- evaluate_present.py和evaluate_absent.py分别计算序列标注present keyphrase和文本生成absent keyphrase的结果
- 提供了merge_prob.py文件，用于测试kp20k时合并.prob
- present评价代码添加了stem开关和读取概率文件计算f1@5(需要在运行decode_seq2seq.py推断时打开output_prob开关）,参数说明如下：
usage: evaluate_present.py [-h] [--stem_evaluate] [--stem_divide] [--label_path LABEL_PATH] [--input_path INPUT_PATH] [--gold_path GOLD_PATH] [--prob_path PROB_PATH] optional arguments: -h, --help show this help message and exit --stem_evaluate whether use stem before evaluation, default: False --stem_divide whether use stem before dividing keyphrases into present and absent parts, default: False --label_path LABEL_PATH the path to the predict .label file --input_path INPUT_PATH the path to the input .seq.in file --gold_path GOLD_PATH the path to the ground truth file, if stem_divide, it is .present.test file; else, it is .keyphrase file --prob_path PROB_PATH the path to the .prob file which records the predicted label probs for calculating present f1@5 - absent评价代码添加了stem开关，参数说明如下：
usage: evaluate_absent.py [-h] [--stem_evaluate] [--stem_divide] [--input_path INPUT_PATH] [--pred_path PRED_PATH] [--gold_path GOLD_PATH] optional arguments: -h, --help show this help message and exit --stem_evaluate whether use stem before evaluation, default: False --stem_divide whether use stem before dividing keyphrases into present and absent parts, default: False --input_path INPUT_PATH the path to the input .seq.in file --pred_path PRED_PATH the path to the predict model.absent file --gold_path GOLD_PATH the path to the ground truth file, if stem_divide, it is .absent file; else, it is .keyphrase file
- evaluate_diff.py：测量词袋误差
- evaluate_present_bieou：present部分采用BIEOU标注方案的评测脚本
- evaluate_all.py：不区分present/absent的评测及其启动脚本