台灣新聞媒體及輿論對待陸生的情感傾向研究

Abstract

「陸生」是中國大陸赴台就讀學生的官方稱謂，由於海峽兩岸互通往來時間未久，長期以來的隔閡使得台灣民眾在對待「陸生」這個特殊群體時可能會有較為複雜的情感。而公眾對於「陸生」的了解通常基於新聞媒體的報導和其他民眾的評論，本研究意圖探究台灣新聞媒體及公眾輿論對待「陸生」的情感傾向，同時試圖發現前者對於後者的影響。本研究首先通過基於語義的情感詞典匹配之方法對新聞文本進行情感評分並判斷傾向；然後通過基於人工標註語料的遞歸神經網絡語言模型（RNNLM）對公眾評論文本進行傾向判別；最終正規化兩者評分試圖發現對應規律。

‘Chinese Mainland Students’ is the official name of those students studying in Taiwan who come from Mainland China. Due to the short time from both sides of the straits communicating to each other, peoples in Taiwan may harbour some complex emotions towards this special group of students for a long estrangement. Generally peoples’ concepts towards those students comes from news and public opinions. Our research aims to discover emotional tendencies towards ‘Chinese Mainland Students’ in news and public opinions in Taiwan, at the same time, to find whether a guidance is between both sides. Firstly, we use emotion dictionary matching based on semantic analysis to give sentiment scores of each news articles, meanwhile, to judge sentiment polarity. Secondly, we train Recurrent Neural Network Language Models based on artificial labeled corpus to estimate sentiment polarity of public opinions, in other words, peoples’ comments. Finally, we normalize both scores and try to find some correspondence rules.