Created
April 30, 2020 09:18
-
-
Save polm/f444f8f506639d3a64840b41202af73b to your computer and use it in GitHub Desktop.
GSD CoNLL spaCy prep script
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #!/usr/bin/env python3 | |
| """ | |
| Convert GSD conll format to a format the spaCy convert script can use as-is. | |
| There are two main changes: | |
| 1. POS tags format is changed slightly. | |
| old: 名詞-普通名詞-一般 | |
| hyphen separated, variable fields | |
| new: 名詞,普通名詞,一般,* | |
| comma separated, always four fields, padded with asterisks | |
| 2. The tenth field has a list of | separated attributes. BILOU NER tags are | |
| included, but O tags are omitted. spaCy looks for just NER data in the 10th | |
| field, so this script discards the other attributes and adds O tags. | |
| """ | |
| import fileinput | |
| def pos_convert(oldpos): | |
| """Convert pos format. | |
| """ | |
| parts = oldpos.split('-') | |
| newfields = (parts + list('***'))[:4] | |
| return ','.join(newfields) | |
| def ner_convert(nerfield): | |
| exfields = nerfield.split('|') | |
| ner = [xx for xx in exfields if xx[:3] == 'NE='] | |
| if len(ner) == 1: | |
| return ner[0][3:] # discard the prefix | |
| elif not ner: | |
| return 'O' | |
| else: | |
| raise RuntimeError("Multiple NER annotations?") | |
| for line in fileinput.input(): | |
| line = line.strip() | |
| # full sentences and blank lines are unchanged | |
| if (not line) or line[0] == '#': | |
| print(line) | |
| continue | |
| fields = line.split('\t') | |
| fields[4] = pos_convert(fields[4]) | |
| fields[9] = ner_convert(fields[9]) | |
| print(*fields, sep='\t') |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment