印尼文情緒偵測於不均衡微網誌資料之研究

Emotion Detection for Unbalanced Indonesian Tweets

方麗娜(Laina Farsiah)

指導教授 ： 陳宜欣

國立清華大學/電機資訊學院/資訊工程學系/碩士(2015年)

摘要

印尼文情緒偵測於不均衡微網誌資料之研究

近年來，推特資料勘探已成爲研究熱點。而在微網誌上的情緒分析，是眾多研究中的其中一種。最近，相關學者提出了一種基於圖學的情緒模式擷取技術，該技術在多種語言的應用中皆取得良好效果。本研究旨在提升印尼文推特情緒分析的精確度，分析的情緒包括以下八類：開心(senang)、憂傷(sedih)、害怕(takut)、驚訝(terkejut)、噁心(jijik)、希望(antisipasi)、信任(percaya)、生氣(marah)。之前的研究中，印尼文的情緒分析精確度不甚理想，主要原因爲印尼文推特中情緒分佈不均衡。因此本研究提出一種調整情緒模式權重的方法以解決情緒分佈不均衡的問題。實驗結果證明，該方法可（顯著）提高印尼文推特中情緒分析的精確度。

關鍵字

情緒偵測 ； 不均衝資料 ； 維特 ； 印尼文

並列摘要

Emotion Detection for Unbalanced Indonesian Tweets
ABSTRACT
Research concerning Twitter mining becomes an interesting research topic in recent years. Emotion
detection is one of research area which uses microblog, such as Twitter, to discover emotions from
textual data. Recently, a novel technique based on graph-based was proposed to extract patterns that
bear emotion. The system has been achieved a good performance in different languages. By adopting
the system, we are motivated to enhance the accuracy of emotion detection for Indonesian language
which consists of eight emotions, i.e. joy (senang), sad (sedih), fear (takut), surprise (terkejut),
disgust (jijik), anticipation (antisipasi), trust (percaya), dan anger (marah). The data distribution
among the emotions is really unbalanced which make the low precision of system for Indonesian
language. In this study, we proposed an adjusting pattern weight to address unbalanced data problem
for Indonesian language. The experiment results show that the proposed approach can improve the
precision for unbalanced Indonesian data.

並列關鍵字

Emotion Detection ； Unbalanced Data ； Twitter ； Indonesian Language

參考文獻

driguez Perez, and J. M. Jose, An interactive interface for visualizing events

Research & Development in Information Retrieval, SIGIR '14, (New York,

Detection and Earthquake Reporting System Development," IEEE Transactions

elections with twitter: What 140 characters reveal about political sentiment.,"

[4] M. Hu and B. Liu, Mining and summarizing customer reviews," in Proceedings