EmoRumorFramework

Code for the paper An Emotion Analysis Framework for Rumours in Online Social Media

Dataset

Please download and extract data to raw_data directory, the data dir should look like the following:

- raw_data
  - pheme
    - pheme_single_aff.csv
  - twitter15
    - twitter15_single_aff.csv
  - twitter16
    - twitter16_single_aff.csv
  - fig # directory for plotted figure

All data are organized in a unified dataframe format. The specific columns may vary, but the commonly useful attributes are as follows:

id: UID of the tweet
text: tweet text
rumour_type: rumours or non-rumours
emollm_emotion: a string represents multi-label emotion detected by EmoLLM, different emotions are separated by comma
emollm_sentiment: sentiment label detected by EmoLLM
structure: json structure of the conversation, only available for root tweets. For comments, this column will be 'reaction'.
factuality: true, false or unverified