A method for determining a polarity of news sentiment, the method comprising steps of: performing, by means of a predetermined machine learning algorithm, semantic-based scoring of news to undergo prediction to obtain a sentiment score of the news to undergo prediction (S31); adjusting, according to a predetermined rule between event tags and event keywords, the sentiment score of the news obtained by the predetermined machine learning algorithm (S32); and determining, according to the adjusted sentiment score of the news to undergo prediction, a polarity of sentiment of the news to undergo prediction (S33). The method can improve accuracy in determining a polarity of news sentiment.
展开▼