/