News is essential to a functioning society. It keeps people informed on important subject matters and enables them to formulate an opinion on those topics. Thus, the information people consume can be immensely influential. While this puts the onus on media outlets to establish trust and credibility, these values are only part of the equation when it comes to discerning real news from fake news.
Unfortunately, there is no cure for fake news. There is no vaccine. There is only a commitment to understanding the news, or what masquerades as news, for what it truly is. Such a commitment is a monumental responsibility and one few have the time or the resources to dedicate toward.
Bywire News has made it their mission to combat fake news and have taken a unique approach to doing so via artificial intelligence. We discussed that approach in our NLP Stream with Bywire News CEO Michael O’Sullivan and CTO Jetze Sikkema. Here is what we learned:
The Keys to Identifying Fake News
One of the biggest concerns with fake news is how widespread it has become. Not only are there a countless number of news sources disseminating information but an expansive number of media platforms with which they can distribute their information. The key to keeping up with this excessive amount of content is finding a solution that can handle this level of scale.
Bywire News built the first component of their solution by identifying five core pieces of information. This was enabled by the expert.ai NL API, which helped to assess the following:
To begin, you analyze an article to identify the main lemmas, or keywords, within it. This helps you to establish the core elements a real news story on the subject should include.
The next layer the algorithm considers is the entities within an article. While elements provide an understanding of an article’s subject matter, entities reveal the people, places and things that are most important to the story.
While the first two layers help you to arrive at the core subject matter of an article, sentiment helps to identify its intent. Where real news is meant to inform, fake news is meant to persuade. By measuring the sentiment score of an article, you can determine whether the language used skews heavily in a positive or negative direction, either of which would indicate an attempt to influence.
A common theme found in content that intends to influence readers is repetition. Where real news tends to flow smoothly and lay out the facts, fake news uses repetition of words or phrases to embed them into the minds of readers. This is known as the Illusory Truth Effect and is a staple of persuasive language.
The final layer of this algorithm leverages the lexical complexity of an article or, more simply, the quality of writing. A well-written article does not ensure a genuine piece of news, but a poorly written article is a clear red flag that an article may be fake. This, of course, also depends on the intended audience and the age of those in it.
These five factors all delve into the language within an article and provide a solid foundation for distinguishing real news from fake news. It does not, however, tell the entire story. To build upon this foundation, Bywire News uses its own proprietary algorithm to examine important factors such as:
- Who are the sources of information for the article? Does the author use external sources or only link to internal content?
- How reputable is the author of the article? Are they known for spreading false information? Do they typically write in a heavily positive or negative voice?
- Is the media used in an article heavily edited? Was it altered from the original?
Bywire uses these factors and more to generate a score, or Trust Index, that determines how genuine a piece of news is. They then use that score to ensure the content on their website is real and can be trusted.
Consumers should be confident in the news they read. If they cannot trust networks and reporters to share honest information, it is critical that they have something to build that trust for them. Bywire News is a great example of this type of solution. They show how natural language technology can curtail the power of fake news. Hopefully, this is only the beginning.