We utilized a web crawler to climb Twitter to build a TwiCOVID19 dataset of the COVID-19 pandemic.
We replaced URLs and user handles (@user-name) with the symbols “<url>” and “<user>”
We label tweets with positive, neutral, or negative polarities using TextBlob
The timeframe of the TwiCOVID19 dataset is from 2021-01-01 00:00:00 to 2021-12-31 23:59:59.
- covid19_tweet.csv desipts information of tweets.
- covid19_user.csv desipts information of users.
- Event tags: #covid19, #covid-19, #coronaviruspandemic, #coronavirus, and others.
- Tweet information: tweet ID, content, posting time, user ID, tag, number of retweets, number of favorited, number of replies, source tweet ID (unique field of retweets), and sentiment label.
- User information: user ID, gender, nickname, number of followers, number of friends, number of favorites, and number of users’ tweets.
Here is the statistics of TwiCOVID19 dataset (“#” DENOTES “NUMBER OF”):
| Statistic | TwiCOVID19 |
|---|---|
| # Tweets | 17,675 5 |
| # Positive tweets | 10,052 |
| # Neutral tweets | 3,999 |
| # Negative tweets | 3,624 |
| avg. # words per tweet | 23 |
| # Users | 7,489 |
| # Forwarding tweets of users | 15,350 |
| Density of user forwarding network | 0.0274% |