Data and codes to reproduce main results in paper titled Breaking the Spiral of Silence? Investigating the Influence of Opinion Reversal Events on Opinion Expression on Social Media

DataAnnotated DataThe dataset comprises human annotated for the stance of the tweet related to Abe. The annotated data is used to train the stance detection model. Refer to the supplymentary materials for more details.Twitter Data`TWEET_DAILY`: The dataset comprises tweets that were included in the analysis, organized by date. `tweet_id`: The unique identifier for each tweet. The original tweets can be retrieved using the Twitter (X) API. `created_at`: The published date of the tweets. `sentiment`: The predicted sentiment (equivalent to stance) of the tweets.> Due to X (formerly Twitter)’s data sharing policies, we are only permitted to publicly share the Tweet IDs used in this study. In accordance with these policies, users who wish to access the full tweet content must retrieve it themselves via the Twitter API. To facilitate replication and verification of our analysis, we provide a processed dataset containing aggregated data based on the original dataset. This processed dataset reproduces the key variables and statistical measures used in the study. Details of the aggregation procedure are described below.The count of published tweets expressing different stance per user `INDIVIDUAL_USER_SENTIMENT`: We aggregate the number of tweets expressing different stances per user, organized by date. `author_id`: The unique identifier for each user.`retweet_type`: The publication type of the tweet ('retweeted', 'replied_to', 'original', 'quoted')`Negative`: he number of tweets in which the user expressed a negative stance on the specific date`Neutral`: he number of tweets in which the user expressed a neutral stance on the specific date`Positive`: he number of tweets in which the user expressed a positive stance on the specific dateCode`config.py`: Revise the paths and parameters based on your specific environment.`PROJECT_ROOT` should be revised according to the user's environment.`compile_data.py`: Process the data for the following analysis. The processed data will be save at path `INDIVIDUAL_DAILY_SENTIMENT_DATA`.> The data processing requires to use the original Twitter dataset. If the original data is not properly configured, we provide a preprocessed `USERS_DUPLICATION_RATE` file that records the duplication rate of published tweets per user, which can be used to process the data without the original Twitter dataset.```figure1.ipynb```: Reproduce the visualization results in Figure1```figure2.ipynb```: Reproduce the visualization results in Figure2```Regression.R```: Reproduce the regreesion results represented in Table 1 and Table 2Configure the ```INDIVIDUAL_DAILY_SENTIMENT_DATA``` using the processed data shared in the archive, or use `compile_data.py` to process the raw data.