Terminology Workflow in Practice: From Extraction to Approval

Why a Terminology Workflow Makes Localization More Stable

A structured terminology workflow prevents the chaos that arises when translators, subject-matter experts, and market experts use different terms for the same concept. In practice, a lack of terminology control leads to inconsistent translations that confuse customers and weaken brand perception. A well-designed workflow, on the other hand, ensures that everyone involved is on the same page – from initial extraction to final approval.

Without a workflow, typical problems arise: A technical term is Germanized by the subject-matter department, while the market expert prefers a more easily understood expression from the customer's perspective. The translator then often resorts to a third variant due to a lack of guidelines. The result: The manual says "Startknopf," the online help says "Einschaltknopf," and the packaging says "Power-Taste." Such inconsistencies appear unprofessional and can even pose risks for safety-critical products. A defined workflow forces these conflicts to be identified and resolved early – before the material goes to the customer.

Recommendation for action: Implement a centralized terminology workflow that includes the steps of extraction, evaluation, consolidation, and approval. Use a shared terminology management system (TMS) in which all entries have a status (e.g., "under review," "approved," "obsolete"). Define that only approved terms may be used in translations. Communicate the workflow clearly to all parties – from the subject-matter department to external translators. Once established, the workflow significantly reduces correction effort and ensures consistent technical language across all channels.

In practice, companies with a fixed terminology workflow require up to 30% fewer translation revisions. While the initial effort for setup should not be underestimated, the long-term stability and efficiency gains justify the investment. Start with a pilot project for one product or product line and then gradually expand the workflow.

The Key Roles: Terminologist, Subject Matter Department, Market Expert, and Translator

A successful terminology workflow relies on clearly defined roles. Four actors are central: the terminologist, the subject matter department, the market expert, and the translator. Each brings specific knowledge, and only through collaboration does a consistent and target-audience-appropriate terminology emerge.

The terminologist is the guardian of the workflow. They extract terms, maintain the database, moderate conflicts, and ensure compliance with standards. In practice, they often possess linguistic and subject-matter expertise. Their task is to evaluate suggestions from other roles and organize them into a consistent system. However, they do not have the final say—decision-making authority lies with the subject matter experts.

The subject matter department (e.g., development, product management) defines the correct specialized language. They know the products and processes best and determine which term is technically accurate. In conflicts with the market expert, who advocates for customer understandability, the subject matter department must compromise—for example, by including synonyms or defined preferred terms for different channels. The market expert (e.g., from marketing or sales) brings the customer's perspective. They know which terms are established in the target audience and which lead to comprehension problems. Their input is especially crucial during the initial market launch in a new language region.

Finally, the translator implements the terminology. They may only use approved terms and are obligated to use the workflow for doubtful cases rather than making unilateral decisions. In practice, this often fails due to lack of training or missing access to the TMS. Therefore, it is essential to involve translators in the workflow and grant them read access to the approved terminology.

Action recommendation: Define responsibilities and decision-making authorities in writing for each role. Introduce regular coordination meetings where conflicts between the subject matter department and market expert are resolved jointly. The terminologist should have veto power over inconsistent or non-compliant proposals. Ensure translators have a defined point of contact for terminology questions—ideally the terminologist.

Extraction of Specialized Terms from Sources: Tools and Methods

The extraction of specialized terms is the first step in the terminology workflow and determines the quality of the entire data foundation. The goal is to collect all relevant terms from sources such as technical documentation, specifications, glossaries, and existing translations. There are both manual and automated methods, which complement each other depending on the scope and complexity of the project.

For manual extraction, Excel lists or simple text editors are suitable. In practice, the source material is searched line by line for specialized terms. This method is time-consuming but useful for small projects or when the material is highly domain-specific. The advantage: the terminologist can immediately capture contextual nuances. The disadvantage: it is error-prone and hardly scalable. Therefore, professional localization projects rely on specialized tools such as SDL MultiTerm Extract, memoQ Term Extraction, or Sketch Engine. These tools automatically analyze corpora and suggest candidates based on frequency, co-occurrence, and domain-specific patterns.

Automated extraction often yields high recall but also produces many false alarms (e.g., general-language words). Therefore, manual verification is essential. A combination has proven effective: first, a candidate set is created automatically, then the terminologist checks each candidate for relevance and adds context, definition, and source. In practice, the tool should allow linking to the source sentence so that the origin of the term can be traced later.

Action recommendation: Use a tool that integrates both extraction and management of terms in one system. Define which sources are prioritized before extraction (e.g., standardized documents over internal emails). Define criteria for when a term is included: at least three times in different contexts or once in an authoritative source? Ensure that all extracted terms receive a unique status ("candidate", "under review", "approved"). Start with an initial extraction from the most important reference document and gradually expand the database with additional sources. This avoids data clutter and maintains control over the workflow.

Evaluation and prioritization: Which terms need to go into the glossary?

After extraction, you are often faced with an extensive list of potential terms. Not every term is equally relevant for your glossary. Systematic evaluation and prioritization help you use resources efficiently and improve translation quality where it is most needed.

Start by categorizing according to technicality and frequency. Technical terms that appear regularly in your source texts and are essential for product understanding should have the highest priority. For example, in medical technology, the term "catheter" would be high priority because it appears in many documents and requires an unambiguous translation. Less critical terms, such as "control element" in an instruction manual, can be dealt with later. Use a simple matrix for this: combine the factors "frequency of occurrence" and "technical relevance" (e.g., low, medium, high). Terms that are high in both dimensions move directly into the definition phase.

Another criterion is the potential for confusion. If a term is ambiguous in the source text or is translated differently in different contexts, you should prioritize it. Also check whether terms are already defined in existing glossaries or style guides. Duplicate work can be avoided by checking existing entries for currency. In practice, it has proven useful to review the list with the terminologist or a subject matter expert and make a preselection. Then priorities are set in a short meeting with the specialist department and market experts. This ensures that all perspectives are incorporated and that the most important terms are processed first.

Recommendation: Create a simple traffic light system (red = immediate, yellow = next round, green = later). Document the decision criteria in a traceable manner. This avoids discussions in later phases and creates transparency within the team. Remember: An overloaded glossary is difficult to maintain. Better to start with 20 critical terms and expand gradually than to manage 200 unchecked entries.

Definition of Terms: Avoiding Ambiguity and Ensuring Precision

Once you have identified the prioritized terms, it is time to define them. The goal is to formulate a clear, context-based description for each term that all stakeholders—translators, technical departments, and market experts—can understand equally. A good definition prevents misinterpretation and creates a binding foundation for translation.

Start with the source language: note the term, its part of speech, and a precise application example from your original texts. Avoid vague formulations like "a device for measurement." Instead: "A digital thermometer for continuous body temperature monitoring with Bluetooth connectivity." The more specific, the better. Then add the target language definition, which describes the same concept in the cultural context of the target market. For example: the German term "Kostenvoranschlag" is often referred to as "Kostenangebot" in Austrian German. Your definition should capture this regional difference without changing the meaning of the term.

A common issue is ambiguity. If a term has multiple meanings, create a separate glossary entry for each meaning and mark the context. Use semantic tags such as "Technology" or "Finance." For instance: "Bank" can mean a financial institution or a piece of furniture. In your glossary, separate these cases and provide a source reference for each. Synonyms are also critical: avoid them by establishing a preferred term and marking alternatives as "do not use." In practice, it has proven beneficial to validate the definitions in a short review with the technical department and a native speaker of the target market. While this takes time, it significantly reduces later corrections.

Action recommendation: For each entry, define an example scenario in which the term appears. This makes it tangible for translators. Also, use a consistent structure: term, part of speech, definition, context, application example. Determine when a term is considered "approved"—ideally when at least one subject matter expert and one market expert have agreed. This prevents unverified definitions from entering the translation workflow.

