The Dutch parliamentary elections took place between 15 and 17th March. These were the first elections since the outbreak of Covid-19. Taking into account this precedent, we used PublicSonar to monitor election safety. Our goal was to detect early warnings for unsafe situations via sentiment analysis and publicly available data. Such information is valuable for ensuring that each citizen can exercise their right to vote safely and responsibly. To do that, it is crucial for local authorities, such as municipalities, emergency services and local resilience forums, to detect the earliest signs of emerging risks.
Setting up search queries
To begin, we set up searches for over 200 election-related terms, locations of election polls across the country, as well as the list of parties and running candidates. PublicSonar retrieves publicly available data from multiple sources, including social media, blogs, news websites and RSS feeds. Data collection began two weeks prior to the elections.
Then, we refined collected data based on different possible calamities, including fire, violation of the rule of law, crowded places, public safety, aggression and Covid safety measures. PublicSonar allows for easy monitoring of retrieved visuals and location-based messages. This eased in creating real time situational awareness.
Additionally, we deployed our sentiment analysis. Our artificial intelligence automatically detects messages with emotions, so it allowed for saving significant hours of manual work. Moreover, incoming data was previewed into a positive and negative sentiment score, which made analysis really easy.
Preparedness: Campaign phase
The first phase we identified was during the campaigning. PublicSonar detected numerous messages about vandalism over election posters and even streets at different locations across the country.
These messages matched the sentiment analysis with a negativity score of 47%, implying that over the total collected messages almost half contained a negative emotion. Citizen negative emotions were predominantly angry, hateful or confused.
Additionally, we detected early messages around a blasting fire in Den Bosch. This information is valuable to emergency responders in the area. In order to estimate the scope and magnitude of the damage. Luckily, citizen messages came in with visuals, which made the creation of a situational picture easier.
Situational awareness: Election phase
The second phase identified was the election phase, once citizens were allowed to send in their votes by post or in stations. During this phase it was most relevant to monitor potential calamities and the proper application of Corona safety measures.
We evaluated the situation at multiple local voting locations, according to messages retrieved by PublicSonar. The voting took place without severe disturbances. At places where it was crowded, the 1,5 meter distance was respected regardless of long queues at some places.
Nevertheless, local crowded situations were detected in Bussum, where the Covid safety measures were put at risk.
Exit polls: After election.
The last phase of the elections was when polls closed and vote counting began. Due to the unprecedented Corona circumstances, elderly people were allowed to vote by post. This did not come without complications. Around 8% of the votes were deemed invalid, as citizens did not follow the instructions correctly. The situation posed the risk of creating a negative social reaction. Subsequently, authorities found a solution that decreased the number of invalid votes. During the course of this process sentiment and response could be easily monitored with PublicSonar.
Safe elections during Covid-19
The value of insights from publicly available data and real time situational awareness contributed to local authorities in the Netherlands for ensuring safe elections during Covid times. Is your organisation involved in local elections? Do you need the earliest signals of emerging risks? Are you interested in the safeguarding of health measures? Learn how publicly available data and sentiment analysis can help you at email@example.com