On the Benefit of Combining Neural, Statistical and External Features for Fake News Identification

Benchmark Model Rank Results
fake-news-detection-on-fnc-1Bhatt et al.#2Weighted Accuracy: 83.08Per-class Accuracy (Unrelated): 98.04
fake-news-detection-on-fnc-1Baseline based on skip-thought embeddings (Bhatt et al., 2017)#5Weighted Accuracy: 76.18Per-class Accuracy (Unrelated): 91.18
fake-news-detection-on-fnc-1Baseline based on word2vec + hand-crafted features (Bhatt et al., 2017)#6Weighted Accuracy: 72.78Per-class Accuracy (Unrelated): 96.05
fake-news-detection-on-fnc-1Neural baseline based on bi-directional LSTMs (Bhatt et al., 2017)#7Weighted Accuracy: 63.11Per-class Accuracy (Unrelated): 78.27