Using Machine Learning to Monitor Social Media Crises at Data Summit 2018

Machine learning can be applied to sentiment analysis of unstructured data in the context of social media. As more and more people tap into social media tools to voice positive and negative opinions, it’s important to predict a social media “crisis” before one occurs.

Jana Mitkovska, project manager, Raytion, and Christian Puzicha, senior solutions architect, Raytion GmbH, presented their session, “Self-Learning Social Media Systems Through Machine Learning & Data Science” on Wednesday, May 23, 2018 during Data Summit 2018.

Data Summit 2018 is taking place at the Hyatt Regency Boston, May 22-23, with pre-conference workshops on Monday, May 21. Cognitive Computing Summit will also be co-located at the event.

Mitkovska and Puzicha used the Kendall Jenner Pepsi ad scandal to illustrate how brands should handle social media monitoring.

By creating a sentiment model to do sentiment analysis companies with social media teams can control how to respond to an outpouring of positive or negative comments and monitor them.

To create a successful social media monitoring solution using machine learning companies must create a solution that adapts to new emojis, abbreviations, and hashtags. To predict and determine whether voices on social media are positive or negative social media teams need to define indicators and dashboards with alerts that define which parameters to monitor.

“If you have social media team you must take action: respond or release PR,” Mitkovska said. “It depends on your company strategy.”

Indicators that a crisis is upon a brand include:

  • Spiking in usage in hashtag
  • Spike in tweets around certain topic
  • Increase in Retweets from certain individuals
  • Appearance of new hashtag
  • Social media influencers become involved

During the outcry of outrage against Pepsi’s Kendall Jenner ad, it took Pepsi 3 hours to remove the ad and issue an apology. That didn’t stop a reaction, however.

Actions that companies can immediately take include alerting and notifying team the social media team as quickly as possible, whether through email, sms, or an automatic phone call.

If there’s and increase positive reaction the product owner should be alerted but if there’s an increase in negative reaction, the PR team should be alerted. If an influencer is involved the marketing team becomes alerted.

The focus of the social media team should be to review reactions, replies, and anomalies.  Then determine a direct response based on company strategies.

Data Summit 2019, presented by DBTA and Big Data Quarterly, is tentatively scheduled for May 21-22, 2019, at the Hyatt Regency Boston with pre-conference workshops on May 20.

Many presentations from Data Summit 2018 have been made available for review at