There are always some star candidates during the election, who attract media attention, and Taiwan’s elections at the end of 2018 are no exception. People might doubt whether some media have more reports on specific candidates or even biases in the report due to the political inclination. We explore this phenomenon through the data on news reports.
We grabbed daily news reports in Taiwan and used keywords to extract specific politicians and observe the changes in quantity. And pass the content to Google\’s artificial intelligence analysis to see whether the media reports are neutral or not. This project continues until the next presidential and legislative elections in 2020.
What makes this project innovative?
This project is not only the first report in Taiwan which use artificial intelligence to analyze the content of the report, but also the first journalism team explores media differences in the number of reports for specific candidates by data. The project will be continuously updated, which enable readers to observe the performance of these politicians in the media.
What was the impact of your project? How did you measure it?
Recently, Taiwanese readers have expressed displeased toward the media, which over-report the same politician and even launched the movement of "switch the TV channel", asking restaurants, schools and other places with TV to stop playing the same news channel. But it is just a personal perspective, whether the TV channel is over-reported toward the same politician. And that’s the reason why we use data to objectively report this perspective.
This is the first time that the reports can precisely confirm this situation with actual evidence, data. We publish and visualize the data, which we collect every day. The project will be continuously updated. The analysis at that time is no longer groundless but based on evidence with data.
Source and methodology
● Data source： - The news data was collected from multiple news websites, crawled by g0v’s “people-in-news” project. ● Methodology： - We manually pick the keywords according to the name of the election candidates and the trending of news. And then, we use the Google NL sentiment analysis API to analysis the news that contains the keywords and get sentiment score for each such news. After the scores were regularized, we group the news data by its publish date, the keyword it has and the media source the news comes from.
● Forent-end: - This project is made by Vue.js and official router library to implement a single page application. In this project, user can navigating between two features and exploring corresponding chart: 1. A line chart showing the trend of news report’s count by a single keyword which was picked from our available keywords. 2. A stacked area chart showing the trend of news report’s count by two of the keywords which were picked from our available keywords. We use D3.js as our charting library.
Producer: Chien Hsin-chan Journalist: Lee Yu Ju Design: Chen Yi-Chian Web-developing: Hsiung Kai Wen Data:Chen Yen-Yu, Lee Yu Ju, Li Chao Yen