Conflict Case: When the Technical Department and Marketing Insist on Different Terms

In localization practice, it often happens that the technical department prefers a technically correct term, while the market expert insists on a more memorable, audience-appropriate formulation. This conflict can block the workflow if not resolved systematically. A structured conflict resolution process is therefore essential for smooth terminology management.

The first step is to analyze the arguments. Have both sides present their positions in writing: the technical department explains why the term is technically precise and what risks deviation entails (e.g., legal consequences or misunderstandings among technical users). The market expert justifies why the alternative term is better understood or fits the brand strategy. In practice, it has proven useful to create a brief pros and cons protocol. An example from the automotive industry: the technical department insisted on "Kraftstoffrücklaufleitung" (fuel return line), while the market expert preferred "Kraftstoffrückführung" (fuel return). The analysis showed that the technical term is indispensable in workshop manuals, while marketing texts should use the shorter version.

The second step is to seek a differentiated solution. Often, terms can be assigned based on context: in technical documents, the technical term is used; in marketing materials, the market-friendly variant. In the glossary, you then record both terms with a context specification. Important: you must clearly define which context corresponds to which term. If agreement is not possible, the terminologist decides based on defined criteria—such as target audience, document type, or legal requirements. Establish an escalation process for this: in case of disagreement, the issue goes to a higher authority (e.g., product management or legal department). This prevents endless discussions.

Action recommendation: Define how conflicts are resolved in your terminology workflow—ideally by making a clarification in the glossary with context-related synonyms. Document every decision and its rationale. This creates traceable precedents and prevents the same conflict from recurring. Important: consult legal advice if terms are relevant for regulatory or trademark reasons.

Mediation and Decision Matrix: How to Resolve Terminological Disputes

In practice, terminological conflicts arise when specialist departments and market experts prefer different terms for the same concept. The specialist department insists on technical accuracy, while the market prioritizes comprehensibility or local conventions. Without structured mediation, such disputes can block the workflow. A proven method is the decision matrix, which weights criteria like target audience, usage context, legal requirements, and search engine relevance.

Build the matrix as follows: For each disputed term, define three to five evaluation criteria, for example “Technical Correctness”, “Target Audience Understanding”, “Consistency with Existing Glossary”, “SEO Potential”, and “Legal Admissibility”. Assign scores from 1 to 5 for each criterion. Have representatives from the specialist department and marketing evaluate independently. Then calculate the average. The term with the higher total score is preferred. For example, for a medical term, “Technical Correctness” might be weighted double, while for a marketing text, “Target Audience Understanding” counts more.

Documenting the decision is crucial. Record why a term was chosen to avoid later queries. If the matrix yields no clear preference, involve an independent terminologist as mediator. They will moderate a workshop where all parties present their arguments. Often a compromise is possible: for example, including both terms with different scopes – the technical term in internal documentation, the market-friendly term in customer communication.

Avoid lengthy escalation. Set a deadline for agreement, e.g., two weeks. If the conflict persists, the responsible editorial manager decides based on the matrix. This approach ensures that terminological decisions are made transparently and objectively. Note: For legally relevant terms (e.g., in medical or financial sectors), obtain legal advice before the final decision.

Designing efficient approval loops: Who reviews when and how?

Approval loops are a critical point in the terminology workflow, as they can consume time and resources. In practice, a graduated review sequence that avoids redundant work has proven effective. Start with a specialist review by a terminologist, who checks the proposed terms for correctness and consistency with the existing glossary. Then follow with a market review by a local market expert, who evaluates the terms for target audience suitability and cultural appropriateness.

Define binding processing deadlines: a maximum of three working days for the specialist review, and another three working days for the market review. Use a ticketing system or collaboration platform (e.g., in your CAT tool or a separate workflow tool) to make the status traceable. Each reviewer receives a notification when an entry is pending review. After both reviews are completed, the entry is automatically forwarded to the next instance – in case of conflicts to the mediator, otherwise to the final approver (e.g., the editorial manager).

