Introduction: CPN Open Virtual Platform

One major goal of the CPN project is to make sure that our personalisation platform will be 'open' to future third-party integrations. We are doing this by creating a ‘virtual’ platform that can be adapted to the needs of media companies by connecting different services together.

In other words, the CPN platform is not intended to be a physical system that media companies need to install, but an ‘Open Virtual Platform’ – read more about our approach here.

Over the course of the project we foresee three major releases of the CPN Open Virtual platform that include functionalities based on the requirements gathered from the three CPN pilots. Each platform release includes specific functionalities, chosen after a process of evaluation and prioritization of user requirements.

This process followed a Scrum agile methodology, in which the CPN media partners (VRT, DIAS and DW) assumed the role of Product Owner and the technical partners participated as Scrum Team.

The second version of the CPN platform described below is a more mature release and improves the reliability, safety and performance of the platform. In fact, the core components were improved and extended and integration tests were conducted in order to demonstrate the platform’s flexibility.

In addition, this platform release focused specifically on the recommender system configuration and personalization, featuring a new A/B Testing Component that provides media companies with an easy way to test the recommender engine.

Core Components and Technology Bricks

The CPN Open Virtual Platform v2 includes:

  • A toolbox with 10 Technology Bricks deployed as microservices (9 backend services + 1 web dashboard)

  • 1 Technology Brick as client application, in web and mobile version (Reader’s App)

  • 3 Core components

Screen Shot 2019-07-09 at 15.49.20.png

Open Stack: The CPN platform and its core components are based on an open source stack that ensures a high level of expandability and flexibility.

Screen Shot 2019-07-09 at 15.51.04.png
  • Docker is a container platform provider and the de facto standard for containerization. It provides tooling and a platform to manage the lifecycle of a container.

  • Rancher is an open source container management platform that allows for running and managing Docker containers in production.

  • Express Gateway is a Microservices API gateway built on Express.js.

  • A message Broker, based on Apache Kafka and Zookeeper was installed and configured to allow an async communication within the platform.

The CPN platform tools, called Technology Bricks, are divided into 4 main categories:

  • Content analysis and retrieval

  • User Profiling

  • Apps for end users

  • Apps for producers

Screen Shot 2019-07-09 at 15.53.50.png

The CPN platform provides media companies with many services and functionalities that improve the production and delivery of media content to their target audience. Thanks to the flexibility of the system, a media company that participates in the CPN ecosystem can decide to exploit parts or all the functionalities provided by the platform.

How to integrate CPN

In order to evaluate the flexibility of the CPN platform v2, we conducted several integration testing activities. In particular, we worked in collaboration with two media partners of the consortium (VRT and DW), leading to the integration of two other client applications: a new mobile website and a smart TV app. In parallel, we are collaborating with three major European media organisations interested in the functionality offered by the CPN platform. (We are also looking for more media companies to test the tool – read more here.)

The process of integration for media companies is an easy, 3-step process:

  1. Article injection: The articles from media company datasource are injected into the CPN platform

  2. Retrieving of recommendations: The CPN platform exposes a service to retrieve recommendations through the API gateway

  3. Collection of user actions: the client app of the media company has to provide user actions to the CPN platform. The API gateway provides an API for this purpose.

An optional step is the recommender configuration, which lets a media company experiment with the recommendation engine by themselves, allowing them to test it in a deeper way.

Conclusions

The CPN platform has now reached a more mature status in its second version, and is enhanced by new functionalities provided by the technology bricks and by processes that improve reliability, safety and performance. The core components were improved and extended and integration tests were conducted in order to demonstrate the platform’s potentiality. Now, CPN is ready for testing by external media companies.

Highlighting the startups working with CPN

Screen Shot 2019-06-17 at 15.38.35.png

Discover the four startups that are developing innovative services that will expand the CPN platform.

After the CPN recommender was presented in February at the CPN Hackathon in London, a group of startups started working on new services that will enhance and add to the existing features of the CPN platform.

The four companies presented their solutions at the personalisation session in June during the World News Media Congress in Glasgow. Check out a video recap from the event below, and scroll down for more information about the startups and the solutions they are creating for CPN!


Kensai

What does your company do?

Kensai is artificial intelligence that understands the narrative from omni-channel data streams. We hyper-scale various AIs in the cloud to achieve fast and accurate sentiment and keyword extraction around brands, products and people.

What is the solution that you are building for CPN?

Kensai is building a solution for newsrooms to monitor the narrative in real-time on Twitter around any news story to personalise it to their audience. In the future, this can also be used to help eliminate the filter bubble by providing contrasting views underneath the article to give a broader picture.

