Software for translation work
Translation software organises reference material, reuses approved translations and terminology, counts work and supports review. Computer-assisted translation keeps a human translator responsible for the text; its value depends on suitable features, usable interfaces and an understanding of what automated checks miss.
Computer-assisted translation
Computer-assisted translation (CAT) covers software ranging from simple aids to complex systems. CAT tools assist a human translator rather than translate independently, although machine translation can be one of their features. Delegating repetitive tasks to the software leaves the translator more time for complex and creative work. Increasing CAT use had radically changed translation over the fifteen years to 2014. In one translator's experience, the tools were in their infancy in the early 1990s and indispensable for business success by 2005.
The category includes the following tools:
- Project management software to structure projects, assign tasks and track progress.
- Grammar checkers built into word processors or installed as add-ons.
- Monolingual and bilingual electronic dictionaries.
- Local or online terminology databases, such as TERMIUM Plus and the Grand dictionnaire terminologique.
- Terminology managers ranging from spreadsheets and FileMaker Pro databases to specialised systems such as SDL MultiTerm, LogiTerm and Termex.
- Translation memories holding source segments and translations in one or more target languages.
- Full-text search tools or indexers, such as ISYS Search Software and dtSearch Desktop, for translated texts and references.
- Concordancers retrieving words or expressions with context from monolingual, bilingual or multilingual corpora.
- Bitext aligners matching a source with its translation to form a searchable or concordance-ready database.
Tools support successive workflow stages:
| Stage | Activity | Tools |
|---|---|---|
| File submission | Project analysis, workload estimation and pricing | Word count and recurrent-segment analysis |
| Project start | Organising references and linguistic guidelines | Alignment, glossaries and concordancers |
| Translation | Translating | Translation memory |
| Editing and proofing | Editing and proofing | Spellcheckers |
| Quality assurance | Quality checks | Translation memory and linguistic guidelines |
| Project close | Archiving the new translation | Updated memory for analysing, translating and proofing future projects |
Translation memory, alignment and terminology
A translation memory stores source sentences and their translations as segment pairs, displayed side by side. It offers earlier translations when identical or similar segments recur and stores new translations when no match exists. The human translator may reject its suggestions. Three main advantages are:
- Faster turnaround and cost savings on repetitive content.
- Consistency and readability, avoiding confusing differences in repeated sentences.
- Automatic scans for mistranslated numbers, simple typos and incorrect punctuation.
A new memory begins empty; alignment software segments completed source and target documents, attempts to match their segments and creates pairs for import. This makes earlier work reusable.
A terminology database stores words and expressions rather than full segments, including organisation-, customer- and product-specific terms absent from standard dictionaries. It searches source sentences in the background so target terms can be inserted without separate lookups; advanced systems check correct term pairings interactively or in batches across a project's memory segments. Its main advantages are:
- Time and cost savings on research and typing long expressions.
- Consistent wording and improved readability without confusing variation.
- Management of customer- and product-specific terminology absent from standard dictionaries.
A typical CAT sequence has seven steps:
- Upload the source text and relevant reference files.
- Analyse and segment the text, usually at sentence-ending periods.
- Translate each segment.
- Check software suggestions, edit and approve the translation.
- Store matched source and target segments in the translation memory.
- Save the translated file.
- Export it in the same or a new format as required.
The benefits include:
- Increased productivity.
- Consistency across translations.
- Better quality control.
- Simpler terminology management.
- Consistent technical terminology.
- Reduced repetitive work.
- More precise pricing through easier word counting.
CAT tools also support layout, translation management and AI integration. Working across documents and looking up terminology electronically leaves more time for writing and avoids retyping repeated expressions.
Usability and selection
Mainly qualitative studies of general translation reported usability problems, with limitations in sample sizes, methods and populations. The reported problems included:
- Learnability: unclear tutorials, handbooks, information and interfaces.
- Helpfulness: difficulty obtaining technical assistance.