Reduce the number of approval stages to the necessary minimum. Experience shows that two review stages (specialist and market review) followed by a final approval suffice. Avoid sequences with more than four stations, as these unnecessarily prolong lead times. Specify that only authorized persons may make changes to already approved terms – and that any change triggers a new approval loop. Document the entire review process in a change log.

Communicate the workflow rules clearly to all parties involved. Provide a brief guide describing roles and responsibilities. Regularly check whether the approval loops are still efficient, for example by analyzing lead times. Adjust capacities when bottlenecks arise. A well-structured approval process prevents terminological decisions from being unnecessarily blocked and ensures fast, coordinated results.

Integration of glossaries into CAT tools and AI translation systems

A glossary only unfolds its full effect when seamlessly integrated into translators' tools. Modern CAT tools (Computer-Aided Translation) and AI translation systems support importing glossaries in CSV, Excel, or TBX format. For smooth integration, structure your glossary according to a uniform scheme: term, definition, context, part of speech, gender (for languages with grammatical gender), and preferred translation per target language. Avoid special characters that could cause import errors.

In CAT tools such as memoQ, Trados, or Across, import the glossary as a term database. Assign the corresponding metadata to each term. Enable automatic term recognition: when a translator enters a glossed term in the source text, the preferred translation is displayed. This saves time and increases consistency. For AI translation systems (neural machine translation), you can upload glossaries as so-called "term sets" or "custom glossaries." The AI takes these specifications into account during translation and deviates less frequently from the desired vocabulary.

Test the integration before productive use. Translate a small text corpus and check whether the glossed terms are used correctly. If there are deviations, adjust the formatting. In practice, regular updates of the imported glossary are necessary: synchronize the glossary at least once per quarter with the current status from your terminology workflow. Use an interface for this or export the glossary again.

An important note: Glossaries in AI translation systems do not replace human post-editing. The AI cannot always reliably apply the specifications, especially for homonymous terms. Therefore, always have translations containing glossed terms reviewed by a native speaker. For legally sensitive areas (e.g., patents or medical devices), we additionally recommend legal advice from a specialized attorney. With a well-founded integration, you ensure that terminology remains consistent across all output channels.

Versioning and Change Management: Ensuring Traceability

A terminology database lives – it is supplemented, corrected, and adapted to new products or markets. Without traceable change management, inconsistencies arise that directly impact quality in localization. In practice, it has proven effective to provide each entry in the glossary with a unique version stamp documenting the date, author, and type of change. This can be done via a simple field in the glossary or through the change history of the terminology management system (TMS) used.

For change management, you should define clear rules: Who is allowed to create new terms? Who may edit or delete existing entries? In practice, it is recommended to set up a role matrix – for example, with the roles "terminologist," "subject matter expert," and "administrator." Ideally, each change goes through a mini-workflow: suggestion, review, approval. Even minor corrections such as typos should be documented, as they can indicate a recurring problem.

A central issue is synchronization between different systems: CT, CAT tool, AI translation platform. If you use the glossary in multiple tools, you must ensure the versions match. In practice, a regular export function to a neutral format (e.g., TBX or CSV) and an automated check for deviations have proven useful. Also plan a monthly or quarterly review of changes, during which you go through the history for anomalies.

Traceability also includes logging decisions: Why was a term discarded or replaced by another? This can be recorded in a side field or a separate log file. This prevents the same discussion from recurring. Remember: Without versioning, every team change becomes a risk – new employees have no insight into the rationale behind earlier decisions. So invest in clean change management; it is the foundation for lasting terminology quality. (Legal notice: The procedures described in this text do not constitute legal advice; consult your legal counsel for legal questions.)

Updating the Terminology Database: Frequency and Responsibilities

A static glossary quickly becomes outdated – new products, changed legal regulations, or regional language trends require regular updates. In practice, a fixed cycle of three to six months has proven useful, supplemented by ad-hoc reviews for new launches or rebrandings. Responsibility for the update process should be clearly assigned, for example in a routine initiated by the terminologist, with subject matter experts and market experts participating.

