When content speaks the language users want, then the experience turns from good to great!

Recently, Tridion’s partner Coreon GmbH convened with Tridion’s team to discuss and understand how Coreon’s multilingual knowledge systems deliver value to the increasing number of global enterprises.

We had Jochen Hummel, co-founder and CEO of Coreon, lead the presentation. Jochen is also the CEO of ES Team, Vice Chair at LT Innovate and was the former CEO of Trados.

Jochen briefly introduced Coreon as a Berlin-based software company, serving as a leader in Multilingual Knowledge Systems with SaaS offerings. The software makes the search, machine learning and IoT applications to be interoperable. It enables multilingual AI and is deployed across the industry, eGov, and education. It is hosted in German data centers.

 

At the heart of Coreon is the unique fusion of terminology with taxonomy/thesaurus/ontology. This amazing tool marries language with knowledge in a holistic way. It addresses the major gap or black hole that currently exists between terminology and taxonomy (knowledge graphs). Coreon deploys specialists in both areas that have collaborated and built this unique solution. It helps multilingual knowledge workers, content creators, and experts to unlock value across various business processes such as search, text mining, IoT, classification, and conversational AI.

 

 

Jochen touched upon the importance of structured data and knowledge graph to turn information into knowledge. What Coreon does is it combines multilingual terminology list with knowledge map which results in multilingual knowledge system.

Jochen explained the use of this with an example of EuroVoc, a database that the EU uses to classify all the information. The knowledge graph is created only once, then the linguists label it and recycle it in different languages which can be delivered across different countries.

 

To enable Advanced Linguistic Search, Jochen recommends converting key search strings into meaning, which means locating and connecting these in the taxonomy. It can be done by first creating domain heat maps in the system and then linking it to other datasets (DMS, ERP, HR, etc.). This ensures that search strings don’t get lost in silos, but reach to the correct information in the system.

 

All of this is performed through the Coreon Multilingual Classifier tool whether its website or documents:

Jochen mentioned a few examples such as customers finding relevant products on e-commerce platforms leading to higher sales irrespective of the language being used to perform the search. Another example is global machine manufacturing which runs business processes around the world in different languages, it benefited from greater efficiency that led to cost savings.

In the government sector, eGov benefits from seamless and accurate information exchange across different countries in different languages, which leads to greater efficiency.

Jochen emphasized that most enterprises around the world primarily focus on effectively communicating information to their clients, but miss the point that today it’s increasingly a two-way process. Customers are communicating through product searches, chat boxes, comments, blogs, reviews, social media posts, etc., which are spread inside and outside the enterprise systems.

It is therefore critical to understand what customers are communicating and when customers communicate they prefer to use the language that they are most comfortable with. This means that enterprises need to be equipped with a multilingual tool to be able to capture their customers’ feedback.

Obviously, this is not possible manually; hence Coreon offers multilingual AI-based tools such as chatbots, text analytics, and competitive intelligence. AI learns with data and half of this data is in text form.

Jochen explained the entire process in which the content creation and translation works using the knowledge of subject matter experts. The content is created in a form that is consumable by machines.

Coreon has built a process that it terms a ‘Knowledge Factory’, which has various stages or modules that ensure that language translation is accurate and optimized for quality. It involves the use of checks and balances on AI.

Jochen compared and contrasted content translation with content creation to explain the way both works is not too different. Further, he also listed the ways in which content differ based on their degree of expressiveness.

Jochen through a 2x2 matrix explained that while there are many solutions in the market that excel in either linguistic strength and multilingual coverage or semantic expressiveness / machine-readiness, but not at both. Coreon with its multilingual AI for humans and machines coupled with strong translation processing capabilities excels at both!

Further, Jochen explained that many times there are customers that have even 100,000 concepts involved to process in chat GPT and hence taxonomization is complex even with AI. In such scenarios, he depicted an approach that Coreon has adopted. It involves building clusters that combine similar concepts (related to each other) into one bucket. These clusters have sub-clusters beneath them again combining similar but lower-level concepts.

This requires machine learning and humans to work together to arrive at such classification. Jochen believes that Tridion and Trados can come together and make content in any language to connect to such classification in English.

MKS can facilitate seamless two-way communication between customers and enterprises. For example, Jochen mentioned that if a customer sends a query in a local language (not English) then traditionally it will first need to be converted to English for the customer support representative to understand then customer support’s response in English back to the local language for the customer to understand. But with MKS, the customer’s query goes to the classification system and finds a response in the local language and is communicated as is to the customer.

MKS works as a single source of truth, its functions include content translation/creation, term annotation/classification, semantic search, synchronization/interoperability, conflict resolution, natural language generation, and AI customization/tuning / correction.

Jochen believes that when RWS will combine Trados and Tridion it will have localization and under one / single solution. It will be able to seamlessly process content creation and translation under one roof and thus deliver multilingual AI.

 

As closing remarks, Jochen reiterated that Coreon currently is the only efficient solution on the market that offers multilingual knowledge for both content creation and translation.