歷年優秀論文獎助 > 2012優秀學位論文得獎名單 >張育蓉
研究生 | 張育蓉 |
研究生(外文) | Yu-Jung Zhang |
論文名稱 | 使用情緒分析於圖書館使用者滿意度評估之研究 |
論文名稱(外文) | A Study on Library Users’ Satisfaction Evaluation Using Sentimental Analysis |
指導教授 | 郭俊桔 |
指導教授(外文) | |
口試委員 | 陳光華、柯皓仁 |
口試日期 | 2012-07-25 |
學位類別 | 碩士 |
校院名稱 | 國立中興大學 |
系所名稱 | 圖書資訊學研究所 |
論文出版年 | 2012 |
畢業學年度 | 100 |
語文別 | 中文 |
論文頁數 | 121 |
中文關鍵詞 | 使用者滿意度、情緒分析、圖書館 |
外文關鍵詞 | users’ satisfaction、sentimental analysis、library |
中文摘要 | 圖書館服務品質與使用者滿意度息息相關,圖書館為瞭解使用者滿意度,每年皆會進行使用者滿意度調查。然而,傳統評估方法有些實施上的限制且須耗費大量人力、時間與成本,無法立即反映出圖書館問題,也無法確實得知使用者之不滿意事項,並及時改善。因此,利用情緒分析探究社群網路上(如噗浪等)之使用者對於圖書館的意見屬於正向或負向,並以圖形方式顯示評估結果,可及時提供圖書館業務調整與改善使用者滿意度。 本研究由社群網路上蒐集使用者對圖書館的意見,藉由人工標記方式標記出使用者情緒、情緒詞、否定詞和情緒搭配詞後,擷取出相關情緒分析用之字典。為了將使用者意見自動分類為館員、館藏、服務、設備和空間與環境等五大類,使用KNN、NB和SVM等資料探勘工具分別探討分類成效。接著,分別使用程度詞加權之情緒分析法和情緒極性與類別象限之情緒分析法,針對每一類的語料集探討情緒分析的成效。最後,探討使用長條圖和雷達圖等圖形化方式呈現使用者情緒。 本研究除提出使用情緒分析於使用者滿意度評估的系統架構,也驗證其可行性。主要的貢獻則有:(1)建置情緒詞、情緒搭配詞、程度詞與否定詞辭典,並摘選出專屬圖書資訊學領域使用之情緒分析辭典;(2)語料前置處理,例如,特殊句型之情緒詞判定,建議反諷句以情緒搭配詞標記,比較句則加以改寫;(3)使用KNN分類器進行語料類別分類成效最佳;(4)利用情緒極性與類別象限搭配Plurk語料進行情緒分析,可得到令人滿意的情緒分析結果;(5)將情緒分析以長條圖和雷達圖呈現,除可清楚顯示圖書館優劣之處,更可以改善圖書館服務而得到更高的使用者滿意度。 |
外文摘要 | As the quality of library service is closely linked with the users’ satisfaction, the libraries usually employ the questionnaire to collect and analyze the users’ satisfaction annually. However, due to difficulty of questionnaire design, low recycle rate and difficulty of statistics analysis and explanation, the results obtained from questionnaire cannot show the real users’ satisfaction. Moreover, as the questionnaire method cannot provide the on-line users’ satisfaction, the librarians cannot deal with the related service problems immediately. On the contrary, as the rapid growth of social network, the opinions can be collected from Internet and analyzed to understand sentiments, i.e., positive or negative, and sentimental trend of the users easily. Thus, this dissertation employs the sentimental analysis to extract the users’ positive or negative opinions from the social network, e.g. Plurk. First, the library users’ opinions are collected from social network and used as the corpus. This corpus is annotated manually to tag sentimental words, negative words, sentimental collocation words, degree words and users’ sentiments. Then, the related dictionaries for sentimental analysis are extracted. In order to cluster the users’ opinions into 5 categories, i.e., librarians, collections, services, equipments and environment, KNN, NB and SVM are used to evaluate the classification performance. Additionally, two sentimental analysis methods, which are weighted degree words methods and polarity-strength methods, respectively, are proposed to study the performance of sentimental analysis. As the experimental results are similar to the results of both using questionnaire method and manual annotations, the feasibility and effectiveness of the proposed methods can be proved. Besides, as the graph representations, i.e., radar charts and bar charts, approaches are employed to show the evaluation results, the users’ satisfaction can be easily obtained and understood so as to be able to improve the library services in time and enhance the users’ satisfaction. |
連結 | 臺灣博碩士論文知識加值系統 |