Will+Headshot+1.jpg

Will Crosthwait, CEO / Co-Founder

Twitter: @WillCrosthwait

Email: will (at) kensai.tech

Website: kensai.tech


Loomi

What does your company do?

Loomi.ai is a London-based technology company focusing on creating human-focused artificial intelligence in order to take the next step in the information evolution.

Our mission is to provide simple, intelligent choices for people to manage the endless flow of information that is modern life. We build technology that removes admin complexities and information overload to create space in our customers' lives and minds, so they can achieve more. We believe highly developed personal AI assistants are the solution, assistants that know your preferences, priorities and goals and organise information in one seamless interface.

What is the solution that you are building for CPN?

Loomi.ai is building a personalisation news focused ontology for the CPN project with the main objective of enhancing the accuracy of the personalisation recommender algorithms. The ontology can be used to formalise both the user and her/his interests, and the news content. The developed ontology allows matchmaking between user and content at different levels. The service will have the capacity to help other news providers increase the accuracy of the personalisation they do through either the CPN project or accessing the service directly.

al-ramich-loomi (1).png

Name: Al Ramich, Founder and CEO

Twitter: @LoomiAssistant

Email: al.ramich (at) loomi.ai

Website: www.loomi.ai


U-Hopper

What does your company do?

U-Hopper is an SME based in Trento, a city in northern Italy, among the Alps. We are a deep-tech company developing IoT and BigData analytics solutions with the aim of helping our customers to create value out of their data. We are specialised in big data analytics and data visualisation, business and customer intelligence solutions, chatbots and blockchain applications. Our team is young, international and gender balanced.

What is the solution that you are building for CPN?

Our solution is called Tapoi, a customer intelligence service that builds profiles based on customer online actions, allowing businesses to provide targeted and tailored services, personalising user experience and achieving higher conversion.

Currently, CPN understands the user's interests analysing what the user has read on the platform; Tapoi will provide a new source of information to enrich such profiles, inferring user's interests from their social activities. In this way it is possible to mitigate both the cold start and the filter bubble problems.

pomini_nicolo (1).jpg

Nicolò Pomini, Software Developer

Twitter: @nicolopomini

email: nicolo.pomini (at) u-hopper.com

Website: www.u-hopper.com


Yoop

What does your company do?

Yoop is a data management and intelligence company that uses machine learning, blockchain and big data analytics to empower organisations and customers make better decisions. Spun off research at the University of Nottingham, the company is now looking to commercialise its innovative technology - starting from the news media sector.

What is the solution that you are building for CPN?

Yoop is building a unified user identity for content providers - a simple way for users to reap the benefits of news personalisation without losing control over their personal data. The product, called ID ward, is unique in three ways. First, it creates a much simpler and more attractive experience for user login. Second, it builds much richer user profiles than anything else on the market - this is the sort of data underpinning personalisation, product development, marketing and advertising that big tech amass in huge quantities, but content providers can only dream of. Finally, it gives users unprecedented transparency and unparalleled control over their own personal data, minimising the trade off between privacy and personalisation. ID ward is not tech for tech's sake, far from it: it is tech with a very direct impact on the content providers' product and revenues.

Mattia_Fosci (1).jpg

Contact person: Dr Mattia Fosci

Twitter: @MattiaFosci

email: mattia.fosci (at) yoop.io

Website: www.yoop-tech.com


Winner of the Most Promising Startup award: At the WAN-IFRA Congress, the CPN award for the most promising startup was given to Yoop! The judges said that all companies gave a clear description of what their applications would provide. Yoop stood out thanks to the clarity on how their solution could be applied in media specifically. Moreover, the judges said the company articulated clearly its application’s potential impact on the media industry.

Rebuilding audience’s trust in algorithms: personalisation at the World News Media Congress 2019

IMG_3491+%281%29.jpg

How are newspapers currently approaching content personalisation? Could it open new ways for publishers to build stronger relationships with their audiences? How can tech startups support the news industry in exploring the benefits of personalisation?

These and many other questions were explored at the session “Personalisation of news matters!“ at the World News Media Congress 2019.

‘We want to offer the best front page to everyone’

Valtteri Varpela, managing editor at Ilta-Sanomat, kicked things off with a presentation on how the Finnish newspaper currently uses personalisation on its homepage. The evening tabloid has an exceptionally strong online presence in Finland, with its website boasting 3,6 million unique monthly visitors. (For reference, Finland has a population of 5,5 million.).

