CONCLUSIONS AND DISCUSSIONS
5.3. Research Contributions
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
results (salient concept analysis) in this study suggested a more correlated relationship between the news media and political actors, it is possible that politics related concerns blur the focus of the referendum issue. Although the the whole semantic network comparison suggested the more correlated relationship between news media were and the environmental groups, from the result of top-ten salient concepts in RQ1, it revealed that several concepts news media put emphaisis on were more similar to what the DPP and KMT actors focused on. The relationship between the referendum and the influences of political concerns due to the connection with an election is suggested in future studies, through using different ways of analysis.
5.3. Research Contributions
This research added contributions upon theoretical, practical and methodological and aspects.
Based on the theories and past studies, this research proposed to fill up the research gap, which advanced the previous theoretical or conceptual insights, practical research procedure, and methodological explorations and extensions.
Theoretically, this research made efforts on investigating classical theory or concepts and adding insights through analysis. First, this research contributed on media slant research through text-based analysis and the measurement with cross-media contents. As media slant seemed not attributed to a theorist, the core definition, measurement, and impact of media slant were still diverged (D'Alessio & Allen, 2000; Lichter, 2017), this research explored and proposed to use the social media contents as parameters to measure the slant, which enriched the insights and applications on concepts of “partisan slant” or “media slant”. Second, this research reviewed literatures of framing and semantic network analysis, and proposed to
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
explore framing through semantic network analysis, using text-based salient concepts to depict the framing of Shenao power plant issue.
Third, this research investigated Taiwan’s environmental referendum issue, and added insights on the past studies of politics framing in environmental issues in news media. Under the context of two-system party politics, this research reviewed the politics struggle and conflict between the two major political parties (i.e., KMT and DPP), and applied Shenao’s issue as the case for partisan slant, framing and semantic network analysis. As past studies discussed the news frame of political conflict or strategy in environmental issues (Bolsen &
Shapiro, 2018; Nisbet, 2010; O’Neill, Williams, Kurz, Wiersma & Boykoff, 2015), through Shenao power plant issue, this research also examined the politics or strategy framing upon environmental issues in local news media. From the results of the analysis, it could trigger more discussion about whether a referendum should be tied up with an election. The relationship between the referendum and the influences of political concerns is worth more investigations in furure studies.
Forth, this research compared the salient concepts regarding the Shenao power plant issue between the digital native and traditional news media. Due to the emerge of web environment that gave birth to digital-native news media, this study proposed to compare different types of news media on semantic analysis and partisan slant exploration. Although the results did not suggest salient differences between two types of media in this research, the applications and explorations added the research directions on the two types of news media for future studies.
Methodologically, this research combined the applications of semantic network analysis, partisan slant research, and framing analysis to investigate the salient concepts among different
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
types of media (i.e., social media and news media), and to explore news media’s partisan slant.
The semantic network analysis was the core methodology in this study, through UCINET 6 to conduct semantic network analysis on salient concepts, Gephi to visualize the network layouts, and QAP correlation analysis to measure partisan slant. Using these tools extended the applications on word-based semantic networks, as these three tools played the different roles upon the semantic network analysis.
At the issue level, this study provided valuable additions on Taiwan’s journalistic research on 2018 referendum issues through semantic network analysis. Due to the two-party system and political turf wars in Taiwan, news about referendum issues was easily laden with political conflicts rather than issue-based focus. On top of this, this research proposed to used the semantic lens to enrich the understanding of Taiwan referendum news research. It used the software of UCINET 6 to extracted salient keyworks through the metric based on the node’s influence in the whole network, for salient concept analysis. Besides, it used the Gephi to graph the layout of netwoks for storytelling.
Second, it extended the use of the semantic correlation analysis on partisan slant research. The QAP analysis on UCINET 6 measured the partisan slant of the news media.
Based on the past studies, Jiang et al. (2016) used QAP to compare the semantic networks between Associated Press and Xinhua News Agency in the event of Arab Spring protect. For the cross-media comparison through text-based content analysis, the practice and procedure of QAP analysis would provide valuable additions for future studies. It not only provided another way to explore and measure media slant, upon the text-based level, but also explored the possibility of doing quantitative content analysis, and framing analysis.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Third, this research extended the application of intermedia contents. Through using the semantic way, this research investigated possibilities of using intermedia contents to measure partisan slant. Past research (e.g., Gentzkow & Shapiro, 2010) used different media contents as an index or parameter to measure news media’s partisan leaning, while this study explored to use Facebook contents as an index to measure news media’s partisan leaning. Additionally, this research explored different framing and salient issues between news media and social media on the text level.
Forth, as for the data cleaning procedure, this research enriched the applications of the computer assisted softwares to process data analysis. This rearch processed data cleaning before the analysis through R programming language. The word segmentation was conducted by JiebaR package (Qin & Wu, 2016), and the co-word matrix was processed by Quanteda package (Benoit et al., 2019). It contributed to the practices of data cleaning on text mining and text analysis, through application of R programming langue and the packages for future studies.
The utilizing of JiebaR package on Traditional Chinese natural language processing would be a valuable application as it was first launched for Simplified Chinese language as a word segmentation tool.