| import json | |
| q_main_types = {} | |
| q_sub_types = {} | |
| for line in open('question_types.txt'): | |
| if line.startswith(" "): | |
| sub_type, sub_desc = line.strip().split(maxsplit=1) | |
| q_sub_types[f"{main_type}:{sub_type}"] = f"{sub_desc}" | |
| else: | |
| main_type, main_desc = line.strip().split(maxsplit=1) | |
| q_main_types[main_type] = main_desc | |
| print(q_main_types) | |
| print(q_sub_types) | |
| label2int = {} | |
| label2int_coarse = {} | |
| def prepare(in_path, out_path): | |
| with open(in_path, encoding='utf8') as fIn, open(out_path, "w") as fOut: | |
| for line in fIn: | |
| label, text = line.strip().split(maxsplit=1) | |
| label_coarse = label.split(":")[0] | |
| if label not in label2int: | |
| label2int[label] = len(label2int) | |
| if label_coarse not in label2int_coarse: | |
| label2int_coarse[label_coarse] = len(label2int_coarse) | |
| fOut.write(json.dumps({'text': text, | |
| 'label': label2int[label], | |
| 'label_text': q_sub_types[label], | |
| 'label_original': label, | |
| 'label_coarse': label2int_coarse[label_coarse], | |
| 'label_coarse_text': q_main_types[label_coarse], | |
| 'label_coarse_original': label_coarse})+"\n") | |
| prepare("train_5500.label", "train.jsonl") | |
| prepare("TREC_10.label", "test.jsonl") |