Tilt Insights
Insights enables you to limit your bias and respond to online manipulation effectively. With insights we process large amounts of data and derive patterns into highly coordinated communities, narratives and quantifiable markers of disinformation techniques.
This month’s update include: new scrapers, network analyses, sentiment analysis and hate speech classification.
We are happy to announce that we have implemented a new version of our Tilt Insights Platform this month. This enables us to scrape near real-time data from Twitter, Instagram, YouTube, and Reddit, which broadens our view of online manipulation and how narratives spread across multiple platforms.
We have also added new machine learning functionality to the platform which enables us to perform network analyses. These models focus on how information spreads through a network of accounts without analyzing the content itself. We detect signs of anomalous coordinated (inauthentic) behavior indicative of online manipulation. Such activity is then flagged for further investigation by analysts.
Finally, we have added natural language classification models that detect sentiment as well as hateful content. These models help our analysts to filter on specific data to look at in more detail, and they can also be used to compute statistics such as a ‘hatefulness index’ for an entire dataset. In this way, the overall degree of hateful content (or some other metric of interest) can be compared between different topics, or for a specific topic over time.
This new version of the Insights Platform empowers analysts to provide a deeper understanding of online manipulation.
"*" indicates required fields