A slug is the readable part of a URL, and making one means reducing arbitrary text to characters that survive being put in an address. Accented letters are decomposed and their marks dropped, so é becomes e; Cyrillic and Greek are transliterated character by character; and a table handles the cases where one letter becomes two, such as ß to ss and æ to ae. Symbols that carry meaning are spelled out — an ampersand becomes and, a currency sign becomes its code.
Decomposition runs before the lookup tables on purpose. An accented Greek letter only matches its mapping once the accent has been stripped, so doing it the other way round leaves those characters untranslated. After transliteration the text is split into words, optional stop words are dropped, the words are joined with your separator, and an optional length limit cuts at a word boundary rather than mid-word.
Choices that affect the address
-
Stop word removal shortens but can obscure:
dropping the and of and for reads well for a long title and badly for a short one, where the remaining words may no longer make sense together.
-
Hyphens are the convention, not underscores:
search engines have long treated a hyphen as a word separator and an underscore as a joiner, so an underscored slug reads as one long word.
-
Scripts with no mapping disappear:
Arabic, Hebrew, Chinese and Japanese have no transliteration table here, so text in them is stripped rather than converted. Provide a Latin title for those.
-
Changing a published slug breaks its links:
the old address stops working unless a redirect is added. Slugs are worth getting right before publishing rather than after.
What actually happens to the text
A slug is the readable part of a URL, and producing one means reducing arbitrary text down to characters that survive being placed in an address. The [msm_tool slug="slug-generator"] decomposes accented letters and drops the marks, so é becomes e; transliterates Cyrillic and Greek character by character; and applies a table for the cases where one letter genuinely becomes two, such as ß to ss and æ to ae. Symbols that carry meaning are spelled out rather than dropped — an ampersand becomes "and", a currency sign becomes its code.
Why decomposition happens before the lookup tables
Splitting an accented letter from its accent mark has to happen before the transliteration tables are consulted, because an accented Greek or Cyrillic character only matches its mapping once the accent has already been separated out. Doing the steps in the other order would leave those characters untranslated rather than correctly converted.
Hyphens are the convention, not underscores
Search engines have long treated a hyphen as marking a boundary between words and an underscore as joining them into one, so a slug built with underscores tends to read as a single long word rather than several separate ones. That is a practical, historical convention rather than a technical requirement, but it is the one worth following.
Removing stop words is a trade-off, not a free improvement
Dropping small words like "the", "and", "of" and "for" shortens a long title cleanly, and does the opposite for a short one — the words left over can stop reading as a coherent phrase once enough of the connective tissue is gone. Whether to remove them is worth deciding per title rather than leaving on by default for everything.
Scripts with no mapping simply disappear
Only Cyrillic and Greek have transliteration tables here, so Arabic, Hebrew, Chinese and Japanese text is stripped out entirely rather than converted into anything resembling the original. A Latin-script title is the only reliable input for those cases.
A published slug is expensive to change
Every existing link to a page stops working the moment its slug changes, unless a redirect is added to bridge the old address to the new one. That makes a slug worth getting right before publishing rather than treating it as something to tidy up afterwards.