近年隨著輿情分析、市場調查與聊天機器人的蓬勃發展,情感分析的重要性逐漸增加,而情感分析的結果與情緒詞庫以及情緒語料庫高度相關。本論文參考網路上少數的情緒辭典例如大連理工大學情緒辭典,以 Facebook數百萬篇粉專貼文的語料庫,透過詞頻分析與人工篩選,整理出一份六千多詞的情緒辭典,並開發一個情緒分析應用程式可以用來分析文章的正負向與主要情緒類別。實驗的結果顯示,本論文的情緒辭典比大連理工的情緒辭典在 Facebook 文章情緒分析的結果上更為精準。 In recent years, Public Opinion Analysis, Market Research and Chatbot have beenbooming day by day. Sentiment Analysis is a basis for these applications and the resultof the Sentiment Analysis is related to the quality of the lexicons and the corpus beingused.In this thesis, we built a Chinese Sentiment Lexicon. We referred to the popularLexicons such as DUTIR and proposed a modified classification structure. We crawledmillions of Facebook articles as corpus to compute the term frequency, and selectedterms with non-trivial frequency that are related to Sentiment into our Lexicon. A webapplication is developed to perform sentiment analysis on articles. The program willprovide a positive-vs-negative assessment plus sentiment classification for the articles.The experiments showed that our Lexicon is more precise than the DUTIR on theFacebook articles.