結合來源與內容之虛假資訊偵測機制

A Novel Source-Content-Topic Dictionary Model for Misinformation Detection

社群媒體成為大眾獲得新聞的主要來源，撼動了新聞生態，隨著社群擴散的假新聞造成積非成是，混淆大眾視聽，為此偵測「虛假資訊」成為必要且迫切的手段。本研究提出一個基於來源－內容－主題的信任網路模型架構，以偵測虛假新聞，演算法考慮到資訊來源過去行為記錄，可視為資訊傳播者的信任額度，應該納入在虛假資訊的偵測，可幫助預警某些虛假資訊的散播，提早遏止這些資訊在社群媒體繼續流通。本研究採用g0v零時政府「Cofacts真的假的」專案中被標記的真假新聞資料，這個是台灣寶貴的真實虛假新聞的資料集，這次研究重點以醫藥新聞為主，當作實驗的資訊標的。

With the rapidly growth of social media, the spreading of information has become faster than ever. It’s a challenge job for reader to justify the reliability of information. Large amounts of online fake news has the potential to cause serious problems in society. To overcome previous problem, this study proposed a novel“Source-Content-Topic dictionary”model to detect misinformation. Our approach does not only rely on the news content classification but also incorporated with information source. To improve the efficiency of content detection, we also apply some deep learning models including Word2Vec, Doc2Vec etc. Finally, we evaluated the performance in a real dataset which was collected by a famous organization called g0v. Results show that the proposed model is both effective and efficient in addressing the misinformation detection problem. In this study, we only focused on the misinformation at medical area. In the future, the proposed model could be evaluated in different aspects of news including politic, etc.