[update]add data

README.md

@@ -9,15 +9,18 @@ size_categories:
 ---
 ## 中文闲聊数据集
 
+role 的取值有: "unknown", "human", "assistant", 三种.
+
 
 数据集从网上收集整理如下:
 
 | 数据 | 原始数据/项目地址 | 样本个数 | 语料描述 | 替代数据下载地址 |
 | :--- |     :---:      |   :---:  |  :---:    |     :---:     |
 | ChatterBot | [ChatterBot](https://github.com/gunthercox/ChatterBot); [chatterbot-corpus](https://github.com/gunthercox/chatterbot-corpus) | 560 | 按类型分类，质量较高 | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
-| douban | [Douban Conversation Corpus](https://github.com/MarkWuNLP/MultiTurnResponseSelection) | 352W | 来自北航和微软的paper, 开源项目, 噪音相对较少，原本是多轮 (平均7.6轮) | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
+| douban | [Douban Conversation Corpus](https://github.com/MarkWuNLP/MultiTurnResponseSelection) | 352W | 来自北航和微软的paper, 噪音相对较少, 多轮(平均7.6轮) | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
 | ptt | [PTT中文語料](https://github.com/zake7749/Gossiping-Chinese-Corpus) | 77W | 开源项目, 台湾PTT论坛八卦版, 繁体, 语料较生活化, 有噪音 | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
-| subtitle | [电视剧对白语料](https://github.com/aceimnorstuvwxz/dgk_lost_conv) | 274W | 开源项目, 来自爬取的电影和美剧的字幕, 有一些噪音, 对白不一定是严谨的对话, 原本是多轮 (平均5.3轮) | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
+| qingyun | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao | 10W | 青云语料, 相对不错, 生活化 |  |
+| subtitle | [电视剧对白语料](https://github.com/aceimnorstuvwxz/dgk_lost_conv) | 274W | 来自爬取的电影和美剧的字幕, 有一些噪音, 不严谨的对话, 说话人无法对应起来, 多轮(平均5.3轮) | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
 | tieba | [贴吧论坛回帖语料](https://pan.baidu.com/s/1mUknfwy1nhSM7XzH8xi7gQ); 密码:i4si | 232W | 多轮, 有噪音 | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
 | weibo | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao | 443W | 来自华为的paper |  |
 | xiaohuangji | [小黄鸡语料](https://github.com/candlewill/Dialog_Corpus) | 45W | 原人人网项目语料, 有一些不雅对话, 少量噪音 | [阿里云盘](https://www.aliyundrive.com/s/qXBdAYtz5j5); 提取码: 81ao |
@@ -27,11 +30,8 @@ size_categories:
 <summary>参考的数据来源,展开查看</summary>
 <pre>
 <code>
-
 https://github.com/codemayq/chinese_chatbot_corpus
-
 https://github.com/yangjianxin1/GPT2-chitchat
-
 </code>
 </pre>
 </details>

chinese_chitchat.py

@@ -12,6 +12,7 @@ _URLS = {
     "chatterbot": "data/chatterbot.jsonl",
     "douban": "data/douban.jsonl",
     "ptt": "data/ptt.jsonl",
+    "qingyun": "data/qingyun.jsonl",
     "subtitle": "data/subtitle.jsonl",
     "tieba": "data/tieba.jsonl",
     "weibo": "data/weibo.jsonl",

data/qingyun.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e0b4d1f69df4085d6d54c7449174ab62a22167dd31e301704e4d5883e5f3d3b
+size 22507854

examples/process/process_qingyun.py

@@ -0,0 +1,67 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import argparse
+import json
+from pathlib import Path
+
+import pandas as pd
+from tqdm import tqdm
+
+from project_settings import project_path
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--data_file",
+        default=(project_path / "original_data/qingyun-11w/12万对话语料青云库.csv").as_posix(),
+        type=str
+    )
+    parser.add_argument(
+        "--output_file",
+        default=(project_path / "data/qingyun.jsonl"),
+        type=str
+    )
+
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = get_args()
+
+    with open(args.data_file, "r", encoding="utf-8") as fin, \
+         open(args.output_file, "w", encoding="utf-8") as fout:
+        for row in fin:
+            row = str(row).strip()
+            splits = row.split(" | ")
+            if len(splits) != 2:
+                print(row)
+                raise AssertionError
+
+            question = splits[0].strip()
+            answer = splits[1].strip()
+
+            row = {
+                "conversation": [
+                    {
+                        "role": "human",
+                        "message": question,
+                    },
+                    {
+                        "role": "assistant",
+                        "message": answer,
+                    },
+                ],
+                "category": None,
+                "data_source": "qingyun",
+            }
+            row = json.dumps(row, ensure_ascii=False)
+            fout.write("{}\n".format(row))
+
+    return
+
+
+if __name__ == '__main__':
+    main()