目录

问题描述：

问题解决：

---

问题描述：

原始的标注的三元组格式如下：

需要转换的格式如下：

tips:有一个小的难点：

1. 针对多三元组的情况，需要额外考虑

2. 最后一个样本，也记得需要处理

问题解决：

from pdb import set_trace as stop
import os
from tqdm import trange
generated_path= "/public/home/hongy/qtxu/UniCOQE_20230812/data/tuple/ele/train_new_generated.txt"
Unicoqe_path = "/public/home/hongy/qtxu/UniCOQE_20230812/data/tuple/ele/train.txt"

raw_data = []
with open(os.path.join(generated_path), 'r') as f:
    for line in f:
        raw_data.append(line)

with open(Unicoqe_path, 'w') as fw:

    line_id, i = 0, 0
    text_line, label_line = '', ''
    for line_id in trange(len(raw_data), desc= "procesing data ……"):
        cur_line = raw_data[line_id]
        if len(cur_line.split('\t')) != 2:
            label_line += '\n' + cur_line
        else:
            if text_line !='':
                sent, label = text_line.strip().split("\t")
                fw.write(sent+"####")
                label_list = label_line.strip().split('\n\n')
                span_index =[]
                all_span= ()
                for label_i in label_list:
                    cur_span = label_i.strip()[1:-1].split(';')
                    sub, obj,asp = cur_span[0], cur_span[1], cur_span[2]
                    try:
                        sub_index = [int(index) for index, word in (pair.split('&', 1) if '&' in pair else [pair, ''] for pair in sub.strip()[1:-1].split())] # 针对14&&这样的特例，进行处理
                        obj_index = [int(index) for index, word in (pair.split('&', 1) if '&' in pair else [pair, ''] for pair in obj.strip()[1:-1].split())]
                        asp_index = [int(index) for index, word in (pair.split('&')for pair in asp.strip()[1:-1].split())] 
                    except:
                        print(text_line)
                        stop()

                    span_tuple = (sub_index, obj_index, asp_index)
                    span_index.append(span_tuple)
                
                fw.write(str('['+', '.join(str(span) for span in span_index)) + "]\n")


            text_line = cur_line
            label_line=''

    fw.write(text_line.strip().split("\t")[0]+"####")
    label_list = label_line.strip().split('\n\n')
    span_index =[]
    all_span= ()
    for label_i in label_list:
        cur_span = label_i.strip()[1:-1].split(';')
        sub, obj,asp = cur_span[0], cur_span[1], cur_span[2]
        sub_index = [int(index) for index, word in (pair.split('&', 1) if '&' in pair else [pair, ''] for pair in sub.strip()[1:-1].split())] # 针对14&&这样的特例，进行处理
        obj_index = [int(index) for index, word in (pair.split('&', 1) if '&' in pair else [pair, ''] for pair in obj.strip()[1:-1].split())]
        asp_index = [int(index) for index, word in (pair.split('&')for pair in asp.strip()[1:-1].split())] 

        span_tuple = (sub_index, obj_index, asp_index)
        span_index.append(span_tuple)
    
    fw.write(str('['+', '.join(str(span) for span in span_index)) + "]\n")