处理数据集的python脚本

def segment(src_dir):
    segment_id = 0
    lines = open(src_dir,'r').readlines()
    temp = lines[0].split(' ')[1].split('_')[1]
    wf = open('./segments_new', 'a', encoding='UTF-8-sig')

    for line in lines:
        utt = line.split(" ")[1]
        utt_id = utt.split('_')[1]
        start = line.split(" ")[2]
        end = line.split(" ")[3]
        if utt_id == temp:
            segment_id_str = "{}_{}".format(utt, str(segment_id).zfill(4))
            print(segment_id_str,utt,start,end)
            segment_id += 1
            wf.write(segment_id_str + ' ' + utt + ' ' + start + ' ' + end + '\n')
        else:
            temp = utt_id
            segment_id = 0
            segment_id_str = "{}_{}".format(utt, str(segment_id).zfill(4))
            print(segment_id_str)
            segment_id += 1
            wf.write(segment_id_str + ' ' + utt + ' ' + start + ' ' + end + '\n')


segment('./segments')

处理前：AMI中MDM格式的音频文件，用kaldi的脚本处理后得到的segment文件

处理后：改了第一列的数据，从0000开始，一个文件结束后，下个文件从0000重新开始命名文件，从左到右分别为：segment_id_str,utt,start,end