structure_assignment/scripts/pipeline.py

import sys
import os
import shutil
import codecs
import re
import lxml.etree as lxml

import classla
from classla import Document
from classla.models.common.conll import CoNLLFile

input_file_name = sys.argv[1]
output_file_name = sys.argv[2]

TMP_DIRECTORY = '../tmp/structure_assignment'

CONLLU_TEI_SCRIPT_NAME = 'conllu_to_xml.py'
MWE_EXTRACTION_SCRIPT_NAME = 'wani.py'
STRUCTURE_ASSIGNMENT_SCRIPT_NAME = 'assign_structures.py'
STRUCTURE_CREATION_SCRIPT_NAME = 'create_structures.py'
STRUCTURE_CREATION_SCRIPT_NAME = 'create_structures.py'
TEI_DICTIONARY_SCRIPT_NAME = 'tei_to_dictionary.py'

OBELIKS_JAR_FILE_NAME = '../resources/obeliks.jar'
TRANSLATION_FILE_NAME = '../resources/dict.xml'
CLASSLA_MODELS_DIRECTORY = '../resources/classla'
STRUCTURE_CURRENT_FILE_NAME = '../resources/structures.xml'
STRUCTURE_SCHEMA_FILE_NAME = '../resources/structures.xsd'
DICTIONARY_SCHEMA_FILE_NAME = '../resources/monolingual_dictionaries.xsd'

STRING_LIST_FILE_NAME = TMP_DIRECTORY + '/strings.txt'
OBELIKS_RAW_FILE_NAME = TMP_DIRECTORY + '/obeliks_raw.conllu'
OBELIKS_TWEAKED_FILE_NAME = TMP_DIRECTORY + '/obeliks_tweaked.conllu'
CLASSLA_FILE_NAME = TMP_DIRECTORY + '/classla.conllu'
TEI_INIT_FILE_NAME = TMP_DIRECTORY + '/tei_initial.xml'
TEI_STRUCTURE_1_FILE_NAME = TMP_DIRECTORY + '/tei_with_structure_ids1.xml'
TEI_STRUCTURE_2_FILE_NAME = TMP_DIRECTORY + '/tei_with_structure_ids2.xml'
MWE_CSV_1_FILE_NAME = TMP_DIRECTORY + '/mwes1.csv'
MWE_CSV_2_FILE_NAME = TMP_DIRECTORY + '/mwes2.csv'
STRUCTURE_NEW_FILE_NAME = TMP_DIRECTORY + '/structures_new.xml'
DICTIONARY_XML_FILE_NAME = TMP_DIRECTORY + '/dictionary.xml'

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_obeliks4J(STRING_LIST_FILE_NAME, OBELIKS_RAW_FILE_NAME)
    fix_xml_ids(OBELIKS_RAW_FILE_NAME, OBELIKS_TWEAKED_FILE_NAME)
    run_classla(OBELIKS_TWEAKED_FILE_NAME, CLASSLA_FILE_NAME)
    run_tei_conversion(CLASSLA_FILE_NAME, TEI_INIT_FILE_NAME)
    run_mwe_extraction(STRUCTURE_CURRENT_FILE_NAME, TEI_INIT_FILE_NAME, MWE_CSV_1_FILE_NAME)
    run_structure_assignment(STRING_LIST_FILE_NAME, TEI_INIT_FILE_NAME, MWE_CSV_1_FILE_NAME, TEI_STRUCTURE_1_FILE_NAME)
    run_structure_creation(STRUCTURE_CURRENT_FILE_NAME, TEI_STRUCTURE_1_FILE_NAME, STRUCTURE_NEW_FILE_NAME)
    validate_structures(STRUCTURE_NEW_FILE_NAME)
    run_mwe_extraction(STRUCTURE_NEW_FILE_NAME, TEI_INIT_FILE_NAME, MWE_CSV_2_FILE_NAME)
    run_structure_assignment(STRING_LIST_FILE_NAME, TEI_INIT_FILE_NAME, MWE_CSV_2_FILE_NAME, TEI_STRUCTURE_2_FILE_NAME)
    run_dictionary_conversion(TEI_STRUCTURE_2_FILE_NAME, DICTIONARY_XML_FILE_NAME)
    validate_dictionary(DICTIONARY_XML_FILE_NAME)
    shutil.copyfile(DICTIONARY_XML_FILE_NAME, output_file_name)

def run_obeliks4J(obeliks_file_name, classla_file_name):
    print('Running obeliks ...')
    obeliks_command = 'java -jar ' + OBELIKS_JAR_FILE_NAME + ' -d -if ' + STRING_LIST_FILE_NAME + ' -o ' + OBELIKS_RAW_FILE_NAME
    os.system(obeliks_command)

def fix_xml_ids(input_file_name, output_file_name):
    print('Fixing xml ids ...')
    output_file = codecs.open(output_file_name, 'w')
    input_file = codecs.open(input_file_name, 'r')
    regexp = r'^(# sent_id = )(\d+\.\d+)$'
    for line in input_file:
        match = re.search(regexp, line)
        if (match):
            line = match.group(1) + XML_ID_PREFIX + match.group(2) + '\n'
        output_file.write(line)
    input_file.close()
    output_file.close()

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_tei_conversion(classla_file_name, tei_file_name):
    print('Converting to tei ...')
    convert_command = ' '.join(['python', CONLLU_TEI_SCRIPT_NAME, classla_file_name, tei_file_name, '--translate', TRANSLATION_FILE_NAME])
    os.system(convert_command)

def run_mwe_extraction(structure_file_name, tei_file_name, mwe_csv_file_name):
    print('Extracting MWEs from tei ...')
    extraction_command = ' '.join(['python', MWE_EXTRACTION_SCRIPT_NAME, structure_file_name, tei_file_name, '--all', mwe_csv_file_name, '--skip-id-check', '--fixed-restriction-order'])
    print(extraction_command)
    os.system(extraction_command)

def validate_structures(structure_file_name):
    print('Validating updated structure specifications ...')
    xml_schema = lxml.XMLSchema(lxml.parse(STRUCTURE_SCHEMA_FILE_NAME))
    xml_tree = lxml.parse(structure_file_name)
    xml_schema.assertValid(xml_tree)

def run_structure_assignment(input_file_name, tei_file_name, mwe_csv_file_name, output_file_name):
    print('Assigning structure ids ...')
    assignment_command = ' '.join(['python', STRUCTURE_ASSIGNMENT_SCRIPT_NAME, input_file_name, tei_file_name, mwe_csv_file_name, output_file_name])
    os.system(assignment_command)

def run_structure_creation(input_file_name, tei_file_name, output_file_name):
    print('Creating missing structures ...')
    creation_command = ' '.join(['python', STRUCTURE_CREATION_SCRIPT_NAME, '-infile', input_file_name, '-tei', tei_file_name, '-outfile', output_file_name])
    os.system(creation_command)
    
def run_dictionary_conversion(tei_file_name, xml_file_name):
    print('Converting to dictionary xml format ...')
    convert_command = ' '.join(['python', TEI_DICTIONARY_SCRIPT_NAME, '-infile', tei_file_name, '-outfile', xml_file_name])
    os.system(convert_command)

def validate_dictionary(dictionary_file_name):
    print('Validating output dictionary file ...')
    xml_schema = lxml.XMLSchema(lxml.parse(DICTIONARY_SCHEMA_FILE_NAME))
    xml_tree = lxml.parse(dictionary_file_name)
    xml_schema.assertValid(xml_tree)

run_pipeline(input_file_name, output_file_name)