Page 123 - SUSTAINABILITY ISSUES & COVID-19

Keyword, hashtag, mentions or replies Original Tweets
Data Collection Pre-processing

Indexing Positive, negative,
Sentiment Sentiment or neutral Tweets Sentiment
Index Analysis
Analysis

Figure 3
General Process of Sentiment Analysis

After twitter data is collected, sentiment analysis is applied to measure citizens’ opinions (Castelló, Etter, & Årup Nielsen,
2016). This method is based on natural language processing techniques that measure sentences’ affective orientation
toward an object (Bo Pang & Lillian Lee, 2008). The result of sentiment analysis is categorized into three categories,
positive, negative, or neutral tweets.

The result ranged between infinitive (absolute negative) to zero (absolute positive), as shown in Figure 4. If positive
tweets equal negative tweets, the sentiment index would be neutral. The closer the value to zero, the better sentiment
indicates a better reputation (Etter et al., 2019). However, the range might be confusing because the lower score
indicates better legitimacy than a higher score. Thus, in the data analysis section, sentiment indexes are multiplied by
-1 to simplify the interpretation because now the higher score indicates better legitimacy. Figure 4 shows the range
and categorization of analysis before and after adjustment. After adjustment, the higher absolute index indicates the
higher sentiment score of an organization.

negative neutral positive

Absolute Absolute
negaitive Positive
Before adjustment ∞ 1 0
After adjustment ∞ -1 0

Figure 4
The Range of Sentiment Index

122 International Conference on Sustainability
(5 Sustainability Practitioner Conference)
Th