elliottlash12 / UD_Old_Irish-CritMinorGlossesMilan

A Universal Dependencies Treebank for the Old Irish "Minor Glosses"
0 stars 0 forks source link

Protocol for generating a new batches #15

Closed elliottlash12 closed 3 years ago

elliottlash12 commented 3 years ago

The current procedure for generating a new batch is the following: (1) run the conllu_maker script on a csv file [it is not simply possible to insert concats + punctuation at this stage] (2) run part of edit_conllu on the output of conllu_maker to generate a new conllu with inserted concats. DO NOT run fill_deps_in at this point. (3) Push the new batch to git. (4) After the new batch has head:deprels, run fill_deps_in.

The above protocol results in some errors see issue #14 .

A better protocol would be the following: Incorporate relevant portions of edit_conllu into conllu_maker, e.g. check_concats, getpunct, and insert_chunks. testing.py is part of this process.