Are there more to the UK news than rugby and the royal family?
In a world where millions of articles are posted online each day, it is impossible to manually keep track of the complete news space.
At Adlede, we want to cover as much as possible of that news space but our own personal media bubbles may give us a skewed view of the full content.
To get an overview of the news space, we use the latest techniques in machine learning to find topics. Let us have a look at what the top news sites in the UK are writing about in this demo:
Try out an interactive version: https://adlede.com/media/uk_topics.html
This grouping of articles and keywords is called Topic Modelling. It works by converting news articles from plain text into numerical vectors. These articles then have a mathematical relation to each other, meaning that two articles that talk about the same topic, are close together in a vector space.
For example, two sports articles are closer together than a sport and an economy article. We can then find clusters of articles closer together and label them as a topic. The topic keywords come from finding what words inside the cluster are more common than outside it.
This is useful for us not only when entering new markets, like the UK, but also when our customers are looking for the right contexts for their campaigns.
By having an initial idea of where to target a campaign we can easily find clusters of closely related topics and create better categories, catching more of the content.
By Anton Eklund, PhD student at Adlede