A Study on Satisfaction Evaluation by Community Platforms' Consumer Review Using Sentiment Analysis─A Case Study of TripAdvisor

Along with explosion of network information, the massive data have nearly 80 percent of non-structured information data (Ye, 2013). Because of the massive data set, the column-'The Age of Big Data' in The New York Times at 2012 that announced the era of Big Data has begun. Therefore, timely and cost-effectiveness of big data analytics has become common in each industry. In addition, online forums, web blogs, Twitter and Facebook were appeared, that has generated massive consumer reviews (Gunter, Koteyko, & Atanasova, 2014). Scholars have been doing research using consumer reviews, such as O'Connor et al. (2010) used Twitter's reviews to investigate US presidential election, confirmed that Twitter's reviews and Obama's supporting data similar, but this have yet to elaborate the validity of between sentiment analysis and ratings (Gunter, Koteyko, & Atanasova, 2014). Moreover, many studies investigate consumer satisfaction by asking consumers, less studies using sentiment analysis compare with satisfaction. Thus, we use online review data on TripAdvisor to do sentiment analysis, calculate sentiment score, and understand its relevance by regression analysis. Then we will classification of reviews to understand which item are consumer like, which item are consumer don't like. This study found that sentiment score and satisfaction has validity; sentiment score can represent consumer satisfaction with the company's products or services, and also found the effect of aspect rating, satisfaction and sentiment score, 'Value' is the most important item of hotel's consumer. Only the effect of high rating's hotels' aspect rating and sentiment score, 'Service' and 'Rooms' are consumer most important item.隨著網路資訊的爆炸，龐大的數據資料中有近80%的資料為非結構性資料（葉乃嘉, 2013），龐大的數據集，也讓《紐約時報》在2012年的專欄文章「The Age of Big Data」中正式宣布了大數據時代的來臨，因此，及時、具有成本效益的大數據（Big data）已成為各行業分析的主流。再加上線上論壇、部落格、推特（Twitter）、臉書（Facebook）等網路平台的出現，產生數量龐大的消費者評論 (Gunter, Koteyko, & Atanasova, 2014)，學者紛紛使用消費者評論做研究，像是O'Connor et al. (2010)使用Twitter評論與美國總統選情做分析，證實評論與歐巴馬的支持數據相近，但情緒分析與支持率之間的效度為何卻尚未詳加說明 (Gunter, Koteyko, & Atanasova, 2014)，再者，許多研究透過詢問消費者本身探討消費者的滿意度，但較少人使用情緒分析與滿意度相比較，因此，本研究使用TripAdvisor線上評論資料作情緒分析，計算出情緒分數，透過迴歸分析了解其相關性，並將評論分類了解消費者滿意不滿意之項目。 本研究證實情緒分數對於滿意度具有效度，情緒分數可代表消費者對企業的產品或服務之滿意度；並且發現七大類滿意度（Aspect Rating）與滿意度以及情緒分數的影響中，價值（Value）都是飯店消費者最重視的項目，唯有評分較高飯店七大類滿意度與情緒分數的影響中，服務（Service）與房間（Rooms）為消費者所最重視的項目。謝辭..............................................i 摘要.............................................ii Abstract........................................iii 目錄.............................................iv 圖目錄............................................v 表目錄............................................v 第一章、 緒論......................................1 第二章、 文獻探討..................................3 第一節 大數據....................................3 第二節 消費者評論................................4 第三節 情緒分析..................................6 第三章、 研究設計.................................10 第一節 研究資料.................................10 第二節 研究方法.................................11 第三節 資料分類.................................17 第四章、 分析結果.................................19 第一節 資料結構.................................19 第二節 情緒分數與滿意度分析之結果.................19 第三節 情緒分數與七大類滿意度分析之結果...........21 第五章、 結論與建議...............................27 第一節 結論....................................27 第二節 討論....................................27 第三節 管理意涵................................28 第四節 研究限制與未來發展 .......................29 第六章、 參考文獻................................30 附錄一 AFINN詞彙與分數表.........................38 附錄二 七大類特徵詞表............................5