- Interface complexity: unintuitive operations and many mouse clicks.
- Layout: poor source–target presentation and overcrowded menus.
- Segmentation: only part of the text visible at once.
- Formatting: inflexible fonts, tags and numbers, difficulties highlighting text or deleting and copying tags and quotation marks.
- Incompatibility: problems with non-Roman character sets, including Arabic and Asian languages.
In one comparison, a more subdivided screen layout produced significantly fewer translated words in the same time, 194 versus 222. Usability problems can interrupt work, create cognitive overload, irritation and anxiety, and deter adoption; these problems are not specific to medical translation. ISO/IEC 25010:2011 defines software-quality and user-experience characteristics, including usability. Learning time and support needs may explain why most CAT users were not freelancers in the literature discussed in 2024.
In 2024 no CAT tool was designed specifically for medical translation, so selection depended on budget, desktop or cloud access, integration with everyday software and supported file types. OCR is essential for translators handling scanned medical or laboratory reports, while compatibility with supplied project packages avoids extra work. Trados SDLXLIFF files were common and supported by many other tools.
Features to compare include:
| Feature | Options and considerations |
|---|---|
| Work system | Desktop tools work offline and are safer against hacking but less flexible for collaboration; cloud tools work across internet-connected devices, save automatically and update and debug faster, but depend on connections and face hacking risks. |
| Price | Paid subscriptions or perpetual licences offer more functions, capabilities and safety; indicative 2024 prices were about €300 annually or €650 perpetually. Free versions offer a more limited introduction. |
| Layout and design | Visible windows and tools should avoid overcrowding and support intuitive operations, reducing interruptions and cognitive fatigue; some tools allow customisation. |
| Terminology management | Term bases hold jargon, words, expressions, acronyms and brands to improve consistency for a client. |
| Project management: collaboration | Multiple translators can work on a project and communicate with stakeholders. |
| Project management: budgeting | Expenditure and revenue calculations support project finances. |
| Project management: reporting and analytics | Progress, volume and other measurable data inform budgets, deadlines and decisions. |
| Support: OCR | Printed or handwritten material becomes editable, helping with scans, uneditable PDFs and photographs. |
| Support: segment analysis | Estimates how much work a project requires. |
| Support: integrated or plug-in machine translation | May speed work in a new domain when correction is faster than translation from scratch. |
| Support: quality assurance | Checks spelling, grammar, number differences, missing tags and format. |
| Import and export formats | Support may include SDL XLIFF, XLIFF, Word, PowerPoint, Excel, RTF and HTML; wider compatibility gives clients more options and eases translation. |
Descriptions, user forums and free trials help assess fit; colleagues and agencies can identify preferred tools, but a colleague's choice may not suit another translator. Client privacy conditions should be checked before online tools are used; private clouds and permission-controlled server tools can support agency collaboration securely. Proper training before adoption is recommended.
Tool examples and project fit
A 2013 survey of over 3,000 translation professionals in different countries found primary use of SDL Trados at 43.2%, Wordfast at 11.5%, MemoQ at 8% and other software at 37.3%. A non-exhaustive 2014 comparison listed:
| Tool | Supported formats | Platform | Licence |
|---|---|---|---|
| MetaTexis | Word, Excel, PowerPoint and text formats | Word add-in | Proprietary |
| MemoQ | Office, HTML, FrameMaker, InDesign, XLIFF, XML, SVG, Typo3 and others | Windows | Proprietary |
| OmegaT | Word, Excel, PowerPoint, XHTML, HTML, OpenDocument, MediaWiki, plain text and others | Cross-platform Java | GNU GPL |
| SDL Trados | Word, Excel, PowerPoint, OpenOffice, InDesign, QuarkXPress, PageMaker, Interleaf, FrameMaker, HTML, SGML, XML and SVG | Windows | Proprietary |
| Wordfast | Word, Excel, PowerPoint, PDF, SGML, HTML, XML, InDesign, FrameMaker, tagged documents and XLIFF | Cross-platform Java | Proprietary |
In 2024 paid examples included SDL Trados Studio, MemoQ, Wordfast, Memsource, Déjà Vu, Across, Wordbee and XTM Cloud; free options included OmegaT, CafeTran Espresso, Smartcat, MateCat and Wordfast Anywhere. One agency favoured Trados Studio for multilingual and back-translation projects because packages were easy to assemble and it covered more languages and subvarieties than Across, including Asian and African languages; it also suited lay-language summaries. Some colleagues preferred Across's usability. Across supplied a server link, while Trados supplied a package for autonomous work. A reported Trados constraint concerned simultaneous quality checks against one memory, although the account did not establish the precise restriction clearly. Both tools had supporters and detractors.
