Table 1 Summary of related works.
From: Multimodal based Amharic fake news detection using CNN and attention-based BiLSTM
Category | Author(s) | Methodology/approach | Key contributions/results | Limitations |
|---|---|---|---|---|
Amharic Fake News Detection (Text Only) | Gereme et al.7 | Deep learning; Custom word embedding (AMFTWE); ETH_FAKE dataset | Introduced dataset and embedding for Amharic; Model trained on new corpus | Used homogeneous data only |
Tazeze8 | ML (Naive Bayes, SVM, etc.); Feature extraction (TF-IDF, n-grams) | Achieved up to 96% accuracy; strong performance from Naive Bayes & Passive Aggressive | Used homogeneous data only | |
Arega9 | TF-IDF-based ML model | Social media fake news detection | Missing algorithm and performance details | |
Woldeyohannis10 | Feature fusion (TF-IDF, word2vec); ML (RF, SVM, LR) | RF + word2vec reached 99.67% precision | Small dataset; limited to homogeneous data | |
Multimodal Fake News Detection (English) | Palani et al.4 | BERT + Capsule Neural Network | Enhanced performance with text + image fusion |  |
Wang et al.5 | Multimodal fusion with attention | Fine-grained feature fusion improved detection | Â | |
Kumari and Ekbal11 | ABS-BiLSTM, CNN-RNN, MFB, MLP | Multilevel attention-based fusion; improved accuracy | Â | |
Fake Image Detection (social media) | Shankar et al.12 | LBPNET (CNN with pair-wise learning) | Robust detection of GAN-generated faces | Used homogeneous data only |
AlShariah and Jilani Saudagar13 | CNN, AlexNet (transfer learning) | Achieved 97% accuracy; Instagram-focused | Focused on specific platform only |
