Retriever trust data insight for better business

Improving business by data competence

Can AI provide a new understanding of the media? And can media data aid in understanding languages?

These were the questions Claes Lyth Walsø, chief information officer at Retriever, was asked to answer at Fibep, the world federation of media intelligence companies at its global Tech Day on May 21st 2021.

Retriever is the number one data analysis company and communication insight provider in the Nordics represented in four countries.

The Fibep community organizes 170 corporate members employing over +14000 people in over 72 countries working on analyses of media and database content. The companies have evolved from mere “clipping” services to full scale media monitoring.

Digital focus

But few companies, however, have focused their innovation work on digital development like Retriever has done since the turn of the century. Furthermore, even fewer has joined a research based consortium like NorwAI with its three universities, two research institutes and ten other companies. 

Retriever has been leading the digital transformation of the media monitoring industry for over a decade, by introducing the latest technology to the market and by adapting our business to the constantly evolving data landscape, says Claes Lyth Walsø.

Retriever’s data are particularly interesting for doing research in what is called Natural Language Processing (NLP).  This involves understanding the human language and then creating AI models. Retriever has huge amounts of textual content with good quality. It is an excellent starting point for creating better models in entity recognition, sentiment analysis and the like.

Today, we have 70 people working on media analyses. By using machine learning, one of the AI technologies, we will be able to automate, categorize and code our data to be more effective and broaden our services to the benefit of our customers, says Claes Lyth Walsø.

Improved services

To give an example, Retriever was asked by DnB some years ago to monitor how economic content was covered in media also reflecting the increased complexity of economics. Manually, it was impossible to give an answer good enough for decision insights. New data competences revealed that this content was reduced over the years, helping DnB decide to launch their own news services.

One of the success criteria for the research center is precisely that the research must be able to continuously lead to innovation among the partners. Personally, I think over time we will be able to work both broadly and in depth with NorwAI, Claes Lyth Walsø says.

By depth, I mean that we are able to do what we do today, but with even better accuracy. In addition, we will work to increase the breadth of what we do by delivering new method analyzes or perhaps think completely new about how we deliver our services.

So the short answers to the Fibep questions, Claes?

No doubt, machine learning will improve our capabilities to serve our customers. And no doubt as well, text databases of good quality are the stepping stones to build AI models in our own language, not only translating methods from foreign languages.