structure_assignment/scripts/pipeline1.py

71 lines
2.7 KiB
Python

import argparse
import os
import shutil
import codecs
import classla
from classla import Document
from classla.models.common.conll import CoNLLFile
from constants import *
arg_parser = argparse.ArgumentParser(description='Parse Slovene strings and convert to TEI.')
arg_parser.add_argument('-inlist', type=str, help='Input list file')
arg_parser.add_argument('-outtei', type=str, help='Output TEI file')
arguments = arg_parser.parse_args()
input_file_name = arguments.inlist
output_file_name = arguments.outtei
NLP_CONFIG_MAP = {
'treebank': 'sl_ssj_jos',
'processors': 'tokenize,pos,lemma,depparse',
'tokenize_pretokenized': True,
'models_dir': CLASSLA_MODELS_DIRECTORY
}
XML_ID_PREFIX = 's'
def run_pipeline(input_file_name, output_file_name):
shutil.rmtree(TMP_DIRECTORY, True)
os.makedirs(TMP_DIRECTORY, exist_ok=True)
shutil.copyfile(input_file_name, STRING_LIST_FILE_NAME)
run_obeliks(STRING_LIST_FILE_NAME, OBELIKS_RAW_FILE_NAME)
tweak_conllu(OBELIKS_RAW_FILE_NAME, OBELIKS_TWEAKED_FILE_NAME)
run_classla(OBELIKS_TWEAKED_FILE_NAME, CLASSLA_OUTPUT_FILE_NAME)
run_jos_translation(CLASSLA_OUTPUT_FILE_NAME, CLASSLA_TRANSLATED_FILE_NAME)
run_tei_conversion(CLASSLA_TRANSLATED_FILE_NAME, TEI_INIT_FILE_NAME)
shutil.copyfile(TEI_INIT_FILE_NAME, output_file_name)
def run_obeliks(list_file_name, conllu_file_name):
print('Running obeliks ...')
obeliks_command = ' '.join(['obeliks', '-c', '-if', list_file_name, '-o', conllu_file_name])
os.system(obeliks_command)
def tweak_conllu(input_file_name, output_file_name):
print('Tweaking conllu results ...')
tweak_command = ' '.join(['python', CONLLU_TWEAK_SCRIPT_NAME, '-infile', input_file_name, '-outfile', output_file_name])
os.system(tweak_command)
def run_classla(obeliks_file_name, classla_file_name):
print('Running classla ...')
doc = Document(text=None)
conll_file = CoNLLFile(filename=obeliks_file_name)
doc.conll_file = conll_file
nlp = classla.Pipeline('sl', **NLP_CONFIG_MAP)
result = nlp(doc)
result.conll_file.write_conll(classla_file_name)
def run_jos_translation(input_file_name, output_file_name):
print('Translating JOS ...')
translate_command = ' '.join(['python', TRANSLATION_SCRIPT_NAME, '-infile', input_file_name, '-dict', TRANSLATION_FILE_NAME, '-outfile', output_file_name])
print(translate_command)
os.system(translate_command)
def run_tei_conversion(classla_file_name, tei_file_name):
print('Converting to tei ...')
convert_command = ' '.join(['python', CONLLU_TEI_SCRIPT_NAME, '-o', tei_file_name, classla_file_name])
print(convert_command)
os.system(convert_command)
run_pipeline(input_file_name, output_file_name)