Specifically, you can establish an annual rhythm with two review sprints: one mid-year (focus on new terms from projects) and one before year-end (focus on deletions and corrections). In between, only urgent cases are updated ad hoc. Set a next review date for each entry so that old entries are automatically reminded. In practice, such a date is maintained in fields like 'last review' and 'next review' in the glossary.

Updating involves not only new terms but also deleting obsolete entries – a frequently neglected point. Mark terms that have not been used for a while as 'inactive' and set a deadline (e.g., one year) after which they are permanently removed. Document the reason so that it remains traceable why an entry disappeared. Once a year, also check the completeness of the glossary by comparing it against your current customer or product documentation.

To ensure the update does not get lost in day-to-day business, it is helpful to block a fixed date in the calendar and inform participants early. Small teams can integrate the update into their weekly routine – for example, as an item in the regular meeting. It is important that the process does not start only when inconsistencies arise. Establish a continuous improvement process where anyone on the team can submit change requests at any time. (Legal notice: The procedures described in this text do not constitute legal advice; consult your legal counsel for legal questions.)

Quality Assurance: Spot Checks and Automated Rule Checks

Even the best terminology management is of little use if the entries are not correct or not applied consistently. Quality assurance (QA) ensures that the glossary is both content-accurate and technically error-free. In practice, companies combine two approaches: random manual checks and automated rule checks in translation tools.

For manual QA, a monthly sample of 10 to 20% of the total inventory is advisable, where you check entries for completeness, unambiguous definitions, and correct source references. If possible, have the samples reviewed by a second terminologist or a subject matter expert who was not involved in creating the entries – this helps uncover blind spots. A simple traffic light system (green = ok, yellow = anomalies, red = errors) helps visualize the status.

Automated checks can be integrated into your CAT tool or AI translation platform. Typical rules are: - No entry without a definition - No entry without a source reference - No duplicate entries for the same term - Consistent use of upper/lower case according to house rules - Link to related terms present (optional) These rules can usually be implemented via QA plugins or custom scripts. In practice, it has proven effective to check the rules directly when creating a new entry and issue a warning if violated – this reduces subsequent errors.

Another element of QA is checking glossary usage in translations. Using concordance queries in the TM, you can determine how often a term appears in the target texts and whether it is used consistently. If actual usage deviates from the glossary entry, you should take action: either the entry is wrong or the translators are not trained. Therefore, plan a quarterly short report listing the deviations. This way, QA is not limited to the glossary itself but ensures it is also alive in practice. (Legal notice: The procedures described in this text do not constitute legal advice; consult your legal counsel for legal questions.)

Checklist: How to Establish a Smooth Terminology Workflow

A stable terminology workflow does not run itself; it requires clear processes and responsibilities. The following checklist summarizes the essential steps you should consider when setting up or optimizing your workflow. Go through the points one by one and adapt them to your corporate structure.

1. Define roles and responsibilities: Determine who is the terminologist, who is the subject matter expert, and who is the market expert. Assign a responsible person for each area and document substitution rules. In practice, it has proven effective to record decision-making authority in writing in a governance document.

2. Standardize the extraction and evaluation process: Specify from which sources (e.g., technical documentation, marketing materials) terms are extracted. Use tools such as term extraction in CAT systems or specialized terminology management systems. Establish criteria for prioritizing terms—such as frequency, relevance for translation, or potential for misunderstandings.

3. Provide definitions and context: Each entry should include a precise definition, an application example, and, if necessary, a note on synonyms or forbidden terms. Work closely with the relevant department to ensure technical accuracy. Avoid ambiguities by clearly distinguishing similar terms.

4. Establish approval loops and conflict resolution: Define who must approve which changes. In case of conflicts between the technical department and market experts, use a decision matrix that considers criteria such as technical precision, comprehensibility for the target audience, and consistency with the existing glossary. Conduct regular review meetings to resolve open issues.

5. Ensure integration into the translation environment: Make sure your glossary is integrated into CAT tools and AI translation systems. Configure terminology checks that warn translators when they violate established terms. Allocate time for maintaining interfaces.

