近年來眾多研究探討文字資訊(如公司年報致股東報告書、新聞報導、分析師報告等)對於投資人決策之影響。這些研究所採取之方法中,內容分析法為最主要者之一。 內容分析法多根據情緒詞典衡量文字資訊所隱含之悲觀或樂觀情緒(sentiment),因此良好之情緒詞典實為文字資訊分析最之根本。然而檢視當前之中文情緒辭典,缺乏專門針對財經領域所編撰。本研究之目的即在建置一專門應用於財經領域之中文情緒辭典。 本研究以Loughran and McDonald所建置之財經領域英文情緒辭典為基礎,首先將其翻譯成中文,並以電腦比對中文新聞進行篩選。同時,進一步以群組共識之方式擴編與刪減,編製成中文財經領域之情緒辭典。 為驗證此一中文辭典之有效性,本研究進一步選取兩百篇新聞報導,分別以本研究所建立之辭典與台大資工系情緒辭典,衡量這些新聞報導的情緒分數,並以之與股票之表現進行相關與迴歸分析。結果發現,依據本研究所建立之辭典確實可以有效衡量新聞報導所隱含之樂(悲)關情緒。
Many studies have investigated the effect of textual information (e.g. MD&A, news reports, analyst reports) on investors’ decision. One of the most important methods adopted by these studies is content analysis. Content analysis usually tries to measure the sentiment contained in textual information based on some sentiment dictionary; therefore a good sentiment dictionary is the foundation for content analysis. However, currently there is no such Chinese dictionary in the area of economics and finance. This study intends to fill this gap by building a Chinese economic/finance sentiment dictionary. Our dictionary is based on the Englished economic/finance sentiment dictionary build by Loughran and McDonald. We first translate the English economic/finance sentiment dictionary into Chinese. We the use computer to calculate the number of times each word appear in 100 selected news reports. We further manually extend and remove some words by using a group consensus method. In order to verify the validity of our dictionary, we select 200 news reports as a sample and measure the sentiment of these news reports using our dictionary and the dictionary built by the department of computer science at National Taiwan Universituy (NTU). We then use correlation and regression analyses to investigate the relationship between the sentiment scores and stock performance. The analysis results show that our dictionary can effectively measure the sentiment of news report and outperforms the NTU dictionary.