The official implementation of findings of NAACL 2022 paper ''Analyzing the Intensity of Complaints on Social Media''. Here, we release the annotated corpus and code for this work.
This Weibo data is collected from August 2020 to May 2021 through the official API, which contains 3103 posts, Our Weibo dataset can be found at Datafolder. For more details about the Weibo dataset, please contact fangming@smail.nju.edu.cn. Thanks!
In this work, we create the post complaint intensity dataset, where intensity values are between -1 and +1. Data collection, preprocessing and annotation process of the data are discussed in detail in our paper. The statistics of our corpus are as follows:
| Hashtag | Num. |
|---|---|
| #代表建议让学生在校内完成家庭作业# (#The representative suggested that students should complete their homework on campus#) | 762 |
| #江苏明确教师不得用手机布置作业# (#Jiangsu Province makes it clear that teachers are not allowed to use mobile phones to assign homework#) | 534 |
| #院士不建议普通孩子学奥数# (#Academician does not recommend ordinary children to learn Mathematical Olympiad#) | 627 |
| #西安外国语大学封闭管理# (#Close management of Xi'an International Studies University#) | 598 |
| #人大法硕复试30余人成绩0分# (#More than 30 people scored 0 in the postgraduate examination of law at Renmin University#) | 582 |
-
Preprocessing:
- For each post, we filtered out posts with less than 10 words and more than 200 words. And we removed the name of the author, loca- tion tags, and URLs.
- We combine Weibo posts from different hashtags together and then split them into 80% for training, 10% for validation, and 10% for test.
-
All annotated post are stored in
data.csvfile underDatafolder. There are two columns "post" and "intensity" respectively, where "post" is the post after preprocessing, and "intensity" is the complaint intensity corresponding to this post.
Code will be uploaded soon.