FAMOUS: Fake News Detection Model Based on Unified Key Sentence Information

Fake news detection causes a challenging problem due to the great influence of communication media over the public. In this paper, we shall present a new fake news detection model using unified key sentence information which can efficiently perform sentence matching between question and article by using key sentence retrieval based on bilateral multi perspective matching model. Our model makes use of one unified word vector for the key sentences of article by extracting them to the question from article and then merging the word vector for each key sentence. It can efficiently perform the sentence matching by executing matching operations between the contextual information obtained from the word vectors of question and key sentences through bidirectional long short term memory. Our model shows the competitive performance for fake news detection on the Korean article dataset over the previous result.