Free tool
Clean Arabic text for matching and search.
Paste Arabic text and get a cleaned copy: tashkeel stripped, kashida removed, alef and taa marbuta normalized on demand, and digits converted between ٠١٢٣ and 0123. Live word and character counts included.
Everything runs locally: nothing you paste leaves this page.
Cleaned output
Counts
| Words | 0 |
| Characters (with spaces) | 0 |
| Characters (no spaces) | 0 |
| Output characters | 0 |
Why these cleanups matter
Arabic text carries optional marks that change bytes without changing meaning. The word "التقرير" and the same word with tashkeel or a stretched kashida are different strings to a computer, so a search misses them and a translation memory scores them as different segments. Cleaning text before it enters a term base or memory is how you make yesterday's approved translation findable today.
- Tashkeel (diacritics). Vowel marks like fatha and shadda are written inconsistently, present in some documents and absent in most. Stripping them collapses those variants into one canonical form for matching.
- Kashida (tatweel). A purely typographic stretch character (U+0640) inserted for justification. It carries no meaning and should almost never survive into a term base.
- Alef variants. Hamza seats on alef (أ إ آ) are a common source of spelling variation between authors. Normalizing them to a bare alef makes fuzzy matching and search dramatically more forgiving; it is also lossy, which is why it is a checkbox and not a default.
- Taa marbuta. Word-final ة is frequently typed as ه and vice versa. Converting one to the other builds match keys that tolerate the confusion.
- Digits. The same date can arrive as ٢٠٢٥ or 2025. Converting Arabic-Indic digits to Latin (or back) keeps figures consistent across a document set and makes number-matching QA checks reliable.
Worked example (hypothetical)
A reviewer at Al Dana Holdings searches the memory for a sentence from "Press Release.docx" and gets no match, because last year's copy was written with tashkeel and justified with kashidas. After stripping both, the two sentences are byte-identical and the approved translation surfaces instantly.
TranslateX applies this understanding natively: Arabic and English with real RTL/LTR handling, plus Arabic-specific QA checks for hamza seats, taa marbuta, and kashida in the review editor. See Arabic and RTL support, and build the cleaned terms into a file with the glossary CSV builder.
Frequently asked questions
What exactly does the toolkit remove or change?
Stripping tashkeel removes the combining marks U+064B through U+065F plus the superscript alef U+0670. Kashida removal deletes the tatweel character U+0640. Alef normalization maps أ, إ, and آ to bare ا; the taa marbuta option maps ة to ه. Digit conversion swaps ٠ through ٩ with 0 through 9 in either direction.
When should I normalize alef or taa marbuta?
When building search keys, term bases, or translation-memory match text, where you want spelling variants to collapse into one form. Do not apply them to text you will publish: the changes are lossy and alter correct spellings.
Why do diacritics break translation memory matching?
A memory matches on the text itself, and a segment written with tashkeel is a different string from the same segment without it. Cleaning input to a canonical form means previously approved translations are found instead of retranslated.
Is my text sent to a server?
No. Every operation is plain JavaScript running in your browser; there are no network calls and nothing you paste leaves the page.
Translate the document. Keep the design.
Right-click a file, work inside Office, or press F6 on anything on screen. TranslateX returns the same document in the other language: fonts, tables, and layout intact, terminology on brand.
Arabic, English & more · Layout preserved · On-premises available