Weibo User Depression Detection Dataset

微博用户抑郁症检测数据集

Introduction

• The Weibo User Depression Detection Dataset (WU3D) contains normal user samples and depressed user samples collected by a crawler using the crawling API provided by the Weibo official.
• The dataset is stored by JSON files, with depressed.json for the depressed users and normal.json for the normal users. It can be download from Google Drive or Baidu Netdisk (百度云盘，提取码yar1).
• The detailed information fields of a user include:

Field Name	Description	Example
nickname	User nickname, a short text.	小陈不要再抑郁了
gender	User gender	男
profile	A sentence of the user’s self-description, self-introduction, short text.	这世界那么多人又不缺我一个活着
birthday	The birthday of the user, stored as a string in the format of "year-month-day". If the user does not fill in the year of birth, the "month-day" format is used. If the user does not fill in the information, the default content is "无" ("None" in English).	1991-07-25
label	User label, 1 for depressed user, 0 for normal user.	1
num_of_follower	The number of user followers.	3
num_of_following	The number of user followings.	12
all_tweet_count	The total number of all the tweets.	137
original_tweet_count	The total number of original tweets.	128
repost_tweet_count	The total number of repost tweets.	9
tweets	Tweet field, a list of user tweets.	-
tweet_content	Tweet text content.	抗抑郁的第72天 这次抑郁发作的起因是因为感情和友情的处理不当无法解决这些问题的我有些头疼好像世界都已经把我抛弃了... :(
posting_time	The posting time of the tweet.	2020-05-01 00:32
posted_picture_url	A list of URLs of pictures posted in the tweet. If no pictures are sent, the default value is "None".	["http://wx2.sinaimg.cn/wap180/xxxx.jpg", "http://wx2.sinaimg.cn/wap180/yyyy.jpg"]
num_of_likes	The number of likes.	1
num_of_forwards	The number of forwards.	0
num_of_comments	The number of comments.	3
tweet_is_original	An identifier that identifies if the tweet is an original tweet.	True

• As of May 30, 2020, the candidate samples collected by the crawler and the various information of the samples in WU3D are shown in the following table. Once new data is updated later, the statistics in this table will be updated.

Dataset	Category	User	Tweet	Picture
Candidates	Depressed	125479	5478806	2354701
	Normal	65913	4927904	3631537
	Total	191392	10406710	5986238
WU3D	Depressed	10325	408797	160481
	Normal	22245	1783113	1087556
	Total	32570	2191910	1248037

Explanations and Issues

• In order to protect the privacy of depressed users, we have deleted the user IDs in WU3D.

• Instead, we have uploaded the crawler script and the script for data preprocessing. Researchers can further collect user samples based on our scripts. All the scripts and the deployment document README.md are placed in the folder WU3D_crawler_preprocessing.

• If encoding='utf-8' doesn't work, try: encoding='utf-8-sig' when using Python to load the JSON file.

• Other issues will be updated soon...

Contributions and Acknowledgements

• All the depressed user samples were manually data labeled by anonymous data labeling specialists and reviewed by psychologists and psychiatrists. We are grateful for the suggestions and help of these anonymous individuals.
• Chenghao Li (李丞浩) developed a Weibo data crawler and Zhang Yilin (张怡霖) carried out data screening and cleaning. The subsequent data labeling and expert review were also completed under the supervision and cooperation of both of them. We are very grateful for their work and contributions.
• The interpretation right of this dataset belongs to Chenghao Li (李丞浩) and Zhang Yilin (张怡霖). If there are any questions about the dataset, please contact:
  • Chenghao Li (李丞浩): [email protected] (Crawler Developer & Dataset Owner)
  • Zhang Yilin (张怡霖) : [email protected] (Dataset Owner)
  • Yiding Wang (王一丁): [email protected] (Github Project maintainer)
  • Zhenyi Wang (王臻懿): [email protected] (Github Project maintainer)