However, among the newspaper’s vast audience there are naturally various kinds of news consumers, and each represents a different kind of challenge.

“We publish about 170 articles a day, and people who read our website once a day will miss a massive amount of that content,” Varpela said. “On the other hand the heavy-users, who may check the website up to 30 times a day, may run out of new content during the day.”

“Our aim with personalisation is to offer a website that serves all types of readers. Instead of presenting the same front page to different kinds of users, we want to offer the best front page to everyone.”

Ilta-Sanomat believes that personalisation will unlock the third era of online news consumption.

Ilta-Sanomat believes that personalisation will unlock the third era of online news consumption.

More broadly, Varpela said Ilta-Sanomat sees personalisation as the next major step in the evolution of digital news publishing. Online news first took off on desktop, while mobile has since become the leading platform for digital news.

“But we think that we have now reached the peak with mobile. The next step will be a move towards personalised news feeds.”

Bursting the filter bubble

While personalisation algorithms are becoming more and more common in the news media, many worry about the so-called filter bubble effect – that algorithms end up displaying audiences only articles that correspond with their taste and preferences, causing them to miss out on other important content.

Ilta-Sanomat addresses this by making sure that their algorithm not only relies on users’ profiles but also takes into account editorial judgement in how the articles are placed on the page. Moreover, the top articles on the page are the same for every user, and the article list includes specific “Exploration modules”, which ensure that readers are also exposed to articles that don’t match their preferences.

Ilta-Sanomat uses various types of modules on its personalised home page.

Ilta-Sanomat uses various types of modules on its personalised home page.

According to Varpela, the newspaper’s personalisation efforts have been a clear success: pageviews have not only gone up, but they are spread more widely across different sections and different types of stories than before. Moreover, heavy users are now seeing more stories, while others are getting more relevant content.

“In short, we have become a better news service.”

Rebuilding trust through transparency

Next, Tilman Wagner, innovation manager at Deutsche Welle, gave a description of the Content Personalisation Network (CPN), which aims to build an industry-leading personalisation solution for the news media (check out our vision here).

Wagner emphasised that as personalisation algorithms increasingly define what news content audiences are exposed to, it’s becoming crucial that the news media addresses related questions around data control, openness of algorithms, and transparency. As a research project CPN looks closely at these issues, especially transparency.

CPN aims to explore and learn about a wide range of topics related to content personalisation.

CPN aims to explore and learn about a wide range of topics related to content personalisation.

“Social networks and particularly Facebook’s News Feed have given algorithmic personalisation a bad name, since users don’t exactly know how they function,” said Wagner. “We want to bring more transparency into the equation. We believe that it will help rebuild audiences' trust in personalisation.”

Although several publishers have started to offer personalised content, the ways in which the different recommender systems work can vary greatly. “No one has the perfect algorithm, but what we try to do with CPN is experiment with mixing different approaches, and see if we can at least get closer to the right formula.”

Curation vs. personalisation

The panel discussion that followed the presentations expanded on the theme of editorial judgement, and why it should be incorporated into news recommender systems.

“During our research, we found that curation was an important reason why people came to specific publishers. They put their trust in editors, and expected them to choose the content that is relevant to their readers,” said Jamie Harrison, head of innovation programmes at Digital Catapult, one of the partners in the CPN project.

“Among publishers, there was a clear concern that if you put your content in the hands of a machine, you might lose some brand value.”

The panelists (left to right): Valtteri Varpela, Tilman Wagner, Ilke Lemmelijn and Jamie Harrison, with the moderator Rolf Dyrnes Svendsen.

The panelists (left to right): Valtteri Varpela, Tilman Wagner, Ilke Lemmelijn and Jamie Harrison, with the moderator Rolf Dyrnes Svendsen.

Regarding the filter bubble issue, the CPN recommender aims to address this by displaying a mixture of personalised and selected content. This ensures that users see the content that the publisher doesn’t want them to miss, said Ilke Lemmelijn, who is an innovation project leader at VRT, another CPN partner.

“Transparency is also crucial here: it’s very important to communicate to users why they see what they see.”

Start-up pitches

The session came to a conclusion with pitches from startups that are working with the CPN team. The following companies are developing innovative solutions that will expand the functionalities of the CPN platform.

We’ll soon publish more information about these startups and their work with CPN. In the meantime, check out these tweets to read more about their pitches:

Our jury chose Yoop as the most promising startup!

Congratulations to Yoop! You’ll soon be able to read about it, as well as the other startups, on the CPN blog.

