The 2025 edition of ETUG brought together almost 40 Trados power users at a new venue, the Seminarwerkstatt in Vienna, along with Daniel Brockmann and Luis Lopes from RWS, who were on hand to field questions from the cases studies and to present the latest developments from RWS’ perspective. It was a particular pleasure to welcome participants from as far away as Spain, Belgium, Croatia and Greece.
Move to specific presentation:
- Day 1: Fleischmann, Bevington, Elbargisi, Brockmann, Panel, Okoniewski.
- Day 2: Ackermann and Brockmann, Shaw, World Cafe, Lane and García Magariños, RWS Roadmap
Day 1: Wednesday 24th September 2025
Julia Traub-Teubl, Plasser & Theurer and ETUG Chairperson, opened the conference with delegates from numerous countries from around Europe. Ralf Lemster again moderated the conference, the programme of which focused strongly on AI in Studio, Terminology, using XML to extend Trados for non-conventional workflows, and use cases on moving from file-based to cloud-based solutions in Trados as well as from GroupShare to the cloud-based Online Editor. In addition, there was an AI panel and a World Café session. The new venue, Seminarwerkstatt proved an excellent location for encouraging dialogue between participants and ensuring two days of lively discussions about Trados products.
Terminology as a Guardrail for AI and moving from file-based to cloud environments
Klaus Fleischmann, Kaleidoscope, opened with a presentation on how terminology can act as a guardrail for AI. As highlighted at several points during the conference, terminology is enjoying a renaissance in the LLM/GenAI era and has a driving seat in translation. Currently over 90% of companies are investing in GenAI but practically none have reached AI maturity, with many failing to achieve and scale value. Despite the claims that “translation is solved” a lack of context and linguistic control means that beyond the tech hype, there are still many problems with achieving high-quality translation results.
Terminology and AI/LLMs have symbiotic roles: AI/LLMs can be used as part of the terminology management process for extraction, clustering relations and generating metadata as well as verification and correction of terminology. In turn, terminology in the form of Terminology Augmented Generation can use natural language data for context and for precision and control. Termbases’ precise and structured data model allow easy retrieval of relevant information. In additional terminology being accessed in real time is faster than searching pre-embedded data – which also uses far fewer tokens, which might also prove cheaper if charging by quantity of tokens. Additional metadata also helps handling edge cases such as disambiguating homographs. For language professionals, such as terminologists, it also means that they remain in a lead position, rather than subordinate to the machine.
Robert Bevington from dSPACE, and a past ETUG Steering Committee member, presented the first case study – about dSPACE’s journey away from file-based setups to cloud(s) – as he remarked initially there should have only been one cloud, but there ended up being two. Moving away from a file-based setup had partially been triggered by issues presented by hybrid working (and problems of working with large files over a VPN). It also proved a timely way to reduce workflow steps – and ensure that workflow steps were realigned. At the same time there were also improved access and rights controls via roles, permissions and groups.
The switch had also been necessitated by difficulties sharing terminology with LSPs, with the terminology solution having become outdated and “bloat” of the number of technical resources (e.g. TMs and project templates). Some concerns existed about the lack of an offline option, if the cloud server was down, but this had not been an issue to date. One con of the new setup had been the large number of mail notifications. In concluding his presentation, Robert mentioned that the reduction in technical resources and tooling had been an unexpected bonus. Future evolutionary steps will be to replace Trados Studio with the Online Editor and to use the task management tool to automate project creation.
Using XML filetypes for subtitle captioning and format heavy translation
Amr Elbargisi from IAEA presented about how XML filetype definitions can be used to permit non-conventional workflows in Trados Studio. His use cases particularly highlighted the use of XML for translating heavily tagged formats (e.g. InDesign) as well as for subtitling. In the case of subtitling this approach helps ensure that non-textual elements (e.g. timestamps for subtitles) are not erroneously translated. In addition, since a mistake in an auto-propagated timestamp can cause a subtitling file to crash, it also eliminates this risk.
The platform-neutrality that XML provides has become increasingly important as demand for non-standard formats increases, especially for overcoming issues of segmentation not aligning with structure. Such non-conventional workflows are particularly useful for multilingual output that needs to be released in multiple languages simultaneously. However, due to their non-standard nature, definitions often need to be re-created if upgrades are conducted. Other use cases highlighted related to accessibility such as “alt” descriptions for graphics, and in the case of multilingual publishing, it removes the need to recreate elements in every target language.
Translation and AI
Daniel Brockmann, RWS, highlighted how AI capabilities are used in translation software, with automation of translation, enhancing translation QA, refining/improving translations and terminology management being the most common uses. Accuracy and translation quality issues, ahead of security risks and compliance and regulatory concerns weigh heavily as do job displacement issues.
Trados has worked on incorporating AI and LLMs into its products and now allows LLMs to be incorporated in addition to more established NMT solutions (with Trados allowing connectivity to over 100 NMT providers and 1,000s of LLMs). It has already implemented numerous AI use cases for linguistic technology. Generative translation is one example – where the output is adapted to the audience based on a prompt. As translation becomes more multimedia- based and while there is less traditional “written translation”, new features include generative subtitling and dubbing a further possible extension. Additionally, Trados is bringing AI-powered terminology extraction, while Trados AI will soon be able to be integrated with Microsoft Copilot in Microsoft Teams.
Several trends are apparent: the rise of GenAI has heralded a terminology renaissance, while there are indications of a future move away from sentenced-based to document-based segmentation. There is heightened interest about transcreation using tools and multilingual source content, while multimedia translation enjoys a greater role. AI can also be used to support project managers and engineers, while in the future linguistic review might use style guides, and picture-based context might assist in generative translation, and with LLMs that are prompted to use translation memories and termbases.
Panel on challenges faced by translation posed by AI
In the panel chaired by Kerstin Berns, Berns Language Consulting (blc) and ETUG, Daniel Brockmann was joined by Dagmar Gromann, University of Vienna, and Frank Drefs, Volkswagen and ETUG. The panel focused on challenges facing their respective sectors in relation to translation. “Staying or keeping relevant” could be one challenge. Similarly, language technology companies like RWS face similar issues of being heard by tech firms, who both consider translation as an afterthought and are keen to unnecessarily reinvent the wheel, when the tools they need for language processing including translation are readily available. Similarly, a prominent challenge faced by in-house language services is the issue of apparent “data and budget takeover”: such departments are acknowledged as a great source for well-curated monolingual and bilingual language data, but are very visible in terms of the budget they also consume. AI-readiness of tools used by language services also cause potential issues.
Companies have started to come round to GenAI’s flaw in terms of language processing, especially translation, but it has taken the implications of potential damage/harm arising from LLM/GenAI-generated content to reconvince them: they were initially frequently “seduced” by the seeming fluency of their output. Translation has also suffered from being a “shy business”, while AI companies are well known for making their bold claims. Europe lags behind the US and Asia, partially due to high concentration of SMEs. SMEs typically do not have the resources to be able to implement their own proprietary AI solutions so instead opt for “off the peg” solutions. Language services units of large companies need to be active in internal guidelines for regarding the use of AI – particularly for translation purposes like highlighting that language-related activities need to always involve language services. The simplest way to go about this is to produce a matrix diagram about what types of texts may use AI, and to what extent and for what purpose.
From a research and educational perspective, there has been a big shift from the “old world”, with newer generations taking little convincing to get on board, although a bifurcation exists in terms of certainty about whether to use GenAI and uncertainty about how it works. Oral defences of theses and practical exercises under “lab conditions” have taken on a greater meaning with many students happy to use LLMs/GenAI to write theses or to complete coding assignments, to an extent that instructors almost have to look over their shoulder. “Defence cramming” is also emerging as a phenomenon, where students have less substantive retention of their theses, and need to revise intensively ahead of defending their theses. Translation studies pathways in higher education are changing, with the onus on “educating better” to improve employment perspectives – for example with the BA (Transcultural Communication) and MA (Translation) pathway now changing its focus to have more practical components in relation to language technology.
XML, filetype definitions and document structure information: lightening the load for translators.
Grzegorz Okoniewski from the European Commission presented about how Akona Ntoso for the EU (AKN4EU) will help reduce the cognitive load for translators and financial cost of translation. 2024 output amounted to 2.35 million pages, with nearly 93% from English. Of this, 56% is for legislation, so preparing filetype definitions (FTDs) in Studio to handle translation of legislation in XML (currently produced in Word but there is an interinstitutional project to move to XML) and using XML structure to disambiguate linguistic content and automating “tedious” segments (typically multi-level numbering) has massive potential. Stylesheets also help to ensure that text appearance is displayed in accordance with the Document Structure Information (DSI) at segment level as well as ensuring the correct text appearance at subsegment level. This is a marked improvement over the using Word, which loses context in Studio due to not fully using DSI to provide necessary context (e.g. failing to extract automatic numbering). In addition, AKN4EU also ensures that segments are merged in such a way that one segment has one meaning entity.
A Studio plugin called “AutoPreparator”, which performs predefined regex-based search/replace actions manages this. The AutoPreparator project will be extended to ensure accurate translation of Official Journal References (which feature frequently throughout legislative texts) as well as for URLs. Following the initial evaluation, there will no longer be a need for human supervision, resulting in cognitive (for the translators) and financial (in terms of translation cost) gains. Considerable financial gains are achievable by automating the process that accounts for 1.5% to 3% of translation volume– especially since the segments automated take longer per character than others). DGT’s QA landscape consists of a range of checkers to identify formal errors, checking of numbers and tags, terminology consistency, a smart autosuggest feature and, smart pre-translate (by the AutoPreparator).
The first day was rounded off by ETUG’s General Assembly, and the re-election of its Board Members, Julia Traub-Teubl and Anita Wilson. Thanks were given to Giuseppe Forte, from the Centre de Traduction, who was standing down from the Steering Committee, for his work on the programme for ETUG 2025. In the evening, conference participants gathered for typical Austrian fayre at Huth Gastwirtschaft, for a seasonal dinner of Austrian specialities.
Day 2: Thursday 25th September 2025
The symbiotic relationship between terminology and AI
Day two opened with a presentation by Sophia Ackermann, Berns Language Consulting (blc), and Daniel Brockmann – again looking at the symbiotic relationship between terminology and AI in machine translation. They demonstrated how AI can support terminology work (e.g. for term extraction (both for monolingual and multilingual term extraction) and synonym identification) and in turn terminology can enhance AI-driven tasks (e.g. for use in chatbots, information retrieval). They highlighted how training and fine-tuning MT is able to increase term consistency, and how the use of AI for creating example sentences that used terminologically correct translations can in turn be used to fine-tune NMT.
The fine-tuning approach involves fine-tuning a generic baseline model using your own data, to ensure better results in your domain-specific language and bringing through your own tone of voice. For example, it is possible to trained company-specific terminology into the Trained MT model, which sets it apart from Generic MT, which does not have knowledge of company-specific terminology. In turn this can also help to rectify term errors present in the source language text too. Daniel also reiterated how much of Trados’ AI capabilities that have been launched to date or are currently in the pipeline actively leverage terminology.
Security, Flexibility and Teamwork using the Trados Online Editor
Christopher Shaw outlined the European Investment Bank (EIB)‘s experience in moving to using the Trados Online Editor – especially for working with freelancers and LSPs, concentrating on how the bank addressed security issues, how the new environment fostered flexibility in the workflow and also enhances teamwork (e.g. through real-time comments and improving freelancer access to TMs and TBs). In terms of the business case there were many security benefits, in that content was stored on EIB servers and thereby eliminating accidental retention of content by external contractors, while risk mitigations also were useful for audit purposes. This had required a lot of negotiations with IT security, as well as a lot of firewall testing.
In terms of flexibility the move to the Online Editor had negated some of the project management overheads, reducing some steps, as Robert Bevington had also highlighted, such as needing to export packages or to split documents, while internal translators and PMs could then use Studio for working on files that had been translated with the Online Editor. There were some issues with Single Sign On, and there were limitations in terms of usefulness for language combinations with small TMs. In addition, for the purpose of teamwork, with live translation, the revision stage can be started earlier (even during the translation stage) and by working in the Online Editor freelancers enjoyed direct access to both termbases and translation memories. So far Online Editor has not yet been rolled out for all language combinations but is due to in due course. Once this happens, there might also be the possibility to reduce the number of Studio licences as some translators become used to Online Editor, especially if the latter’s features are extended to a level that is close to that of Trados Studio.
World Café on Productivity, the Trados Cloud platform and Terminology and AI
After two use cases, a World Café session focused on the three core areas addressed over the two days: translation productivity, the Trados Cloud platform and Terminology and AI. From the table on translation productivity the need to move away from post-editing, seen as a dumping rate activity towards more valuable and better remunerated review was highlighted, particularly considering the changing but nonetheless heavy cognitive load that translators encounter, especially post-editing machine translation. Payment models were also discussed. Given some of the talk about XML extension of Trados and layout heavy formats, future enhancements could be to provide linguists with lighter versions of Studio to allow them to focus purely on translation, while having another product flavour that allows genuine in layout editing. This was also touched upon in the Roadmap presentation. Potential boosts in the future might see moving from fuzzy matching to semantic matching, the use of conversational AI and AI for repetitive project management tasks. For all the innovations, some participants still make use of some of Trados’ longest-standing features like alignment and mentioned concordance searching as still being a very important feature, and improvements to term recognition.
Migration to the cloud environment was still hampered in some cases by limitations – for example Online Editor being limited to 100 files. Concerns still exist in regulated industries and certain sectors about the security of the cloud offering – this is also one reason why RWS is continuing to offer on premise and file-based solutions. With the advent of the new cloud platform, there were requests from some participants to also have new training possibilities in relation to Trados Studio and Online Editor, while the acceptance of cloud solutions also now brings Content Strategy to the fore.
The Terminology and AI table revisited the importance of terminology – including the topics of pre-processing of terminology, terminology verification, and fixing terminology through to the value of “wrong” terms and project-specific termbases. Further use cases for terminology were also addressed e.g. for chatbots, and how AI could be used to assist in building up relations between terms, and adding new visualisation options like concept maps, knowledge graphs and relations between terms. In addition, there was discussion about the role of metadata in connection with terminology and AI.
AI Plugin in Studio 2022 – ECB’s “AI Professional Companion”
Catherine Lane and Daniel García Magariños presented about the European Central Bank‘s use of the AI PlugIn in Studio 2022, which it calls its “AI Professional Companion”. Importantly, when translators are concerned about their ongoing role, a “Human in the Lead” approach is advocated: AI is there to support linguists but does not relegate them to being “humans in the loop”. The examples that were demonstrated show how prompts were being used for tasks like improving clarity and inclusivity of texts, or for making texts more informal and conducting ambiguity checks.
Some of the Q&A and discussions also dipped into the mindset change among in-house translators and how a “champions” approach works well, i.e. with technology savvy linguists taking the lead, but also ensuring that their more reluctant colleagues are not left behind. There had also been some interesting and serendipitous outcomes. For example, it had emerged by chance that the prompts were very successful in helping ensure that segment tags were successfully applied from the source into target units), but the gains from using AI were not the same across all language pairs and were certainly better in some combinations than in others. Planned future migration to Studio 2024 would also enable the integration of new features, that were currently not possible in the Studio 2022 environment.
RWS Roadmap
Luis Lopes and Daniel Brockmann rounded off ETUG 2025 with RWS’s Roadmap – including covering some of the technical and feature-related highlights from the recent release of Studio 2024 SR1 and what lies ahead. Forthcoming developments include the realisation of a perennial and popular request in the RWS Community. While other CAT tools are discontinuing their development of “on premise” solutions (and going “all in” on cloud-based solutions), on premise solutions will continue to be available, although the gap between Trados Studio and the Online Editor in terms of their respective number of features will close to a certain extent.
Cloud adoption has clearly increased considerably in 2025 in terms of volume. With the empowerment of subject matter experts (SMEs), it is necessary to consider how to involve them in the translation and review process, while also not scaring them off with the Trados views that translators take for granted – which is why an In-Layout Editor for Word and PowerPoint is a useful feature. Layout heavy file formats, as had been highlighted on day 1 by Amr Elbargisi, were another reason for a paying add-on in the Online Editor for InDesign Preview. Similarly, vendor improvements in the project management side of Trados were also being rolled out. New features in the Online Editor include being able to open multiple files in Online Editor (i.e. virtual merge) and segment history, as well as adding live video preview to subtitle translation.
In addition, Perfect Match Automation, functionalities to leverage adjacent languages and project conversations were also additions for project management. In terms of AI, Connected AI offered the chance to have connections to a wide range of LLMs and there were new possibilities for AI Video Dubbing.
Regarding Studio there is now a Trados Copilot to help users use in the form of Smart Help Across Trados, while the AI Assistant is a foundation for plugging in any LLM as well as Local AI for translators through Hugging Face. Similarly, there are also improvements in terms of both cloud and GroupShare integration. The AppStore is also becoming more accessible. TMX files will be able to be added directly as file-based TMs. By modernizing Studio, files will open more quickly, and the speeds for locking large numbers of segments will improve dramatically, and navigation between different open document tabs will become faster. And last, but no means least, it was announced that a 64-bit version of Trados Studio using a modernized desktop platform will appear in 2026.
The ETUG Steering Committee thanks speakers and participants for their lively interactions over the course of the two days and looks forward to welcoming participants to its ETUG BiteSize webinars in 2026, and the next physical conference in 2027.
Conference participants can access the presentations (and a publication by Daniel Brockmann) through this password-protected page.