A Cantonese rumor dataset for the paper 《Identifying Cantonese Rumors with Discriminative Feature Integration in online Social Networks》
We develop a web crawler to collect Cantonese tweets from Twitter and finish the data annotation to construct this dataset. This dataset contains 27,328 tweets, including 13,883 rumors and 13,445 non-rumors.
CR-Dataset-example.csv: This file provides the source tweets and labels in a format like: 'text, label'.
Example:
| Text | Label |
|---|---|
| 今日警察已經光明正大喬裝示威者到處破壞而唔需要隱藏！光明正大拿槍對準市民 #929globalantitotalitarianism #hkpolicebrutality https://t.co/OURXJZyfag | 1 |
The time of data collection:
- February 2020 - April 2020
- November 2021 - December 2021
The reports from three sources are recognized as the bases of facts:
- Hong Kong Government News
- Hong Kong Police Force
- Top 10 Rumors of Mobs in Hong Kong in 2019 from Southern Metropolis Daily
Data annotation is strictly in accordance with reports from the above three sources, and does not involve personal political positions. This dataset is used for academic research only.
If you find this dataset useful, please cite our paper.