Fine-Grained Hate Speech Community Detection

Twitter communities

This API allows to visualize the different hate speech target community networks obtained by mixing data views of a different nature (feature or graph space)

  • Clustering techniques: K-means & K-medoids {KM}, Spectral clustering {SC}, and Multi-view Spectral Clustering {MvSC}
  • Feature sets
    • mBERT view {mBERT} - multilingual BERT
    • mUSE view {mUSE} - multilingual Universal Sentence Encoder
  • Affinity matrices
    • Projection view {mBERT-PRO & mUSE-PRO}
    • Network view {mBERT-NET & mUSE-NET}

Parameters

Network Projection

(click on nodes to explore)