Unlocking the Power of Archives: Keskisuomalainen’s Search Challenge
Keskisuomalainen maintains a vast archive of articles, images, audio, and video content that journalists and editors rely on to craft new stories. However, their search functionality was limited—based solely on article titles and automatically generated tags, which are often overly broad or inaccurate. This made it difficult for journalists to efficiently locate relevant reference material, resulting in hours of lost productivity. Additionally, the same flawed tagging system underpins the platform’s advertising engine, leading to mismatched ad placements and low clickthrough rates.
Our member Nemeon figured out the way forward for the challenges Keskisuomalainen was facing.
Keskisuomalainen
Keskisuomalainen is Finland’s leading regional and local media group, with a legacy dating back to 1871. Headquartered in Jyväskylä, the company reaches over 3.1 million Finns—more than 70% of the population—through a powerful mix of print, digital, radio, and digital out-of-home (DOOH) channels.
With a portfolio of nearly 80 newspapers and a growing suite of marketing, research, and distribution services, Keskisuomalainen empowers communities and advertisers alike to stay informed, connected, and visible—locally and nationally.
By combining deep journalistic roots with forward-thinking media innovation, Keskisuomalainen continues to shape the Finnish media landscape—delivering trusted content, scalable reach, and meaningful engagement across every platform.
Smarter Content Tagging and Search for Keskisuomalainen
Our member Nemeon, a data and AI partner for logistics and media companies, collaborated with Finnish media group Keskisuomalainen to build a next-generation tagging and search system—designed to enhance editorial workflows and advertising precision.
Leveraging Neo4j, local multi-language embedding models, and a privacy-first LLM deployed via AWS Bedrock, they developed a solution that transforms raw editorial content into a rich, searchable knowledge graph.
The system, hosted on AWS, begins by extracting text and metadata from Keskisuomalainen’s content management system. This data is then processed to generate embeddings and construct a knowledge graph that captures:
- Named entities and relationships
- Broader contextual themes
- Article tone and style
- Advertising-relevant categories
This graph is automatically imported into Neo4j, where it’s linked to document vectors to enable hybrid graph and vector search—giving reporters powerful tools to explore content semantically and contextually.
Finally, the system feeds enriched tags back into the CMS, where they serve as the foundation for targeted advertising selection.
The results
The project successfully delivered a proof of concept for graph-based article search and automatic tagging powered by large language models. This resulted in a significantly improved search experience for business stakeholders, enabling faster and more intuitive access to relevant content. The enhanced tagging accuracy not only streamlined editorial workflows but also contributed to more precise advertising targeting—leading to higher clickthrough rates. Beyond internal efficiencies, the system laid the groundwork for a future customer-facing solution, opening new possibilities for personalized content discovery and monetization.
About Nemeon
Nemeon is a specialized data and AI partner for logistics companies, offering tailored solutions that combine deep industry expertise with cutting-edge technology. Operating across Western Europe and North America, Nemeon’s diverse team—the Nemeonites—brings together professionals from eight nationalities, each contributing unique skills in data strategy, visualization, engineering, and artificial intelligence