IIIT-MLNS at SemEval-2022 shared task # 8.
Base architecture: Siamese network
- Features: News text, metadata, NER-extracted features
- Base encoder model: XLM-RoBERTa, Multilingual DistilBERT
- Encoder representation concatenations: [|x1 - x2|; (x1 + x2)/2], [x1; x2; |x1 - x2|]
- Data augmentation
- Output activation
- Ensembling