task4.py

'''
Python Training
'''
import argparse
import codecs
import xml.etree.ElementTree as ET
import pandas as pd

INPUT_XML = 'CortanaLiveData_ja-JP_2015_Audio_test_Result.xml'
INPUT_HYP = 'task3_output.hyp'
OUTPUT_HYP = 'task4_output.hyp'

def run(input_xml, input_hyp, output_hyp):
    '''
    start program
    '''
    xmlp = ET.XMLParser(encoding='utf-16')
    f = ET.parse(input_xml, parser=xmlp)
    transDict = {}
    for elem in f.iter('SpeechIn'):
        wave_file_path = elem.get('WaveFilePath').split('.')[0]
        trans = elem.getchildren()[0].getchildren()[0].getchildren()[0].get('Text')
        transDict[wave_file_path] = trans

    data = pd.read_csv(input_hyp, sep='\t', header=None, encoding='utf-16')

    for index, item in data.iterrows():
        key = '\\'.join((item[1], (item[0])))
        if key in transDict:
            item[8] = transDict[key]
            # print(key)

    with codecs.open(output_hyp, 'w', encoding='utf-16') as f:
        for item in data.values.tolist():
            f.write('\t'.join(item) + '\n')


if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='prodcess file')
    parser.add_argument('--input_xml', type=str, default=INPUT_XML, help='input xml file')
    parser.add_argument('--input_hyp', type=str, default=INPUT_HYP, help='input hyp file')
    parser.add_argument('--output_hyp', type=str, default=OUTPUT_HYP, help='output hyp file')
    args = parser.parse_args()
    run(args.input_xml, args.input_hyp, args.output_hyp)