This repository contains a dataset of social media posts about artificial intelligence along with the date, platform, username, and country code where available. The data comes from YouTube (66.6%, comments), Reddit (28.5%, comments), Twitter (4.9%, tweets). It also contains all the source code that was used for gathering the data, as well as useful visualizations and analysis which pulls insights from this data.
This work was done as group coursework for the Applied Data Science unit at the University of Bristol.
The dataset located in AI_sentiment_dataset.csv contains the following attributes:
| Attribute | Description |
|---|---|
| Content | The text of the social media post/comment |
| Date | Date the post/comment was posted. Format: YYYY-MM-DD |
| Reactions | Number of likes/upvotes. Can be negative for Reddit data. |
| N_Children | Number of replies to the post/comment. Note that the dataset does not include replies |
| Post Title | Title of the YouTube video/Reddit post the comment is under |
| Platform | One of ['YouTube', 'Reddit', 'Twitter'] |
| Country | 3-letter country code following ISO 3166 Alpha-3. Note that this column is empty for all Reddit data and most of YouTube data. |
The code used to gather the data is in src/gathering. To present findings
from the dataset, we used an ensemble of sentiment analysis models. This
analysis can be found in src/analysis. We then visualize our findings
in a series of Jupyter notebooks in src/visualization. The charts this
generated are also present in src/visualization/figures.
Data in various stages of being gathered and analyzed is in data.