Recommendation: Enable the Glossary Generator now to lock in terminology across translate tasks, especially when you load ファイル or pull content from источник, and use the chrome extension for quick access.

What’s new: The Glossary Generator analyzes your источник content and word tokens, builds a glossary up to 50,000 terms, and preserves バグ修正 terminology across english translations. It supports import from ファイル, and export as CSV for reuse in other files or teams. Glossaries can be reused with claude and claudegemini references to keep terminology aligned.

How to use: Open DeepL, select Glossary Generator, import your ファイル, choose target languages including english, and configure authorization settings for team access. Pair translations with str_translated mappings to ensure consistency across translate tasks.

Tips for Chrome users: Install the chrome extension and enable automatic glossary application during translate tasks. Sync glossaries with your источник data and share to your team through controlled authorization channels; this helps maintain alignment across English and other languages in real-time.

With these improvements, you gain faster, more consistent translate results and better control over terminology, powered by DeepL’s Glossary Generator and broader feature set.

DeepL Updates: Glossary Generator, Feature Enhancements, and Generative AI Translation Considerations

Implement the Glossary Generator now to lock terminology across translations. Use deepl-auth-key for authorization and include it in the Authorization header for each requestsposturl call. Set str_origin to english to anchor the source language, and store glossaries in a dedicated filename. Build entries with fields for term, context, and str_translated, and keep the ファイル organized for quick review. Track greent翻訳性能の改善 in your reports to show concrete gains over time.

Glossary Generator and Data Flows

Generative AI Translation Considerations and Quality

How to leverage the Glossary Generator to standardize terms across projects

Build a master glossary as the single source of truth. Create glossary_master.csv with fields: word, str_origin, entries, filename, and source. Populate canonical terms and store str_origin to capture provenance; attach a concise context and a sample translation for each entry. Keep the file in a shared location (ファイル) and tag機能追加 notes for future refinements.

Audit current projects to identify gaps. Pull existing glossaries, map synonyms to the canonical term in the master list, and consolidate variants by adding them to the entries array. For each term, tag each variant with its language and the source term (источник) to preserve the lineage of every translation.

Leverage the Glossary Generator to produce per-project glossaries. Use glossary_master.csv as the source, select target languages, and apply a consistent naming scheme (filename) for outputs. The generator will retain the canonical word, while delivering translate-ready variants and a clear str_origin trail for translators.

Automate updates through requestsposturl. When new terms appear, push a patch to the project glossary and authenticate with deepl-auth-key via the authorization header. Enable incremental changes and maintain a changelog that includes バグ修正 and 機能追加 notes for traceability.

Bridge browser workflows with chrome and local files (ファイル) to speed edits. Include 東京での説明会案内 as a cross-team alignment example, and invite contributors to review changes in a shared glossary workspace. Integrate Claude and claudegemini as additional engines to validate terms and expand coverage without breaking the canonical mapping.

Quality checks translate into concrete gains: run a targeted test set, compare results against the master glossaries, and track greent翻訳性能の改善 metrics. When mismatches appear, adjust the corresponding entries, update the source notes (источник), and re-run the generator to publish refreshed glossaries.

Keep a long-term guard against drift by making glossary reviews a routine: schedule quarterly term audits, collect feedback from translators, and log all 機能追加 and バグ修正 activities in the glossary history. The result is consistent terminology across projects, with predictable translate behavior and auditable provenance.

Example entry demonstrates handling complex phrases: いびきをかいて眠っている彼に気づかれないようそっとベッドから抜け出し着古した祭服の袖に手を通す. Store this string as a glossary entry with its str_origin, include a concise context, and link it to a sample translation to prevent drift in multi-language outputs.

Importing, editing, and reusing glossary terms to ensure consistency in multiple documents

Import terms from a master glossary file (ファイル) and map them to str_origin and str_translated, then propagate translations across documents via the glossary. Use entries and translations to keep every occurrence aligned, and reference the filename to organize each document variant. Capture origin in источник for auditing, and leverage deepl-auth-key for API-driven updates. Review terms in chrome, and use claude or claudegemini as alternative engines when needed.

Data model and workflow

Practical implementation and examples

  1. Set up a master glossary file (ファイル) with fields: str_origin, str_translated, translations, entries, filename, and источник.
  2. Fetch updates via requestsposturl, apply deepl-auth-key for translations, and validate results in chrome before publishing.
  3. For reuse, tag each term with a target document in filename and attach related entries so edits propagate to all linked docs.
  4. Include sample terms such as 東京での説明会案内 and the long phrase いびきをかいて眠っている彼に気づかれないようそっとベッドから抜け出し着古した祭服の袖に手を通す to ensure consistent handling of multilingual strings.
  5. Monitor quality with greent翻訳性能の改善 notes and address バグ修正 promptly to avoid regression in downstream documents.

What’s new: breakdown of translation memory, batch processing, and API improvements

Recommendation: enable glossary-driven translation memory for all projects and pair it with the enhanced batch processor to cut average processing time and improve term consistency across the board.

Translation memory now breaks down into exact matches, high-confidence fuzzy matches, and lower-confidence reuses. The glossary links terms directly to translations, so entries like glossary terms surface in translations even when the surrounding sentence changes. In practice, memory contains 2.4 million entries, with exact matches at 18% and high-confidence fuzzy matches at 44%; low-confidence fuzzy matches account for 28%. This mix boosts consistency on terms such as 東京での説明会案内 and 各ファイルの専門語, while reducing post-edit workload on long translations. The system also associates each term with a source (источник) and supports multiple languages, including english and japanese phrases, so str_origin and str_translated align with the glossary’s intent.

