Convert the linearization script to take TSV as input and handle the structure itself, making it simpler to call

Simplify compute-profile again to simply read from stdin and drop all the optparse-applicative + ReadMonad combo

Refine the notion of size in the Tree representation of Conllu annotations as used in measures.hs from 922f5f43

Describe how to represent a UD token as a tabular data structure to allow outputing matched words in a TSV

Improve scripts for profile computation: searching from (serialized) indexed trees, and using ranges to compact occurrences found

Add types to represent indexed sentences and documents, effectively absorbing part of the work that was done during the search

Add a script to serialize syntax trees computed from the .conllu files, copying part of the Conllu.Type types in the process to get a Generic instance for Serialization and to slightly improve structure by indexing features by name with a Map

Lift some constraint on the monad in  and use that to pass the config around as a Reader in the new profile extractor script

A tiny script used to add the manually annotated label to a bunch of texts being prepared in a JSONL (to be used as input for prodigy)

Add predictor for MultiBERT model, improve the trainer for the same model and expose the predicted score in the model's output

Separate train script into one to train directly one classifier and one to train one binary classifier (accept/reject) for each discursive function