Multilingual support
English, Turkish and German are handled by the same engine, including stemming and date formats.
- Turkish stemming: A Snowball Turkish stemmer in keyword matching.
- German compound splitting: "Softwareentwicklung" becomes "Software" and "Entwicklung".
- Per-language stopwords: Tuned for each language.
- Encoding corruption check: Garbled ı İ ş ğ ç ö ü ä ß characters are reported as a Parseability finding.
- Date formats: "Oca 2022", "Ocak 2022", "01.2022", "Jan. 2022", "März 2022", "heute", "halen", "devam ediyor".
- Market-based advice: Photo, date of birth, marital status and military service, depending on the target market.
- Europass detection: Europass layouts are recognised and flagged.
How does the CV checker handle Turkish and German CVs?
The CV checker runs English, Turkish and German CVs through the same scoring engine, with language-specific handling where the languages differ: Turkish casing and suffixes, German compound words, date formats, stopwords and section headings. The score stays out of 100 whichever language the document is in.
The document language is detected by comparing how often each language’s stopwords occur in the first 4,000 words, and that choice selects the matching dictionaries. Turkish keyword matching lowercases İ and I correctly, strips suffixes down to a stable key and folds diacritics, so “geliştirme” and “gelistirme” count as one word. German matching splits compounds against a dictionary and also tries spaced and hyphenated forms, so “Softwareentwicklung” can meet “Software Entwicklung”. Section headings such as Berufserfahrung and İş Deneyimi are recognised.
Which date formats and character problems does it recognise?
The CV checker reads dates such as “Oca 2022”, “Ocak 2022”, “01.2022”, “Jan. 2022” and “März 2022”, and open-ended ranges ending in “heute”, “halen” or “devam ediyor”. It also reports broken character encoding as a parsing finding.
Encoding problems are checked in three ways: replacement characters or CID placeholders in the text layer, mojibake sequences left when UTF-8 text is read as another encoding, and a high share of letters outside the Latin ranges a European CV normally uses. Any of these can destroy ı, İ, ş, ğ, ç, ö, ü, ä and ß, after which names, headings and skills stop matching. Each finding quotes example lines from your document and gives a concrete fix, such as embedding the fonts when you export.
| Area | English | Turkish | German |
|---|---|---|---|
| Keyword matching | Standard word matching | Suffix stripping and diacritic folding | Compound splitting and spaced or hyphenated variants |
| Open-ended date words | present, current, now | halen, devam ediyor | heute, aktuell, bis heute |
| Section headings | Experience, Employment | İş Deneyimi, Deneyim | Berufserfahrung, Werdegang |
| Special letters checked | Not applicable | ı İ ş ğ ç ö ü | ä ö ü ß |
What market advice does it give for a CV in each language?
The CV checker judges the personal-details block against a target market you choose, English (US/UK), German (DACH) or Turkish, or against the document’s own language when you leave it on auto-detect. It looks for photo references, date of birth, marital status and military service.
For an English-market CV, any of those fields costs 2 points, because many employers discard them. For a German CV, a military service line costs 1 point, and a missing date of birth only produces a note that costs nothing. For a Turkish CV, a missing date of birth likewise produces a no-cost note. A Europass layout is detected by the brand word or by its own section labels and costs 2 points, because its two-column tables are read badly by parsers. Absence of a field never lowers the score.
What are the limits of the language support?
The CV checker supports three languages only: English, Turkish and German. A CV in another language is still scored, but without dictionaries for it, so its keyword and section checks are less reliable.
The Turkish stemmer is a practical subset of the Snowball approach, not a full linguistic one, and it deliberately leaves out endings it cannot tell apart from noun endings. German compound splitting is a matching aid that adds forms to compare, not a parser, and a rare compound may not split. Language detection is by stopword frequency, so a very short document can be misread. Market notes are advice about what readers expect, not rules, and no parser requires a photo or a date of birth.
Last updated