博碩士論文 104451004 完整後設資料紀錄

DC 欄位 語言
DC.contributor企業管理學系在職專班zh_TW
DC.creator張偉德zh_TW
DC.creatorWei-Te Changen_US
dc.date.accessioned2018-1-19T07:39:07Z
dc.date.available2018-1-19T07:39:07Z
dc.date.issued2018
dc.identifier.urihttp://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=104451004
dc.contributor.department企業管理學系在職專班zh_TW
DC.description國立中央大學zh_TW
DC.descriptionNational Central Universityen_US
dc.description.abstract現今企業重視聲譽,將其視為重要資產來管理,媒體是傳遞企業意象的重要管道。過去研究為取得企業聲譽,採問卷方式,取得聲譽量表,進行後續分析,但很少有研究是從媒體的評論來推測企業聲譽,媒體一直是各利益關係人獲取企業資訊的來源之一,而媒體報導的觀點會影響企業聲譽的走向。 情感分析(Sentiment Analysis)可以解析文章或評論的內容,確立一個人對某件事物,所想表達的觀點或態度,過去研究,以判斷偏向正面、中立或負面為主,但人的情感複雜多變,人們藉由文字所表達的情感也是,透過情感詞典找出文中情感變數分數進行分析,對研究更有助益。 在網際網路的推波下,取得這些媒體評論集,相較於過去較為容易,藉由過往研究常用的聲譽問卷,向從事媒體工作的人員發放問卷,取得目標企業的聲譽量表,並使用情感分析對媒體所發表的相關企業評論,從中取得所有情感變數的分數,做為後續分析的資料來源。 最後,使用複迴歸分析與倒傳遞類神經網路,兩個預測方法,從情感分數的變數中,找出影響企業聲譽的顯著變數。從結果看來,複迴歸分析所得到的均方差(MSE)在7點尺度上為1.161,均方根差(RMSE)為1.077,倒傳遞類神經網路中所得到的均方差(MSE)為0.290,均方根差(RMSE)為0.538,最後利用顯著變數投入倒傳遞類神經網路中均方差(MSE)為0.245,均方根差(RMSE)為0.495,皆有不錯的預測效果,以顯著變數投入倒傳遞類神經網路最佳,顯示從媒體評論中的情感分析是可以推測企業聲譽。 zh_TW
dc.description.abstractNowadays, the enterprise attaches importance to reputation, treats it as an important asset to manage, and the media is an important channel to convey enterprise image. In the past, the research has been used to obtain the reputation of the enterprise, adopt the questionnaire method, obtain the prestige scale, carry on the follow-up analysis, but very few research is from the media comment to speculate the corporate reputation, the media has been one of the sources that the stakeholders obtain the enterprise information and the media coverage of the view will affect the direction of corporate reputation. Sentiment analysis can parse the content of an article or comment, establish a person′s opinion or attitude towards something, the past research, judge the positive, neutral or negative, but people′s emotions are complex and changeable, people through the expression of emotion is also, it′s more helpful for the study to find out the score of affective variables in the paper through the Affective Dictionary. It is easier to get these media comment sets under the push wave of the internet than in the past, by distributing questionnaires to the people who work in the media and obtaining the reputation scale of the target enterprises and use the emotional analysis of the media published by the relevant corporate comments, from which to obtain all the emotional variables of the score, as a follow-up analysis of the data source. Finally, using the multiple regression analysis and the BPNN, two prediction methods, from the effective fraction variables, to identify the significant variables affecting the corporate reputation. From the results, at the 7-point scale, The MSE of multiple regression analysis was 1.161, RMSE 1.077, the MSE in the BPNN is 0.290, RMSE 0.538, and finally using the significant variable into the BPNN, MSE is 0.245, RMSE is 0.495, all of them have good predictive effect, and the outstanding variables are put into the inverted transfer class neural network best, display from the media The emotional analysis in the comments can be inferred from the corporate reputation.en_US
DC.subject企業聲譽zh_TW
DC.subject情感分析zh_TW
DC.subject複迴歸分析zh_TW
DC.subject倒傳遞類神經網路zh_TW
DC.subjectJieba中文斷詞zh_TW
DC.subjectCorporate Reputationen_US
DC.subjectSentiment Analysisen_US
DC.subjectMultiple Regression Analysisen_US
DC.subjectBack-Propagation Neural Networken_US
DC.subjectJiebaen_US
DC.title應用情感分析從媒體評論推測企業聲譽之研究zh_TW
dc.language.isozh-TWzh-TW
DC.titleApplying Sentiment Analysis to Study Corporate Reputation Estimated from Media Commentsen_US
DC.type博碩士論文zh_TW
DC.typethesisen_US
DC.publisherNational Central Universityen_US

若有論文相關問題,請聯絡國立中央大學圖書館推廣服務組 TEL:(03)422-7151轉57407,或E-mail聯絡  - 隱私權政策聲明