Utility toolkit: alignment, terminology, QA, and delivery checks in one workspace

Translation work begins before the first segment is translated. A PM may align last year's bilingual files, extract terminology, import a client TBX, and estimate volume. During review, the team searches previous translations. Before delivery, it checks TM quality, numbers, dates, tags, and placeholders.

These tasks often sit across CAT tools, Xbench, spreadsheets, and small utilities. The tools may work, but their output is easily stranded: a term list never reaches the project, or a cleaned TM never returns to the workspace the team actually uses.

BabelBee's utility toolkit brings these supporting tasks into the same workspace. Open it from the account menu or the workbench header; you do not need to create a translation job first. Approved terms and aligned pairs can be written directly to the workspace glossary or TM for future projects.

This article explains what each tool does and where it fits in a translation workflow.

Before the job: prepare language assets and scope the work

Scoping is harder when translation memory, terminology, and volume are stored in unrelated files.

Alignment is for bilingual material that is not a TM yet. Upload source and target, or one bilingual file, pair the sentences, edit the pairs, then export TMX 1.4 or write them into workspace TM. That is the right step when a client sends last year's Word files and you need those translations in memory, not on a shared drive.

Term extraction reads source text or a bilingual file and proposes candidates, including Chinese, Japanese, and Korean. Accept the terms that belong in the workspace glossary, or export a CSV for a linguist to review first. Use it before a long contract, a manual, or any job where the same party, product, or clause will recur.

TBX import previews a TBX file, writes accepted terms into the workspace glossary, and can export the glossary back as TBX. Use it when a client or another tool already maintains a termbase — or when you need to return one.

Analysis reports new segments plus exact, internal, and cross-file repetitions. It counts Latin words and CJK characters. It does not invent fuzzy-match bands or weighted pricing. Use it to plan hours and quota, not to produce a CAT-style leverage invoice.

Together, these four tools answer: what memory do we have, what terms should we lock, what did the client already send, and how large is the job.

During the job: search with context

Once work is underway, the expensive question is usually not "what does this word mean?" but "how did we translate it last time, in this domain?"

Concordance searches a string across workspace TM, the open project, or uploaded bilingual files, and returns hits with context. A PM or linguist can answer from the record instead of from memory. That matters on repeat clients, on legal defined terms, and on any batch that reuses last quarter's wording.

Chinese production: script, width, and punctuation

Chinese delivery is often three finishes of the same source: Simplified for the mainland, Traditional for Taiwan, and sometimes a Hong Kong variant. Mixed full-width and half-width characters, or English punctuation in a Chinese paragraph, also read as unfinished even when the translation is correct.

CJK kit (中文生产三件套) applies 简繁 conversion, 全角半角, and 中英标点 as a rules pass — not a second translation. Preview the changes, then download the text or write them onto project targets. Use it on every Chinese job that must ship to more than one region, or whenever mixed punctuation would fail a client review.

Before delivery: clean memory and run mechanical checks

The last hour before export is when cheap mechanical errors become expensive. The toolkit keeps those checks next to the job instead of in a separate QA product.

TM cleanup flags exact duplicates, source-equals-target rows, empty sides, garbled text, and one-to-many conflicts. You preview first, then apply. Use it after a large TMX import, or when a workspace TM has accumulated years of mixed quality.

Numbers / dates / units is a locale pass for separators, date order, and common units. You get a report, plus a conservative fix for unambiguous cases. Use it on financials, specifications, and any file where 1,234.56 and 1.234,56 cannot be left to chance.

Regex find / replace runs across a project or uploaded files. Preview matches, apply, and undo the last apply in that session. Use it for late, systematic changes — a SKU prefix, a product rename — after you have seen the hit list.

Tags & placeholders checks parity for XLIFF-style markers, HTML/XML tags, {0}, %s, and similar tokens. Export a report if a reviewer needs it. Use it on software strings, help files, and any job where a missing tag will break a build.

Custom QA profiles are reusable rule packs: custom patterns, forbidden terms, must-use pairs, numbers, dates, units, tags, brackets, and length ratios. Save a named pack (for example, "Client A legal") and run it on a project or a bilingual file. These checks complement translation review by catching objective issues already defined in a style guide.

How the toolkit fits BabelBee production

The toolkit is not a second product and not a replacement for the workbench. If the task is "translate this document and review it," start with an upload. If the task is "prepare assets" or "check a file we already have," start in the toolkit.

Both sides share one login and one workspace. A term accepted in extraction is available when the next job starts. A TM pair written from alignment is available to concordance and to translation. That is the operational advantage: fewer tool switches, and fewer assets that never make it back into production.

A practical weekly habit is to clean TM and refresh the glossary before the week's uploads, then run jobs against assets the team already trusts.

Who this is for

The toolkit is for LSPs and in-house language teams that already use aligners, term extractors, concordance, and reusable QA packs, and want those tasks in the same workspace as AI production.

It is less useful if you only need a one-off draft and have no TM or glossary to maintain. In that case, upload the file and review in the workbench.

Frequently asked questions

Do I need to start a translation job first?

No. Alignment, term extraction, TBX, analysis, CJK conversion, TM cleanup, and the delivery checks can run on uploads or workspace assets. Concordance and several delivery tools can also attach to the project that is already open.

Does analysis replace a CAT leverage quote?

No. Analysis reports new work and exact repetitions only. BabelBee pricing remains word-based. Use the report to plan effort, not to invoice fuzzy bands.

Can reviewers use the toolkit?

Reviewers can search and run QA. Writing to workspace TM or glossary requires Owner, Admin, or Member.

How does this relate to Agent production?

The toolkit prepares and checks assets around a job. Agent production analyzes the document, creates reviewable terminology and translation strategy, then carries that plan through translation and closed-loop QA. They complement each other; neither replaces human sign-off on high-risk work.

---

Open the toolkit · Agent production vs Standard · Start a free workspace