6. Introduce versioning and change tracking: Use a system that logs every change to the terminology database. Document who made which change when and for what reason. This facilitates traceability and enables targeted rollback in case of incorrect decisions.

7. Conduct regular quality checks: Perform spot checks at defined intervals to verify the correct application of terminology in translations. Additionally, automated rule checks can uncover inconsistencies. Plan a revision of the glossary after every major product launch.

Outlook: Terminology Work in the Age of AI and Agile Localization

Terminology work is facing a transformation: AI-powered translation systems and agile localization processes are changing the requirements for glossaries and workflows. Companies must adapt their terminology processes to benefit from the new technologies without losing control over language quality.

Artificial intelligence can support the extraction and maintenance of terminology: machine learning models recognize recurring technical terms in large volumes of text and suggest candidates for the glossary. In practice, however, human review remains indispensable, as AI does not always correctly assess contextual nuances or company preferences. Nonetheless, AI-powered suggestions can accelerate workflows—especially when initially populating a glossary.

Agile localization means short iterations and frequent releases. Terminology must therefore be updatable in real time. Traditional workflows with weekly approval loops are often too slow. Instead, successful teams rely on decentralized maintenance: subject matter experts and market experts edit terms directly in the terminology management system, and changes are automatically transferred to the translation environment. The role of the terminologist shifts toward quality assurance and governance, while operational maintenance moves to the specialist departments.

Another trend is the integration of terminology into AI translation models: companies feed their glossaries as "contextual examples" or via APIs into neural machine translation systems. This enables consistent translation of company-specific terms. However, note that systems respond differently to terminology specifications. Test how your system applies glossary entries before rollout, and define rules for handling unknown terms.

Finally, collaboration between humans and machines is being redefined: translators correct AI suggestions and feed the corrections back into the glossary—a cycle that continuously improves terminology. Ensure that this feedback loop is structured and does not lead to uncontrolled growth. A central steering committee should validate changes before they take effect for all translation projects. Even if AI simplifies many tasks, the strategic management of terminology remains a human task requiring expertise and organizational sensitivity.

Common Objections to Terminology Work and How to Address Them

In practice, terminology projects often face skepticism. A typical objection is: "That's too time-consuming – we don't have time for that." This overlooks the fact that the initial effort pays off many times over later. It's better to start with a small, focused glossary for the products or campaigns with the highest translation volume. This way, you quickly demonstrate measurable success, such as fewer queries from translators or more consistent texts. Another common objection is: "Our departments can't agree on terms." This is exactly where a moderated workflow with a decision matrix helps to resolve blockages. Some managers fear that rigid specifications will stifle creativity. However, in practice we find that defined core terms provide a stable foundation for writing – even for marketing texts – without limiting creative freedom. And then there is the argument: "We've always done without it, and it works." But anyone who compares the costs of rework, lost brand presence due to inconsistent messages, or legal risks from incorrect technical terms quickly realizes: terminology work is not an end in itself, but quality assurance. A good way to counter objections is a pilot project: choose a glossary of 50 to 100 terms, document the effort from extraction to approval – and compare the error rate in subsequent translations. With this data, you argue based on facts. Remember: external service providers also benefit from clear terminology, as it saves them research time. Regulate cooperation in service level agreements (SLAs): glossaries are exchanged at project start and synchronized regularly in case of changes. This avoids discrepancies. Note that the legal binding nature of definitions in your company should be reviewed by the legal department – especially for protected brand or product names. With these arguments and a step-by-step introduction, you can address most reservations.

