微網誌短句的情感指數分析－以新浪微博為例

Estimating Emotion Index of Short Sentences in a Microblog Website-Taking Weibo.com as an Example

楊亨利(Heng-Li Yang) ； 林青峰(Qing-Feng Lin)

《資訊管理學報》 24卷1期 (2017/01) Pp. 1-28

摘要

隨著個人網誌與社群網路的發展，從個人社群網誌去分析發言資料、互動記錄、交友狀況等最後找出可用的規則，已成為熱門的分析應用。本研究經由分析作者在微網誌發表的狀態文句，希望除了能找出作者的正/負面意見傾向外，更進一步能瞭解作者撰文時可能蘊含的情緒。我們提出一個新的方法，以大陸的新浪微博為例，首先利用演化策略的方法，我們可以建立對微網誌作者正向情緒分類器與負向情緒分類器。若有需要，正負向亦可區分為非常正/非常負向、正/負向兩類別。實驗結果顯示，我們分類的效果在精準率、召回率、F1 分數均達令人滿意水準。其次，我們開發了能找出作者的情感指數推估系統；該系統利用迴歸方法可經由分析作者在其微網誌上輸入的狀態文句，推估作者想表達的心情，給予一個幸福指數；其他的情感（如：喜樂、憤怒、悲傷、厭噁、恐懼）指數也能類似地建立。

關鍵字

微網誌 ； 情緒分析 ； 意見分析 ； 情感指數 ； 演化策略

並列摘要

Purpose- This study aims to propose an approach for mining positive/negative opinions and estimating an emotion index of sentences in microblog website. Design/methodology/approach- After reviewing the related literatures, we proposed an ontology-based approach by using ConceptNet and evolution strategic for mining positive/negative opinions from short sentences posted in a microblog, Weibo.com. Applying regression analysis, we also built a prototype system to estimate its implied emotion. Findings- Using the experiment data, we can build a positive classifier to provide positive sentiment cluster and negative classifier to provide negative sentiment cluster with five or three scales. The levels of precision and recall rates, and F1 scores for those classifiers are satisfactory. In addition, our system can give an index of happiness. Research limitations/implications- The future study can collect more sentences for testing and try other micro-blog or regular blog sites. The efficiency can be also further enhanced. Practical implications- Practically, businesses can apply our proposed approach to understand the emotion of the customers after purchasing their products/services. Social workers or police departments might identify persons with suicidal potentials at the early stage from the web. Originality/value- The academic contribution is to propose a new approach to discover possible emotion.

並列關鍵字

microblog ； emotion mining ； opinion mining ； index of emotion

參考文獻

蕭瑞祥、姜青山、曹金豐、陳柏翰(2015)。基於中文語法規則的情感評價單元抽取方法之研究。中華民國資訊管理學報。22(3)，243-272。

賴正育、楊亨利(2012)。微網誌使用的需求動機及其影響。中華民國資訊管理學報。19(1)，81-103。

Beyer, H.G.,Schwefel, H.P.(2002).Evolution strategies-a comprehensive introduction.Natural Computing.1(1),3-52.

Chaovalit, P.,Zhou, L.(2005).Movie review mining: a comparison between supervised and unsupervised classification approaches.Proceedings of The 38th Hawaii International Conference on System Sciences(HICSS'05).(Proceedings of The 38th Hawaii International Conference on System Sciences(HICSS'05)).:

Dave, K.,Lawrence, S.,Pennock, D.M.(2003).Mining the peanut gallery: opinion extraction and semantic classification of product reviews.Proceedings of International Conference of World Wide Web.(Proceedings of International Conference of World Wide Web).: