透過您的圖書館登入
IP:3.143.168.172
  • 學位論文

以評論內容為基礎之評論特質趨勢分析模式

The Model for Sentiment Analysis of Comments

指導教授 : 侯建良

摘要


當企業推出新產品/服務時,消費者往往透過各類管道針對該產品/服務發表評論,並透過評論表達其對於產品/服務之關注重點、投入情緒、整體評價等,此些評論特質會隨時間產生趨勢的變化(如關注重點的改變、正負情緒比例的增減等),而企業可透過此些趨勢的變化掌握市場的最新需求,以了解消費者偏好之變動趨向,進而發展符合消費者期待之產品/服務。另一方面,消費者亦可透過其關注之產品/服務的相關評論趨勢了解目前市場的輿論與評價趨向,並作出更切合市場脈動趨向之消費決策。然而,此些評論往往包含過多複雜資訊,企業如欲發展符合消費者期待之產品/服務需花費更多時間分析評論內容中所透露之消費者的關注重點、投入情緒、整體評價等趨向。另一方面,消費者為了解其他消費者欲傳遞之產品/服務的正負評價亦須投入更多精力自龐大且複雜之資料中取得產品/服務相關資訊(如使用情形、使用心得、購買價格等)。 為解決企業與消費者在處理評論內容過程中所面臨之各項問題,本研究乃於前置階段中先蒐集評論者於網際網路中針對特定產品/服務所發表之評論內容,並整理各評論內容之評論發佈時間、關注重點、投入情緒、整體評價等特質,再依據此些特質之整理結果建置詞彙庫;之後,本研究乃依據此些評論內容之整理結果發展一套「以評論內容為基礎之評論特質趨勢分析模式」方法。而「以評論內容為基礎之評論特質趨勢分析模式」方法可區分為「評論特質擷取」、「評論資料分群」及「評論特質統整及分佈趨勢分析」等三大階段。其中,「評論特質擷取」階段可依前置階段所建置之詞彙庫擷取評論內容的相關特質(如關注重點、投入情緒、整體評價等);「評論資料分群」階段乃依評論者發表產品/服務之評論內容的發佈時間將各評論進行分群,以作為後續階段進行評論內容趨勢分析之基礎;「評論特質統整及分佈趨勢分析」階段則將各評論內容之關注點、情緒傾向、評價判斷等三大特質以視覺化方式呈現,以凸顯其內容隨時間變化之趨勢。 未來,企業在新產品/服務推出時,即可參考過去消費者對於同性質之產品/服務所透露的關注重點、投入情緒、整體評價等特質之趨勢,讓新產品/服務可更符合消費者需求。另一方面,消費者亦可透過此些趨勢之變化了解目前市場的輿論與評價趨向,進而選擇更合適之產品/服務。

並列摘要


As a company releases a new product service to the market, consumers often express their interests, emotional tendencies and overall evaluations over the internet through comments. The trend of comment characteristics always changes with time. Enterprises can capture the latest market demand based on the tendency explored. On the other hand, consumers can also make their own decisions by these references. However, complex information might exist in the comments and thus companies have to spend much time in analyzing comments characteristics. Furthermore, consumers should also dedicate more efforts to acquire information about product usage, feedbacks and t price from the complicated information. In order to solve the problems, this study develops a model for sentiment analysis of comments. It can be used to visuallyreveals the tendency of the comments. By applying the proposed model, enterprises can refer to the latest trend of interests, emotional tendencies and overall evaluation of consumers. As a result, new products / services can fit more to consumer needs. Moreover, the consumers can acquire useful opinions from the public in order to select more suitable products or services.

參考文獻


1. Abilhoa, W. D. and de Castro, L. N., 2014, “A Keyword Extraction Method from Twitter Messages Represented as Graphs,” Applied Mathematics and Computation, Vol. 240, pp. 308-325.
2. Ahmed, M. A., Azuraliza, A. B. and Mohd, Z. A., 2014, “Fast Dynamic Clustering SOAP Messages Based Compression and Aggregation Model for Enhanced Performance of Web Services,” Journal of Network and Computer Applications, Vol. 41, pp. 80-88.
3. Akter, R. and Chung, Y., 2013, “An Evolutionary Approach for Document Clustering,” IERI Procedia, Vol. 4, pp. 370-375.
4. Alush, A., Friedman, A. and Goldberger, J., 2016, “Pairwise Clustering Based on the Mutual-information Criterion,” Neurocomputing, Vol. 182, pp. 284-293.
5. Alvaro G., 2015, “Techniques to Reduce Cluttering of RDF Visualizations,” Future Generation Computer Systems, Vol. 53, pp. 152-156.

延伸閱讀