Skip to main content

Content Personalisation Network

Article Category

Article available in the folowing languages:

Innovative algorithms for transparent content personalisation

In the saturated market of information, news curation has grown in importance as the only viable way to avoid overdose and keep users interested. American publishers benefit from a head start, but a new app created with the European market in mind hopes to reverse the trend.

Digital Economy

Europe being the sum of its different cultures and languages, it comes as no surprise that it has inherited a very diverse media industry. The diversity adds richness but it also leads to fragmentation which makes the creation of personalised content more difficult. So the question is: How can a fragmented market with small economies of scale, a slower update of cutting-edge technology than its American counterpart, and consumers already overwhelmed with content, effectively transition to personalised content? “In the context of dominance by US-based global players who do not suffer from these problems, the CPN project has developed a new approach to personalisation of digital content,” explains Mike Matton, head of international R&D collaborations at VRT Innovation. “CPN incorporates new technologies and adapts them to European needs, with more focus on ethics, trust and diversity.” CPN’s motto is to ‘deliver the right information, at the right time, in the right context’. From a technological point of view, this translated into the CPN News Recommender Engine. Publishers wishing to personalise content for their users can connect to the CPN platform, use its A/B testing module to identify the best configuration for their needs, and start feeding content recommendations to users accordingly. To make this possible, CPN proposes ‘à la carte’ recommenders that can be tweaked individually and combined with each other, or with existing pieces of software. These include content-based recommendations based on keyword extraction, filtering techniques based on historic consumption, trending items, random recommendations, sentiment-based suggestions, and even a composite recommender that combines various techniques. “The main innovations lie first in the flexibility brought by these multiple microservices, but also in the human-centric focus of user research and the transparency of the recommendation algorithms,” Matton adds. “If you compare them to those of US-based companies like Google, they are less complex and more transparent. We put real effort into tuning them to have a positive effect on society as a whole, instead of just increasing company profits. The value here is how informed users are, rather than just how much time they spend using the app.”

Users come first

CPN is currently in its third and final pilot testing phase. Users have been involved since the early stages of development and have already helped improve the system considerably. Among the most important outcomes of the pilots, Matton cites users’ interest in knowing exactly why a news item is being recommended to them, as well as the importance of user interface. The latter has been largely improved since its first version, and the team now says user feedback is very promising. Whether the system will be widely adopted remains to be seen, but CPN certainly brings something new to the table that will speak to European publishers and consumers alike. The former will benefit from easy-to-implement algorithms and GDPR compliance, while the latter will enjoy news curating in a transparent manner. Everyone stands to gain. CPN was due for completion at the end of April, and its final phases were dedicated to the analysis of usage data, surveys and final conclusions. “We hope that our work will help European media companies to fill the technological gap between themselves and large international companies,” Matton concludes. “We also hope that this can happen in an ethical way, and that we can become an example for the rest of the world to follow.”

Keywords

CPN, content personalisation, app, software, algorithms, filtering

Discover other articles in the same domain of application