Cost and volume
Different client preferences can require several tools, increasing translators' costs, while learning time varies with the tool's complexity and the learner. Client savings come from memory matches and repetitions, but a useful memory takes several projects to build. Preparing a small file can cost disproportionately much and may not justify a CAT tool, although many small files make one necessary. A client's preferred tool may have few users, requiring the agency to find a solution. CAT tools are necessary in agency work because client terminology and updated memories are held there.
Occasional users can learn a tool on previous projects before a live assignment to avoid delays. In 2024 lower-outlay routes included seven- or thirty-day trials, agency-supplied project licences and monthly subscriptions such as those offered by Trados and memoQ. Free tools were also available, although one occasional-use discussion could not establish their confidentiality implications.
Counting software is essential for analysing content across the many file formats used in translation. In 2014 FineCount offered text statistics across formats, with a free single-computer noncommercial version and a professional version adding website counts, text export, invoices and reports. CATCount was freeware converting complex repetition-based rate schemes into a single adjusted count, also known as leveraged, weighted or equivalent word count. It could print and export the scheme, original counts and resulting figure. Some clients misunderstand CAT's role and ask for lower budgets.
Voice recognition
A 2014 assessment put human speech at five to seven times typing speed and cognitive processing at ten to fifteen times, with translating closer to those speeds as simultaneous interpreting illustrates. Voice recognition had matured enough for processing to keep up with speech, but professional-quality support had to exist in the target language. Dragon Naturally Speaking version 12 supported US and UK English, French, German, Italian, Spanish, Dutch and Japanese. Dictated translations remain drafts needing review and editing, although practice reduces editing effort; verbalising a translation also calls for an interpreter's disposition.
Review and automated checks
In 2014 ApSIC Comparator was freeware producing side-by-side reports of reviewers' changes across formats for quality evaluation and translator feedback. ChangeTracker similarly reported bilingual changes, allowed comments and error codes and exported to Excel.
Translation quality assurance software compares bilingual segments for:
- Inconsistencies.
- Departures from a project glossary.
- Omissions.
- Target segments identical to their source.
- Punctuation, capitalisation, numerical and formatting errors.
- Incorrect untranslatables and tags.
Such checks cannot detect misunderstandings of the source, poor style or inappropriate register, and may assume an incorrect source is correct. In 2014 ApSIC Xbench combined QA functions in one view, supported many CAT formats and spellchecked 33 languages. Its predefined checks covered completeness, consistency, numbers, tags and key terms; custom checklists and hundreds of simultaneous files allowed editing only files with spelling problems.
Automated QA can compare repeated-segment translations, term-base compliance and consistency with earlier memory translations, as well as punctuation, spaces, capitals and segment length, with custom checks for recurring problems. Number matching helps check dosages but is not infallible: German “im 28. Lebensjahr”, meaning in the 28th year of life, correctly becomes “27 years of age” and may be falsely flagged. Fully automated checks cannot ensure accurate source meaning. Errors in reused terminology can spread through a project, so terminology databases should be cleaned before work begins, a step often missed under time pressure.