[update]add ocnli

README.md

@@ -21,7 +21,9 @@ size_categories:
 | BUSTM | 汉语 | [BUSTM](https://tianchi.aliyun.com/competition/entrance/531851/information); [BUSTM](https://github.com/xiaobu-coai/BUSTM) | 总样本数为：177173，其中，匹配样本个数为：54805，不匹配样本个数为：122368 | 小布助手对话短文本语义匹配比赛数据集 | [BUSTM](https://github.com/CLUEbenchmark/FewCLUE/tree/main/datasets/bustm) |
 | CHIP2019 | 汉语 | [CHIP2019](https://www.biendata.xyz/competition/chip2019/) | 2万 | 平安医疗科技疾病问答迁移学习比赛数据集 |  |
 | COVID-19 | 汉语 | [COVID-19](https://tianchi.aliyun.com/competition/entrance/231776/information) |  | 天池新冠疫情相似句对判定大赛 | [COVID-19](https://gitee.com/liangzongchang/COVID-19-sentence-pair/) |
-| Chinese-MNLI | 汉语 | [Chinese-MNLI](https://github.com/pluto-junzeng/CNSD) | TRAIN: 390K, VALID: 12K, TEST: 13K | 通过翻译加部分人工修正的方法，从英文原数据集生成（原数据是：蕴含，中性，冲突，的句子推理数据集，已转换为句子对。） |  |
+| Chinese-MNLI | 汉语 | [Chinese-MNLI](https://github.com/pluto-junzeng/CNSD) | TRAIN: 390K, VALID: 12K, TEST: 13K | 通过翻译加部分人工修正的方法，从英文原数据集生成（原数据是：蕴含，中性，冲突，的句子推理数据集，已转换为句子对）。 |  |
+| Chinese-SNLI | 汉语 | [Chinese-SNLI](https://github.com/pluto-junzeng/CNSD) | TRAIN: 550K, VALID: 10K, TEST: 10K | 通过翻译加部分人工修正的方法，从英文原数据集生成（原数据是：蕴含，中性，冲突，的句子推理数据集，已转换为句子对）。 |  |
+| OCNLI | 汉语 | [OCNLI](https://github.com/CLUEbenchmark/OCNLI) | TRAIN: 50K, VALID: 3K, TEST: 3K | 原生中文自然语言推理数据集，是第一个非翻译的、使用原生汉语的大型中文自然语言推理数据集。 |  |
 
 
 <details>

data/chinese_mnli.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:994e50bf6964ca0d6073b31fae10af95c2037ebbba434a864591b79c64757045
-size 117184478
+oid sha256:f0b87b6ebc3b8124ca4f1cbb0e5055e4941c0f8f93014248f93d82cbf617a585
+size 117078623

data/chinese_snli.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0616f76d77f83a4bbc4ff0b0cbea1a018dbdcf8c7dc94538a062284d9626704f
+size 125893606

data/ocnli.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12b5f3a0e082d09f29d646024f73c870906c29adc81cdbd9c6c519fe276c3f33
+size 12092258

examples/preprocess/process_chinese_mnli.py

@@ -45,16 +45,22 @@ def main():
                     sentence2 = row["sentence2"]
                     gold_label = row["gold_label"]
 
+                    if gold_label not in ("entailment", "neutral", "contradiction", "-"):
+                        print(gold_label)
+                        raise AssertionError
+
+                    if gold_label in ("-",):
+                        continue
+
+                    label = "1" if gold_label in ("entailment",) else "0"
+
                     if name == "cnsd_multil_train.jsonl":
-                        label = "1" if gold_label in ("entailment",) else "0"
                         flag = "train"
                         category = "cnsd_multil_train"
                     elif name == "cnsd_multil_dev_matched.jsonl":
-                        label = "1" if gold_label in ("entailment",) else "0"
                         flag = "validation"
                         category = "cnsd_multil_dev_matched"
                     elif name == "cnsd_multil_dev_mismatched.jsonl":
-                        label = "1" if gold_label in ("entailment",) else "0"
                         flag = "validation"
                         category = "cnsd_multil_dev_mismatched"
                     else:

