Multilingual CSV import: map files in any language

Your schema is written in one language. Your customers’ files aren’t. A German export says Straße, PLZ, and Ansprechpartner; a Spanish one says Razón Social and Fecha de nacimiento; a Russian one says Фамилия and Индекс. All of them should land on your street, zip, contact, companyName, birthDate, lastName fields — without anyone renaming columns before uploading.

Most importers compare header text to field names, so a header in another language — which shares not a single letter with your field — matches nothing, and the user hand-maps every column. Bolting a translation API onto that creates new problems: data leaves the machine, results drift between runs, and offline or self-hosted deployments can’t use it at all.

Mildport takes a different path: the matcher itself understands what common fields are called across languages.

What works out of the box

• Six languages. English, German, Spanish, French, Italian, and Russian headers map onto a schema defined in any of them — most commonly, local-language files onto an English schema.
• The fields every business file carries. Names (first, last, salutation), the full address block (street, house number, postal code, city, state, country), birth dates, job titles, emails, phone numbers, websites, and geo coordinates are recognized by what they mean, not by how they’re spelled. PLZ finds zip; Fecha de nacimiento finds birthDate; Должность finds jobTitle.
• Business vocabulary, both CRM and property. Company-name conventions like Raison sociale, Razón Social, and Firmenname resolve to your organization-name field; CRM words like Ansprechpartner (contact person) and Kontrahent (account) place columns on the right record type. Property-management imports work in the same way — Mieter, Vermieter, and Objekt columns find tenant, landlord, and property fields.
• Accent- and spelling-tolerant. Prénom typed without its accent, Müller written Mueller, a file exported on macOS versus Windows — all match the same way.
• Deterministic and offline. No translation service, no network call, no run-to-run drift. The same file maps the same way every time, including in air-gapped self-hosted deployments.

Your data is never translated — values pass through byte-for-byte. Only the understanding of headers crosses languages.

When it isn’t sure, it asks — once

Language understanding doesn’t mean guessing. When a header genuinely fits more than one field, Mildport asks a clear question instead of picking silently, and the answer is remembered: the next file from the same source maps automatically. Repeated imports converge to zero questions.

FAQ

Can I import a CSV whose headers are in another language? Yes — headers in English, German, Spanish, French, Italian, or Russian land on the right fields of a schema defined in any of them. No renaming, no template file.

Does it translate my data? No. Cell values are never modified or sent anywhere; only column headers are understood across languages so they can be matched.

Does it need internet access or AI? No. Language-aware matching is part of the deterministic matcher — self-hosted, offline, same result every run. The optional AI assist is a separate, off-by-default layer.

What about accents and spelling variants? Matching is accent- and variant-insensitive: Prénom/Prenom, Müller/Mueller, and differently-encoded files (macOS vs Windows exports) all behave identically.

---

Try it with your own file — the live demo at mildport.com runs entirely in your browser. Related: automatic column mapping and importing rows with multiple addresses.