https://github.com/manoelhortaribeiro/hatefuluserstwitter

Code for the paper "Characterizing and Detecting Hateful Users on Twitter"
https://github.com/manoelhortaribeiro/hatefuluserstwitter

abuse-detection hate-speech twitter

Last synced: 26 days ago
JSON representation

Code for the paper "Characterizing and Detecting Hateful Users on Twitter"

• Host: GitHub
• URL: https://github.com/manoelhortaribeiro/hatefuluserstwitter
• Owner: manoelhortaribeiro
• License: mit
• Created: 2017-10-20T13:59:47.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2021-04-20T11:42:13.000Z (about 4 years ago)
• Last Synced: 2023-08-21T15:32:30.220Z (over 1 year ago)
• Topics: abuse-detection, hate-speech, twitter
• Language: Jupyter Notebook
• Homepage:
• Size: 4.09 MB
• Stars: 69
• Watchers: 5
• Forks: 11
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.txt

Awesome Lists containing this project

README

        
# Hateful Users on Twitter

This folder contains the data and the analysis done in the paper:

~~~bibtex

@inproceedings{ribeiro2018characterizing,

title={Characterizing and Detecting Hateful Users on Twitter},

author={Horta Ribeiro, Manoel and Calais, Pedro and 

        Santos, Yuri and Almeida, Virg{\'\i}lio and Meira Jr, Wagner},

booktitle={Proceedings of the International AAAI Conference on Web and Social Media},

year={2018}

}

~~~

    

The experiments with the *GraphSage* algorithm are in [another repository](https://github.com/manoelhortaribeiro/GraphSageHatefulUsers).

The dataset can be downloaded [here on Kaggle](https://www.kaggle.com/manoelribeiro/hateful-users-on-twitter).

## Data and Reproducibility

This dataset contains a network of 100k users, out of which ~5k were annotated as hateful or not. For each user, several content-related, network-related and activity related features were provided. Some of the files used are not shared because sharing them violates Twitter's guidelines. 

You can download the following files here:

- `bad_words.txt` list of bad words matched in the tweets.

- `lexicon.txt` list of lexicon used in the diffusion method.

And the following files on Kaggle:

- `users_(hate|suspended)_(glove|all).content` files with the feature vector for each user and their classes, the ones with `hate` label users as either `hateful`, `normal` or `other`, whereas the ones with `suspended` label users as either `suspended` or active. The ones with `glove` have only the glove vectors as features, the ones with `all` have other attributes related to users activity and network centrality. This is only for the GraphSage algorithm.

- `user.edges` file with all the (directed) edges in the retweet graph.

- `users_clean.graphml` networkx compatible file with retweet network. User id's correspond to those in `users_anon_neighborhood.csv`!

- `users_anon_neighborhood.csv`  file with several features for each user as well as the avg for some features for their 1-neighborhood (ppl they tweeted). Notice that `c_` are attributes calculated for the 1-neighborhood of a user in the retweet network (averaged out).

## Attributes description

      hate :("hateful"|"normal"|"other")

      if user was annotated as hateful, normal, or not annotated.

      

      (is_50|is_50_2) :bool

      whether user was deleted up to 12/12/17 or 14/01/18. 

      

      (is_63|is_63_2) :bool

      whether user was suspended up to 12/12/17 or 14/01/18. 

            

      (hate|normal)_neigh :bool

      is the user on the neighborhood of a (hateful|normal) user? 

      

      [c_] (statuses|follower|followees|favorites)_count :int

      number of (tweets|follower|followees|favorites) a user has.

      

      [c_] listed_count:int

      number of lists a user is in.

        

      [c_] (betweenness|eigenvector|in_degree|outdegree) :float

      centrality measurements for each user in the retweet graph.

      

      [c_] *_empath :float

      occurrences of empath categories in the users latest 200 tweets.

      

      [c_] *_glove :float          

      glove vector calculated for users latest 200 tweets.

      

      [c_] (sentiment|subjectivity) :float

      average sentiment and subjectivity of users tweets.

      

      [c_] (time_diff|time_diff_median) :float

      average and median time difference between tweets.

      

      [c_] (tweet|retweet|quote) number :float

      percentage of direct tweets, retweets and quotes of an user.

      

      [c_] (number urls|number hashtags|baddies|mentions) :float

      number of bad words|mentions|urls|hashtags per tweet in average.

      

      [c_] status length :float

      average status length.

      

      hashtags :string

      all hashtags employed by the user separated by spaces.

## Folder Structure

These are the main folders, reproducible with the dataset downloaded from **Kaggle**:

- `./analysis/` contains the script exploring the dataset collected.

- `./classification/` contains scripts with boosting classifier.

These folders are not reproducible, but they are present just in for completeness:

- `./crawler/` contains the code used to extract the dataset. You need to set neo4j to run it.

- `./prepreprocessing/` contains scripts to select the users to be annotated, and extract their tweets.

- `./features/` contains scripts to get the features to be analyzed and that will be fed into the classifier.

Auxiliary folders:

- `./data/` data generated by data wrangling.

- `./secrets/` for the API/DB authentication stuff.

- `./tmp/` auxiliary scripts.

- `./img/` images generated by analyses.