Budget and effort: Realistically assessing cost factors and savings potential

Introducing a terminology workflow incurs initial costs – for personnel, tool licenses, and training. However, these costs can be limited through clear planning. Experience shows that creating an initial basic glossary of around 200 terms takes between 40 and 80 working hours, depending on the complexity of the technical language. This is distributed among terminologists, subject matter experts, and market experts. The bulk goes to extraction and evaluation – here, automated extraction tools found in many translation management systems (TMS) help. Ongoing maintenance requires about 10 to 20 hours per quarter, depending on the rate of innovation. On the other hand, there are savings: translators can work faster, and rework and correction loops typically decrease by 20 to 30%, measured against typical correction rates in localization projects. CAT tool efficiency also increases because consistent terminology yields higher overlap rates in translation memories. From a legal perspective, it is advisable to observe the mandatory separation of consulting and legal advice: the figures given are empirical values, not fixed guarantees. When working with service providers, you should request modular offers: pure glossary creation, workflow setup, or ongoing maintenance. A typical model is for the service provider to keep the terminology up to date while you manage approval internally. Ensure that license costs for specialized terminology tools are not underestimated – they can range from 50 to 200 euros per user per month. Open-source alternatives like termWikis are possible but require more internal effort. Also plan training for your translators and subject matter experts: everyone working with the glossary must be able to use it. A short workshop – online or in person – of two to three hours is often sufficient. Keep in mind: the better terminology work is integrated, the lower the long-term costs for quality assurance and error correction. The return on investment becomes apparent after one or two project cycles – when team members use the glossary as a natural tool.

Practical pitfalls and how to avoid them

Even a well-thought-out terminology workflow can encounter unexpected hurdles in practice. A common mistake is incomplete extraction: if only obvious technical terms are captured, product- or industry-specific neologisms are overlooked. Avoid this by including not only documents but also emails, tickets, and chat logs from product development. Another pitfall is over-definition: not every term needs to be included – focus on high-frequency, ambiguous, or difficult terms. Define criteria for when a term is glossary-worthy (e.g., appearing in more than three sources or known translation issues).

A typical problem when working with service providers is differing glossary formats. Standardize the output in TBX or a CSV schema that your CAT tool and the service provider can import. If clear freezes for changes are missing, conflicts can arise between review cycles and ongoing translations. Introduce a defined cut-off: after translation starts, terms are only collected in a separate "holding queue" and integrated into the glossary only after project completion.

Traceability also often suffers: without a comment field for each term entry (e.g., reason for decision, source), later adjustments become a guessing game. Implement a change log with timestamp and responsible person. Finally, overly high expectations of automation can lead to disappointment: AI-supported extractions provide suggestions but do not replace expert review. Therefore, always plan for a manual review step that takes about 20% of the extraction time. With these precautions, you avoid typical stumbling blocks and keep the workflow efficient.

Step-by-Step Example: Terminology Workflow in a Fictional Project

Imagine a medical technology company is launching a new ventilator. Goal: a German-English glossary with 80 relevant terms. The workflow starts with extraction: the terminologist scans the German product specification, clinical studies, and service manuals using an extraction tool (e.g., Sketch Engine). Result: 300 candidates. Next, evaluation: together with the technical department (developers, product managers), irrelevant general terms are removed. 120 terms remain, including "Tidalvolumen" and "PEEP". Definition: a brief explanation is set for each term, e.g., "Tidalvolumen: amount of air per breath on the ventilator, standard range 6–8 ml/kg". The market expert (US subsidiary) checks the English equivalents and finds: "Tidalvolumen" is used as "tidal volume" in US English, but "PEEP" (positive end-expiratory pressure) is accepted but not spelled out. Conflict: the technical department wants to keep the abbreviation, the market expert advocates for the spelled-out form on first occurrence. The decision matrix is applied: since doctors in the US use the abbreviation, the terminologist decides on "PEEP (positive end-expiratory pressure)" at first mention, then "PEEP". Review cycle: first, the terminologist checks the entries (definition, context, source), then the technical department (technical accuracy), finally the market expert (linguistic appropriateness). Duration: two weeks. Subsequently, the glossary is imported as a TBX file into the CAT tool (e.g., SDL Trados). The translator sees a note with definition when working on "PEEP". After the first release update: the product receives a new feature; the terminologist extracts again, adds five new terms, and goes through the workflow again. This keeps the glossary consistent and the translation process stable.