How we select and rank content in the Brave Today news feed

In December 2020, we launched a new service in the Brave browser – the Brave Today news feed (aka Brave News). It collects information from sources verified by us and distributes it into 15 categories, among which each user can find something interesting for himself. To view the news in Brave Today, you need to open a new tab in the Brave browser and scroll down the page. In this article, we will talk about how we select and rank content.

In future releases, users will be able to independently add RSS feeds of interest to them. So far, the simplest and most understandable solution for everyone has been implemented – we simply selected RSS feeds from leading publishers from Comscore, Alexa and Feedly lists, and focused on sources with financial support – either they have advertisers, or paid subscriptions, or both, and other. In addition, these should have been sources with stable and valid RSS feeds – according to this criterion, several large publishers did not pass the selection.

As a result, our list includes about 300 sources. This amount allows us to control the amount of data that our private (like everything else) content delivery network transmits to the browser. In addition, our users appreciate the speed of Brave, and we cannot sacrifice it for the sake of the news feed. Brave Today should not reduce browser performance, so we will roll out the expanded capabilities for adding new sources and customization only after we test our wonderful aggregator and get enough feedback.

To show the user only the freshest and most interesting news, we have come up with our own ranking algorithm.

Naturally, the main catch in the implementation is maintaining the privacy properties of our products. Obviously, the content should be collected by the browser itself, while leaving no identifiers or other traces available to anyone, including us. What content users viewed, where they clicked, which sources turned on or off – what happened in the browser remains in the browser. Therefore, we need a ranking model that can work locally on the client and does not require data transfer anywhere from the browser.

In the first version of our feed, we match the selected sources with the history of page visits in the browser. In the future, we will use a model similar to the Brave Ads mechanism, which takes into account the amount of time and user activity on the pages visited (again, strictly on the client).

In simple terms, the current ranking method looks like this:

  1. Sorting the scores in reverse order

  2. Each news item at the initial stage has 0 points.

  3. On the server side, each news item receives a score equal to the logarithm of the news age in seconds. These points are responsible for the freshness of the content.

  4. On the browser side, each news item gets -5 points for crossing the source with the list of recently visited sites by the user. This is how we make the feed more interesting for you.

To that we add a little local randomization to keep the feed from being predictable and suggesting new and unexpected sources.

Example: You recently read something about the uprising of the machines. A day ago, one of our sources published a news about artificial intelligence. On the server, she gets 11 points (ln (86400 seconds) = 11.37). In the browser -5 points, a total of 6. And the news about street protests from another source, published 12 hours ago, will receive 10 points, because the browser does not think the user is interested in this.

Of course, different people have different tastes, so the sources can be customized. Brave Today lets you turn any of these off or on. The feed also contains news from us and our partners, but they can also be removed. In addition, if the news feed distracts you too much, you can turn it off completely. When we tested Brave Today, many of our team found out more than once that a good part of the working day was lost due to watching the feed, so one of the ideas for the premium version is to set the schedule for the appearance of the news feed. It will also be possible to disable promo blocks in the premium product.

To support authors and publishers, we do not collect news in a separate feed. You read them directly on the sites where they are published, in the form in which they were conceived.

For now, the number of sources is limited, but in future releases we will configure the ability to add them manually. RSS is an essential information flow management tool that allows you to breathe fresh air outside the information bubble of media feeds.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *