When the client wants to order a translation, they are asked to select the target language.<\/p>\n\n\n\n Figure 1: Selection of target languages<\/em><\/p>\n\n\n\n The list shows a set of predefined languages. If needed, users can choose from over 150 languages. The initial suggestions are generated dynamically from recent usage statistics to streamline and speed up the ordering process.<\/p>\n\n\n\n Figure 2: Translation mode and deadline selection<\/em><\/p>\n\n\n\n After an upload is completed, the system immediately proposes seven possible deadlines and displays them as selectable options. Experience shows that 90% of users accept one of the proposed deadlines and simply click \u201cOK.\u201d In addition to these seven suggestions, users may also choose a custom deadline from the pickup schedule. If a user selects a deadline that is not feasible, the system displays an appropriate warning.<\/p>\n\n\n\n Figure 3: The client can select his own deadline<\/em><\/p>\n\n\n\n The order screen is entirely parameterizable. Prices can be replaced by points, the express charge can be suppressed, and the translation modes can be reduced or expanded. The parameterization takes place through the registry or the system operator interface.<\/p>\n\n\n\n Figure 4: Parameterisation through registry or System Operator Interface<\/em><\/p>\n\n\n\n The deadlines are calculated according to the availability of the translation resources. They can be parameterized for each user or user group individually.<\/p>\n\n\n\n Translation Mode and Workflow Schemas<\/strong><\/p>\n\n\n\n The translation mode corresponds to different workflows with distinct pricing or effort models. In this example, four workflow options are presented. TTN TMS is a modular system, allowing each workflow to be easily configured according to specific requirements.<\/p>\n\n\n\n Classic translation<\/strong><\/p>\n\n\n\n In the classic translation model, the text is translated by a human translator and then carefully corrected by a proofreader. The Translation Manager supervises the process, especially if there are several target languages, ensuring consistent formatting across all languages. They review the source files for any images. If necessary, they organize vector files so that they can be overwritten, or they place text boxes covering the pixel legend so that the translators can access them in segmented mode of the XLIFF editor.<\/p>\n\n\n\n Figure 5: Workflow for classic translations<\/em><\/p>\n\n\n\n Before translation starts, TTN TMS makes the linguistic resources available: the client\u2019s translation memory and termbase are assigned to the project and connected in the CAT tool. In parallel, the AI agent TermBoy scans the source text for specialist terms still missing from the termbase. It researches those terms on its own and creates complete entries; terms with low reliability are sent to a subject-matter expert at the client for approval. The translator therefore works with verified terminology from the very first line (see the chapter \u201cTermBoy \u2013 the AI Agent for Terminology\u201d).<\/p>\n\n\n\n Once the translators have completed their work, the text is checked with the QA tool integrated into the CAT system. Quality Assurance verifies the formal aspects of the translation: punctuation, numbers and dates and \u2013 more importantly \u2013 whether the terms stored in the termbase have actually been used in the translated text. If the translator deliberately departs from an entry, they have to validate the corresponding message. The QA settings are configured automatically by TTN TMS when the project is created, so that they match the client\u2019s language-specific requirements. In Switzerland, the guidelines of the federal administration apply unless agreed otherwise.<\/p>\n\n\n\n A second test, the AI check, verifies the semantic correspondence between source and target text. To do so, the entire text is first submitted to an AI model \u2013 usually the current version of ChatGPT or Claude \u2013 so that the model can grasp the connections between the parts of the document. TTN TMS then opens the bilingual, segmented XLIFF file and sends each sentence pair individually to the AI, asking whether the translation fits the overall context and whether it can be improved. The AI check returns a table listing the numbers of the segments the translator should review. These are often insignificant nuances; serious translation errors and omissions, however, are detected reliably and can be corrected.<\/p>\n\n\n\n Once the translator has carried out both checks and uploaded the files, they are automatically routed to the proofreader. The system again generates an AI check and a comprehensive QA report from the CAT tool and provides both as support for the revision. After the proofreader has submitted the revised version, the final AI and QA reports are produced. They can also be retrieved by the client and document the linguistic accuracy and consistency of the text before delivery. Both checks are described in detail in the chapters \u201cAI Quality Check\u201d and \u201cThe AI Check from the Client\u2019s Point of View\u201d.<\/p>\n\n\n\n If the client changes the text after the translation, they can upload it or send it by mail. The file will then be retrofitted, and the latest modifications will be updated in the Translation Memory. The translation memory therefore stays up to date even when the final version was produced outside the system.<\/p>\n\n\n\n Full Post-Editing<\/strong><\/p>\n\n\n\n In full post-editing, the text is first translated by an AI provider \u2013 by DeepL, Claude, ChatGPT or another LLM provider. The pre-translation always runs via the Trados API and uses the same resources as a classic translation: 100 % and 99 % matches from the translation memory are adopted unchanged, fuzzy matches are passed to the AI provider as a hint, and the relevant termbase entries are inserted into the translation prompt segment by segment (term injection).<\/p>\n\n\n\n Figure 6: Full Post-Editing workflow<\/em><\/p>\n\n\n\n A second AI provider then checks the machine translation: if Claude produced the translation, ChatGPT verifies it \u2013 and vice versa. This cross-test is referred to as the AI-AI check. Together with the QA report, the post-editor thus receives a list of the passages where the machine translation is weak before starting work.<\/p>\n\n\n\n The post-editor carefully compares the target text with the source text and brings it up to the level of a human specialist translation. If the system detects any pictures, references, or other unhandled elements, the Translation Manager will check the files and adapt them accordingly before sending the file to the client. The Translation Manager also verifies that the translator can access the relevant resources from the Translation Memory and TermBase, ensuring that the linguistic work is carried out with the correct reference data.<\/p>\n\n\n\n After post-editing, the file goes through the same checks as a classic translation: QA report and AI check. The Translation Manager releases the translation and delivers it; subsequent changes made by the client flow back into the translation memory via Retrofit.<\/p>\n\n\n\n Post-Editing Light<\/strong><\/p>\n\n\n\n Light post-editing works similarly to full post-editing, but does without the revision stage. The post-editor focuses basically only on the target text and consults the source text only if there is an apparent problem in the machine-translated text.<\/p>\n\n\n\n Figure 7: Post-Editing Light workflow<\/em><\/p>\n\n\n\n Here too, the post-editor has the QA report and the AI check of the second AI provider available and can therefore direct their attention specifically to the segments that were flagged. As in full post-editing, terminology and TM matches have already been fed into the pre-translation.<\/p>\n\n\n\n Unless the translator writes \u201cHold\u201d or \u201cCheck\u201d in the window that appears when clicking on check in, the translation is forwarded automatically to the client. This model is suitable for texts where intelligibility and speed matter most and no publication-ready wording is required.<\/p>\n\n\n\n Automatic translation<\/strong><\/p>\n\n\n\n The Automatic Translation model is the simplest. It sends the text to ChatGPT, Claude, DeepL, or any other well-known LLM Provider and translates the text. The translation is always done using the Trados API. The system synchronizes the terminology in the Termbase, inserts the matching terms into the prompt and prompts the Translation provider to implement fuzzy matches. Full matches are always translated in the same way without modifications.<\/p>\n\n\n\n Figure 8: Automatic translation \u2013 the AI check goes to the client<\/em><\/p>\n\n\n\n Automatic Translation via TTN is particularly useful when the client or their organization has large translation memories and the terminology is well defined in the TermBase, since the terminology-aware processes take into consideration existing translations and approved terms. This allows the system to produce output that is more consistent with the client\u2019s established language.<\/p>\n\n\n\n As no human post-editing takes place in this model, quality control goes directly to the client: the AI check by the second provider and the QA report are delivered together with the translation. Both are attached as HTML files and are additionally available in the order details in Order Management.<\/p>\n\n\n\n The client can therefore see, segment by segment, which passages the verifying AI flagged, and decides for themselves whether to use the text as it is, have individual passages revised, or convert the order into a post-editing job afterwards. Automatic translation thus remains traceable and documented even without human revision.<\/p>\n\n\n\n Translation Archives<\/strong><\/p>\n\n\n\n TTN TMS clients are members of different user groups. One of the most important groups is the archive group. A member of an archive group can see and download the translations of other members within that group. Usually, an organisation has different archive groups. It is not suitable for texts from the HR department to be accessible to all other users. However, translations from the Press and Media Department can be seen and downloaded by all members.<\/p>\n\n\n\n TTN maintains a full-text archive for all source and target texts. Therefore, all files are converted to HTML files.<\/p>\n\n\n\n Figure 9: Full-text search in translation archive<\/em><\/p>\n\n\n\n The archive search allows sophisticated searching by search terms, languages, format, and other criteria.<\/p>\n\n\n\n Figure 10: A three-level list provides fast and easy access to the desired search results<\/em><\/p>\n\n\n\n The archive presents results in a three-level list, facilitating quick access to completed or pending orders. All order information, including documentation files, instructions, or XLIFF files, can be found with a few clicks.<\/p>\n\n\n\n Figure 11: Translation archive with pictures in HTML in browser<\/em><\/p>\n\n\n\n Since all files are converted into HTML, the content can be displayed directly in the browser and downloaded. The integration of pictures, graphs, and schemas allows for quick orientation within the text.<\/p>\n\n\n\n Thanks to the conversion into HTML and full-text indexing, the user can search and access all current and executed translations very quickly.<\/p>\n\n\n\n Figure 12: A side-by-side viewer allows analysis of target and source texts<\/em><\/p>\n\n\n\n The side-by-side archives allow translators as well as clients to locate and identify search strings and old translations very quickly.<\/p>\n\n\n\n Figure 13: Users can access and search the TMs without third-party software<\/em><\/p>\n\n\n\n TTN TMS seamlessly integrates with the GroupShare server, providing authorized users with direct access to Translation Memory content. Depending on their permissions, users can download TM data in TMX format without needing to access GroupShare directly.<\/p>\n\n\n\n The same functionality applies to termbases: users can consult and retrieve termbase content either through TTN TMS or via the multilingual online interface, ensuring transparent and efficient access to all linguistic resources.<\/p>\n\n\n\n Figure 14: MultiTerm online database<\/em><\/p>\n\n\n\n The MultiTerm database and its entries are very important for automatic translations, as the terms are exported to DeepL and other providers. The client can add terms and search terms directly in their database. A link to the database is added in the delivery mail, allowing them to affect the next translation.<\/p>\n\n\n\n Direct access to TMs and TBs guarantees full transparency of the translation resources. The client can verify at any time whether they are properly maintained, resulting in greater confidence in the overall process.<\/p>\n\n\n\n Access to Translator Database<\/strong><\/p>\n\n\n\n The client can \u2013 if configured \u2013 get a link to the translator database, where they can access the profile of the translators and proofreaders.<\/p>\n\n\n\n Figure 15: Access to translator and proofreader profile<\/em><\/p>\n\n\n\n The system can be parameterized to insert a link in each translation delivery report that goes to the translator database, where the client can retrieve information about their translation team. This allows them to ensure that the translation team is qualified or, if authorized, restrict the team to their preferred translators.<\/p>\n\n\n\n Feedback Functions<\/strong><\/p>\n\n\n\n TTN offers powerful feedback functions that allow clients to rate each translation. These ratings have a direct impact on the selection of the translation team.<\/p>\n\n\n\n Figure 16: Client feedback function with evaluation<\/em> The client can upload a revised version. The old version will be replaced in the archive, and the translator and the proofreader receive a summary where all changes are listed by email. The new file is retrofitted into the Translation Memory \u2013 this procedure is explained later.<\/p>\n\n\n\n The feedback function allows the Translation Manager to know if the client is satisfied with the service. If there is unsatisfying feedback, they can change the translator or the whole team.<\/p>\n\n\n\n Statistics and Charts<\/strong><\/p>\n\n\n\n The system offers statistics per language, turnover, and much more. The invoices \u2013 if applicable \u2013 can be edited and controlled online.<\/p>\n\n\n\n Figure 17: Graphics for turnover, language demand, words translated<\/em> A complex cost analysis system allows the organisation to determine if the team and the whole department are working efficiently. Invoicing<\/strong><\/p>\n\n\n\n There is no universally standardised invoicing system for translation services, which results in significant complexity. Translation agencies are required to submit tenders, while each organisation defines its own highly specific invoicing procedures within the technical specifications document. These requirements often include customised match rate bands for CAT tools, and invoicing preferences that may be based on source or target text, word count, line count (with or without spaces), and numerous other metrics.<\/p>\n\n\n\n TTN TMS, a translation management system developed over more than 30 years, accommodates all known invoicing variations.<\/p>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Selection-of-target-languages-6.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Translation-mode-and-deadline-selection-5-1024x706.png\"){kind=spacer}
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/The-client-can-select-his-own-deadline-6.png\")
<\/figure>\n\n\n\n![\"\"\/](\"\/Info\/wp-content\/uploads\/2026\/01\/Parameterisation-through-registry-or-System-Operator-Interface.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2026\/09\/Workflow-for-classic-translations-v5-1024x366.png\"){kind=spacer}
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2026\/09\/Full-Post-Editing-workflow-v5-1024x418.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2026\/09\/Post-Editing-Light-workflow-v5.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2026\/09\/Automatic-translation-the-AI-check-goes-to-the-client-v5.png\"){kind=spacer}
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Full-text-search-in-translation-archive-6.png\"){kind=spacer}
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/A-three-level-list-provides-fast-and-easy-access-to-the-desired-search-results-4-1024x749.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Translation-archive-with-pictures-in-HTML-in-browser-5.png\"){kind=spacer}
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/A-side-by-side-viewer-allows-analysis-of-target-and-source-texts-4-1024x265.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Users-can-access-and-search-the-TMs-without-third-party-software-4-1024x594.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/MultiTerm-online-database-4.png\")
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Access-to-translator-and-proofreader-profile-6.png\"){kind=spacer}
<\/figure>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Client-feedback-function-with-evaluation-6.png\")
<\/figure>\n\n\n\n
<\/p>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Graphics-for-turnover-language-demand-words-translated-5.png\"){kind=spacer}
<\/figure>\n\n\n\n
<\/p>\n\n\n\n
<\/p>\n\n\n\n![\"\"](\"\/Info\/wp-content\/uploads\/2025\/12\/Extract-of-an-invoice-with-client-specific-data-5.png\")
<\/figure>\n\n\n\n