تشخيص اخبار جعلي به وسيله طبقه‌بندي موضع

These days increase in unauthenticated internet sources has led to vast spread of fake news. Failure to detect misinformation pro‎mp‎tly can have significantly irreparable damages on the walk of life. Recent researches have improved stance classification as a primitive step to detect fake news in English. Moreover, they have less focus on recognizing fake news. In this work, we develop a deep learning model based on ParsBERT to improve the accuracy of the Persian stance classification. We over-sample the available imbalance dataset in Persian to compensate lack of data in such a low-resource language. Then, we implement a model to detect fake news in Persian by using our best stance classifier and other manually extracted features. Consequently, we achieved an accuracy of 82% for stance classification and 99% for fake news detection on the employed dataset.

205070

دانشگاه علم و صنعت ايران

مهندسي كامپيوتر

دكتر صالح اعتمادي