Rumor Detection Dataset (Twitter15 and Twitter16)

This dataset features labeled source tweets from the widely used Twitter15 and Twitter16 datasets, designed for rumor detection and misinformation classification research. By categorizing tweets into rumors or non-rumors, this dataset is ideal for training supervised machine learning models and exploring the relationship between tweet content and rumor classification.

Key Features

1. Source Tweets:

   • Contains the text content of original tweets.
   • Simplified focus on tweet text for straightforward text classification tasks.

2. Labels:

   • Categories include True, False, Unverified, or Non-rumor, making it versatile for both binary and multi-class classification.

3. Folder Structure:

   • Twitter15:
     • source_tweets: Contains the raw tweet text.
     • labels: Corresponding labels for each tweet.
   • Twitter16:
     • Similar structure to Twitter15 with tweets and their respective labels.

Applications

1. Natural Language Processing (NLP):

   • Preprocessing, tokenization, and feature extraction for text-based models.
   • Fine-tuning transformer models like BERT, RoBERTa, or GPT for rumor classification.

2. Text Classification:

   • Training models to detect rumors and misinformation.
   • Benchmarking machine learning algorithms for accuracy and performance.

3. Social Media Analytics:

   • Analyze how rumors spread and are expressed on platforms like Twitter.
   • Insights into user behavior and content patterns related to misinformation.