Batch processing scales to handle larger workloads without increasing latency. Use batches up to 2,000 files per request or as many as 50,000 segments, with streaming for long documents. For mixed content, start with 200–300 files to monitor memory usage and then ramp up. Batch-level controls connect to the API via requestsposturl and pass filename or ファイル references to keep uploads organized. When you reference a term in a batch, the response returns str_translated alongside the original word (word) and its translated counterpart in english or target languages, making QA straightforward.

API improvements deliver faster, more reliable integrations. New endpoints support glossary-driven translations and batch submissions, while latency dropped from ~450ms to ~120ms under typical loads. Authentication uses deepl-auth-key with header-based authentication, and you can pass str_origin to preserve the request source. The response payload includes translations, entries, and glossary annotations, plus fields like str_translated and str_origin to simplify downstream processing. You can upload files with a filename parameter and reference ファイル arrays, then fetch translations and glossary-consistent results in a single call. The updated docs also cover how to attach glossaries to specific projects and how to reuse a given glossary across translations, including edge cases for CSS/HTML files and content with embedded code.

Glossary management receives a dedicated upgrade: add or update terms via a simple entries-based workflow, and see automatic term alignment in translations. For example, a glossary term set can cover industry jargon and product names (e.g., claude, claudegemini) and be reused across languages, ensuring translations stay aligned with your brand. The system records each action in a changelog, making it easy to track機能追加 and バグ修正 across releases. When you export or share results, the glossary and translations reference the same filename and file type, so downstream tools stay synchronized in Chrome and other environments.

Optional test data demonstrates how the feature set behaves in practice. Use sample phrases like "translate," "translations," and "word" alongside long strings such as いびきをかいて眠っている彼に気づかれないようそっとベッドから抜け出し着古した祭服の袖に手を通す to validate glossary matching and context handling. Include phrases like 東京での説明会案内 and istокий пример to verify multi-language routing. For real-world runs, include a test file with filename="sample.txt" and content in English and Japanese to confirm str_translated and str_origin are accurate. Additionally, reference "requestsposturl" to confirm batch submission wiring and "deepl-auth-key" validity in secure calls.

Guidelines for safe Generative AI translations: QA checks, glossary alignment, and human review

QA checks validate translations against source terms and consistency rules. Run automated checks for term capitalization, numeric formats, and style consistency across english and translations. Maintain a glossary with entries that map str_origin to str_translated, and store each term as a separate word with its filename and context. Verify that the translation pipeline uses the deepl-auth-key and authorization tokens properly, and that each request employs requestsposturl in a sandbox before production. Track results for translations and flag mismatches for rapid remediation.

Glossary alignment keeps terminology synchronized across languages. For every term in the glossary, confirm that translations match the canonical gloss entries, including special terms like 東京での説明会案内, which should be mapped to its target-language equivalent and reflected in translations. Maintain the источник as the canonical source indicator and ensure filename, word, and glossary fields are aligned. Export patches with バグ修正 notes and 機能追加 updates when changes occur, and attach the str_origin and str_translated pairs to demonstrate alignment.

Human review assigns bilingual reviewers to spot subtle context errors, cultural nuances, and policy constraints. Reviewers compare source and translated sentences in chrome, verify that formatting and tone remain appropriate, and confirm that risk-sensitive phrases are translated accurately. Include a manual check for the phrase いびきをかいて眠っている彼に気づかれないようそっとベッドから抜け出し着古した祭服の袖に手を通す to ensure meaning and context are preserved. Use a clear log that records reviewer notes, along with requestsposturl, filename, and word-level feedback. Document improvements in the glossary and flag needs for 新機能 (機能追加) or バグ修正 in the release notes.

Upgrade path for teams: licensing, onboarding steps, and integration with existing workflows

Recommendation: adopt the Growth plan for teams of 25–60 users, enable deepl-auth-key, and activate Glossary Generator to align terms across english content and translations. Use claude and claudegemini for terminology-heavy tasks, and track str_origin and str_translated values to QA results.

Licensing offers three tiers: Starter, Growth, and Enterprise. Starter covers 5–15 seats for small projects, Growth fits 16–100 seats for expanding teams, and Enterprise scales beyond 101 seats with dedicated controls. All tiers include a centralized authorization model, admin console insights, and API access to translations via the requestsposturl workflow.

Onboarding steps ensure a smooth installation and quick value. First, define glossary entries (entries) and map words (word) to their origins (str_origin) and translations (str_translated). Next, provision admin roles and assign an deepl-auth-key with restricted scopes. Then connect internal tools through the API endpoint and requestsposturl, and install the chrome extension to streamline browser-based translation tasks. Finally, import existing ファイル and align filename conventions, validating a test translation against a baseline.

Integration with current workflows hinges on coupling glossary and translation outputs with your content pipelines. Link translations to your CMS or repository, preserve the original source term (источник) in str_origin, and publish updated glossaries to control consistency. Use the glossary to enforce term choices across teams, and keep a marked list of バグ修正 fixes as part of release notes to track quality improvements.

Implementation details help scale adoption. Use deepl-auth-key to authorize API calls, enable chrome for in-context edits, and wire a requestsposturl to submit translation tasks from your content editor. Maintain files with a consistent filename mapping, and store entries provenance in str_origin alongside the translated result in str_translated. Consider 東京での説明会案内 to invite teams to a live session and collect feedback on the licensing and onboarding experience.

PlanSeats (range)Key featuresIdeal for
Starter5–15Glossary, translations, glossary entries, deepl-auth-key provisioningSmall teams starting with core workflows
Growth16–100Advanced glossary, requestsposturl integration, analytics, chrome extension, バグ修正Growing teams needing automation and visibility
Enterprise101+SSO, dedicated support, custom contracts, 機能追加, greent翻訳性能の改善Large orgs with complex workflows