examples/preprocess/process_chinese_snli.py

@@ -0,0 +1,86 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import argparse
+import json
+import os
+from pathlib import Path
+import sys
+
+pwd = os.path.abspath(os.path.dirname(__file__))
+sys.path.append(os.path.join(pwd, '../../'))
+
+from datasets import load_dataset
+from tqdm import tqdm
+
+from project_settings import project_path
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument("--data_dir", default="./data/chinese_snli", type=str)
+
+    parser.add_argument(
+        "--output_file",
+        default=(project_path / "data/chinese_snli.jsonl"),
+        type=str
+    )
+
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = get_args()
+
+    data_dir = Path(args.data_dir)
+
+    with open(args.output_file, "w", encoding="utf-8") as fout:
+        for name in ["cnsd_snli_v1.0.train.jsonl", "cnsd_snli_v1.0.dev.jsonl", "cnsd_snli_v1.0.test.jsonl"]:
+            filename = data_dir / name
+            with open(filename, "r", encoding="utf-8") as fin:
+                for row in fin:
+                    row = json.loads(row)
+                    sentence1 = row["sentence1"]
+                    sentence2 = row["sentence2"]
+                    gold_label = row["gold_label"]
+
+                    if gold_label not in ("entailment", "neutral", "contradiction", "-"):
+                        print(gold_label)
+                        raise AssertionError
+
+                    if gold_label in ("-",):
+                        continue
+
+                    label = "1" if gold_label in ("entailment",) else "0"
+                    category = gold_label
+
+                    if name == "cnsd_snli_v1.0.train.jsonl":
+                        flag = "train"
+                    elif name == "cnsd_snli_v1.0.dev.jsonl":
+                        flag = "validation"
+                    elif name == "cnsd_snli_v1.0.test.jsonl":
+                        flag = "test"
+                    else:
+                        raise AssertionError
+
+                    if label not in ("0", "1", None):
+                        raise AssertionError
+
+                    row = {
+                        "sentence1": sentence1,
+                        "sentence2": sentence2,
+                        "label": label,
+                        "category": category,
+                        "data_source": "Chinese-SNLI",
+                        "split": flag
+                    }
+
+                    row = json.dumps(row, ensure_ascii=False)
+                    fout.write("{}\n".format(row))
+
+    return
+
+
+if __name__ == '__main__':
+    main()

examples/preprocess/process_ocnli.py

@@ -0,0 +1,84 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import argparse
+import json
+import os
+from pathlib import Path
+import sys
+
+pwd = os.path.abspath(os.path.dirname(__file__))
+sys.path.append(os.path.join(pwd, '../../'))
+
+from datasets import load_dataset
+from tqdm import tqdm
+
+from project_settings import project_path
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument("--data_dir", default="./data/ocnli/data/ocnli", type=str)
+
+    parser.add_argument(
+        "--output_file",
+        default=(project_path / "data/ocnli.jsonl"),
+        type=str
+    )
+
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = get_args()
+
+    data_dir = Path(args.data_dir)
+
+    with open(args.output_file, "w", encoding="utf-8") as fout:
+        for name in ["train.50k.json", "dev.json", "test.json"]:
+            filename = data_dir / name
+            with open(filename, "r", encoding="utf-8") as fin:
+                for row in fin:
+                    row = json.loads(row)
+                    sentence1 = row["sentence1"]
+                    sentence2 = row["sentence2"]
+                    label = row.get("label", None)
+                    category = row.get("genre", None)
+
+                    if label not in ("entailment", "neutral", "contradiction", "-", None):
+                        print(label)
+                        raise AssertionError
+
+                    if label in ("-",):
+                        continue
+
+                    if label is not None:
+                        label = "1" if label in ("entailment",) else "0"
+
+                    if name == "train.50k.json":
+                        flag = "train"
+                    elif name == "dev.json":
+                        flag = "validation"
+                    elif name == "test.json":
+                        flag = "test"
+                    else:
+                        raise AssertionError
+
+                    row = {
+                        "sentence1": sentence1,
+                        "sentence2": sentence2,
+                        "label": label,
+                        "category": category,
+                        "data_source": "OCNLI",
+                        "split": flag
+                    }
+
+                    row = json.dumps(row, ensure_ascii=False)
+                    fout.write("{}\n".format(row))
+
+    return
+
+
+if __name__ == '__main__':
+    main()

sentence_pair.py

@@ -13,11 +13,13 @@ _URLS = {
     "bustm": "data/bustm.jsonl",
     "ccks2018_task3": "data/ccks2018_task3.jsonl",
     "chinese_mnli": "data/chinese_mnli.jsonl",
+    "chinese_snli": "data/chinese_snli.jsonl",
     "chinese_sts": "data/chinese_sts.jsonl",
     "chip2019": "data/chip2019.jsonl",
     "covid_19": "data/covid_19.jsonl",
     "diac2019": "data/diac2019.jsonl",
     "lcqmc": "data/lcqmc.jsonl",
+    "ocnli": "data/ocnli.jsonl",
 
 }