By Teemu Henriksson

What we learned from VRT MyNWS: a pilot on news personalisation at the Flemish public broadcaster

image (3).png

At the end of March, the innovation and news department of the Flemish public broadcaster VRT experimented for four weeks with VRT MyNWS. This web application was built and tested as part of CPN and aimed to bring the VRT readers articles based on their interests. After the launch of VRT MyNWS, 949 testers subscribed to give feedback on their experience of personalisation, the usage of the web app and the display of articles. Read on to find out what we learned.

VRT MyNWS: a web app customised to readers

In order to measure whether people are better informed when they receive recommendations or not, the project team built the web app VRT MyNWS, resembling to the general news website of VRT NWS. News articles in the new web app could be found under three different tabs:

  • My news: personalised articles

  • Headlines: articles selected by the news department

  • Just in: most recently published articles

During one month, testers could give feedback via a button in the app, participate in surveys and were kept up to date via a weekly mailing. Overall, the team received over 200 emails with suggestions for improving the app. These mostly concerned the user interface and the recommendation algorithm. On a daily basis, a team of developers processed the feedback.

“It was extremely informative to gather input from end users, as well as the news department, and link those two into practice. Only by working together and experimenting on the production floor, we are able to innovate.”- Ilke Lemmelijn, CPN project coordinator.

Personalisation based on artificial intelligence and popularity

The testing with VRT MyNWS took place in two phases. In the first phase of the experiment, the tab ‘My news’ was loaded with articles that were selected according to three algorithms: collaborative filtering, content-based recommending and a random selection of recent articles. By collaborative filtering, readers received news articles that other readers with similar interests also found interesting. By content-based recommending, articles were selected and offered on the basis of the content and metadata of the articles. Lastly, by including a random selection of articles, the project team aimed to avoid readers receiving too many similar articles and ending up in a so-called filter bubble. The testers were divided into two groups that each received their own composition of the three algorithms.

In a second phase of the testing period, the project team chose to recommend articles based on their popularity during a certain timeframe. The team wanted to find out what the ideal timeframe would be. Testers were once again divided into three groups, each receiving popular articles from the last hour, the last 12 hours or the last two days. The results showed that testers appreciated it more when they received articles that were popular during the last hour, rather than for a longer period.

Feedback on the user interface and the algorithm

Based on the click behaviour, testers seemed to appreciate the simple algorithm from the second phase more than the complex system from the first phase. As such, the tab ‘My news’ was used 68% as opposed to other tabs in the second phase, while it was used 38% in the first phase. During the second phase, testers also used the 'My news' tab more when they were shown popular articles from the last hour old. Overall, testers said that they felt better informed in the second phase.

During the first phase, the user feedback mostly concerned the user interface, such as the width of articles, and the recommendation algorithm, such as irrelevant news. In both phases, the publication date of the articles appeared to be an important factor in personalisation. Apart from the small amount of relevant, older articles, the majority of testers preferred to read articles that are no more than two days old.

“Unique for recommendations in the news area is that the age of the articles is crucial in maintaining the interest of readers.” - Joris Mattheijssens, Data scientist at CPN and VRT Innovation

The next step for news personalisation at VRT

The insights and results of the VRT MyNWS test are further processed within CPN. The CPN team also calls out to other European news organisations to personalise their news stories with the software from CPN. In a new pilot, the Cypriot news organisation Dias and Deutsche Welle are also experimenting with the personalisation of their news articles. The insights and next steps will be published here.

Interested in piloting the CPN software? Contact us here!

Discover the “technology bricks” that power the CPN platform

kvistholt-photography-191153-unsplash.jpg

Our new report outlines the conceptual architecture and the technological infrastructure – what we call “technology bricks” – that power the CPN platform. The full report includes a description of each component, the functionality it provides, along with parameters, inputs, output, and API examples.

These components include Semantic Lifting, Topic Extractor, Uplifting/Depressing Article Classifier, Recommender AB-Testing, and Twitter Analytics

The set of “technology bricks” took as a starting point the user requirements collected earlier in the project. These “bricks” constitute the second version of the platform infrastructure, which will be ready by the end of May 2019.

For the second prototype of the CPN platform, the implementation of the features has been prioritised in a way that allows us to follow the planned schedule, while adding meaningful functionalities that will also be improved and extended in future prototypes.

Subsequent versions of the platform components are expected to provide updated versions of the currently available bricks along with possible new bricks, in order to adapt to possible new requirements and functionalities needed by the constantly evolving CPN platform during the implementation phase of the project.

Read the full report on the "technology bricks" here. Stay tuned for our second CPN platform release!

(Image by Kvistholt Photography on Unsplash)