Carlo Bernardini (Development)
Faculty of Sciences, VU
Amsterdam, NL
E: 1934481@student.vu.nl
Hüsnü Taş (Statistics)
Faculty of Sciences, UvA
Amsterdam, NL
E: 2525616@student.vu.nl
Deepak Soekhoe (Mining)
Faculty of Sciences, VU
Amsterdam, NL
E: 1997424@student.vu.nl
 
The current data set holds 0 tweets that matched 0 tracked parties

Mention Count

This visualization displays how often individual political parties were mentioned within the collected data sample of tweets. This visualization does not take into consideration sentiment classifications.

Sentiment Analysis
API provided by Ai Applied API provided by Ai Applied

Which sentiment do people express towards individual parties on twitter? These donut charts visualize the distribution of positive and negative sentiments towards individual parties on the basis of collected tweet texts. For this analysis we used a sentiment analysis API from Ai Applied, who kindly allowed us access to execute an increased amount of API calls.
Tweets mentioning multiple parties were filtered out of the analyzed set since the judgement concerns the entire tweet text. The confidence threshold was set to >=0.75 — meaning tweets with a confidence grade below 0.75 are ignored — to enhance the reliability of the judgements.

Final Debate Timeline

For this visualization we mined tweets during the final electoral debate on Tuesday March 18th. We collected a set of 6717 tweets when tracking #PS15, the official hashtag for the debate according to the NOS . From these tweets, we selected the ones mentioning one or more hashtags that we assigned to each party beforehand. Then, we split up the tweets in timeframes of 5 minutes from the beginning (20:30) until the end (21:55) of the debate.

 
 
Click on a node in the timeline above to seek to the corresponding moment in the video below.

Term Cloud
API provided by Ai Applied API provided by Ai Applied

In addition, we also performed a text analysis of all tweets, extracting the main concepts for each tweet and summarizing them according to their frequency and impact. Each concept label was assigned a certain weight, which is taken into account in the summary. Again, an API by Ai Applied was used for this analysis.