🚀 Product Release: New Search, Filters, Security Hardenings
The October release of our Multilingual Knowledge System makes search faster, term recognition more flexible and filters more reliable.
New search engines, now in production. After a tech preview earlier this year, the rebuilt search is live. Standard search runs on a high-performance, full-text search engine written in Rust, embedded directly in the application and roughly twice as fast as Lucene. Responses are now near-instant and use way few resources. Term-as-you-type is smarter too: Multiword terms now match even when you type the words in a different order.
Term recognition you can tune. A new in-house built index identifies terms in a sentence in a few dozens milliseconds. New match options let you choose the behavior per request:
- Strict matching for terminology checks in Trados Studio or memoQ, where precision matters most.
- Tolerant matching for chatbots and user-generated content, where typos and messy input are common.
The new release for our Multilingual Knowledge System also brings improvements on filtering, faster imports, quicker navigation and security hardening on the road to ISO 27001.
👉 Read the full product announcement
🏭 Quanos ST4 AI Language Factories
At the Quanos Connect conference in May, we announced our partnership with Quanos, together with our partner company ESTeam. Now, one by one, Language Factories are being built.
Currently, we are fine-tuning various Multilingual Knowledge Systems to be used within the Factories: Various legacy terminology databases and taxonomies are being migrated; flat lists are turned into knowledge graphs and multilingually enriched. Coreon is a key part of the context layer which enables high quality AI translation.
🎬 Watch “Knowledge Graphs – A Time-Saver For Terminologists”
Unfamiliar subject area, bare terms with no context, and a machine translation that renders droit d’alimentation as “food law”: every linguist knows the frustration of a flat termbase.
In our new illustrative video, a linguist puts her toughest questions about Coreon’s Multilingual Knowledge System to a product manager:
- No time to restructure? Coreon uses AI and NLP to draft a first knowledge graph from your existing term list, such as an Excel file. You only review and refine it.
- No time to maintain it? In daily work, you don’t search for a concept’s place. You navigate straight to it, and broader, narrower and related concepts are already in place.
- Just a pretty picture? Seeing how concepts relate makes writing definitions much faster.
- Prefer search? Keyword search misses synonyms, which is how duplicates creep in. In a graph, related concepts sit side by side, so gaps and redundancies are easy to spot.
A concept system isn’t one more task on your list. It replaces search-and-hope with a map of your subject area.
tekom 2026: Meet us in Stuttgart!
This year, we’re bringing many great highlights to tekom:
- Coreon 2: Discover the new UI of our Multilingual Knowledge System, with reworked graph logic and several familiar features back on board.
- The October release, live: Try out the new search engine and see how quickly it recognizes terms, whether the input is polished source text or a typo-ridden chatbot query.
- Subprocessor Watch: Learn about our Eurostars project with Danish compliance company Wult. The demonstrator spots and classifies subprocessors in privacy policies and flags risky data transfers, supporting companies to stay compliant. For example, a knowledge-graph-supported LLM correctly red-flags a transfer to China as disqualifying, while the plain LLM only finds it “concerning.” ▶️ Learn more
Meet us in Stuttgart:
- Jochen: Tue 10th and Wed 11th
- Carina: Wed 11th and Thu 12th
- Michael: all days
What context does AI need to match humans? Do not miss Jochen’s joint presentation Better Translations: They begin in the CMS with Petra Dutz from Quanos!
As in previous years, we’re co-exhibiting with Adapt Localization Services, and this year Caron Publications joins us too.