A Study on Social Sentiment Analysis and Product Review Opinion Mining

[[abstract]]隨著網路環境的快速發展，社群網路上往往有許多公共議題或新產品發表的討論文章，引起了研究人員對於文章中所表達的情緒進行分析之渴望，因為完整的分析結果在某種程度上可以代表著整體網路輿論的風向球，以做為政府施政依據，或是產品行銷策略的調整參考，見解探勘與情緒分析正是針對商品特性以及評論代表情緒的主流分析方法。本研究提出一套將見解探勘與情緒分析進行交叉分析比對的系統，期望能找出產品以及社群事件之間的發展關聯性。 我們對這兩種分析方式各別設計了兩種資料分析模組，在見解探勘方面本研究提出品牌見解探勘模組，以特定品牌為基礎包含所有商品的綜合評價，更具備適地性資料幫助延伸分析深度。而在情緒分析方面本研究提出社群事件情緒探查模組，其中包含一套自我學習的情緒辭典，試圖解決因網路上持續增加新字詞，所導致之情緒分析判斷可能隨時間流逝而失去準確性的問題。 搭配上述兩個模組，本研究的交叉分析能夠找出當某特定品牌相關社群事件發生之後，是否會影響之後該品牌的產品見解探勘的結果，此外經由本系統的分析還能找出資訊的可信度、代表度，幫助過濾不具參考價值之資料。[[abstract]]With the rapid development of the Internet applications people now tend to pose comments on social networking Web sites for emerging public issues or new product unboxing Such phenomenon inspires researchers eager to conduct the sentiment analysis on these comments as a complete analysis can be an important reference for the government's policy or the adjustment for product marketing strategies This research can bring forward a cross-analysis system of sentiment analysis and opinion mining aim of finding the association between product and social event In this paper we designed two different data analysis modules for both analysis methods for opinion mining the module is called brand-based opinion mining module the analysis based on specific brand and the comprehensive review of all products belong to the brand including the geographical information to extend the analysis depth For sentiment analysis the module is called social events sentiment discover module including a self-learning dictionary with mood classification which attempts to help us solve the problem of diminishing effect due to thriving neologisms have been continuously introduced into the Internet community With these two modules our cross-analysis will find out when a social event happened against to a brand weather or not influences their products opinion mining’s result furthermore this system can also bring out the credibility and representivity which will filter out the unreliable dat