drime

Welcome to drime!

Wait, what is this?

This is a3nm's attempt to build a better French rhyme dictionary. It uses the Lexique database with some customisations, and is powered by Python, MySQL and Flask. You can get the code on the drime repository to run it locally.

Ad: if you like drime, you might enjoy the French poetry checker plint.

How do I use it?

In the word field, enter the word you want to get rhymes for.

In the n_syllables field, you can specify a number of syllables to limit on. You can either specify an exact number or a range (eg. "1-3"). You can suffix a "+" to indicate that you can accept one syllable more if the word starts with a vowel or non-aspirated 'h', ie. the previous word ends with an elidable 'e'. The syllable counts for words are approximate: the estimation is always done to overapproximate the results of a given query.

The gender checkbox limits to rhymes that respect rhyme gender. If you're unsure about what this is, see the relevant Wikipedia page (in French).

The classical checkbox restricts the rhymes to those that are allowed by French classical poetry constraints (according to which, e.g., "rang" rhymes with "sang" or "flanc" but not with "grand"). If you select this option, you should probably also enable the "gender" option. (Sorry, I can't find a reference online to explain more about how classical rhymes work.)

The pronunciation indicated by drime is not correct!

drime uses Lexique to find out about the pronunciation of words, and uses a different system to guess that of unknown words. If it fails, you can provide a pronunciation written between square brackets using X-SAMPA. You can also use this to disambiguate between multiple possible pronunciations. For instance: fils [fis], leicester [lEstER].

Which results are missing? (false negatives)

• The system will only provide words from its dictionary. No proper nouns, for instance.
• The system will only provide words which either rhyme (ie. all sounds up to and including the last vowel match exactly) or which share the last three characters (for visual rhymes).
• The system might be fooled because its pronunciation for a word is mistaken or does not include all variants. Besides, the system is picky about phonemic distinctions (notably "brun/brin", ie. "9~/E~").
• The system will eliminate results that it considers redundant, ie. results derived from a previous result.
• The system will not accept rhymes for the classical rule of liason supposée.
• The system will only return the 500 best results.
• The system may be too strict when enforcing classical rules (it shouldn't happen very often, though).

Which results are unnecessary? (false positives)

• The system will provide words which are too similar to the query word to make a decent rhyme with it (eg. "reprendre" for "prendre").
• The system will provide results with a wrong syllable count.
• The system will provide abusive visual rhymes (at the end of the results).
• The system will provide rhymes which do not follow classical constraints, because its interpretation of them may be too liberal.

How are results sorted?

Results favor, in this order, the number of common phonemes, the number of common letters, and the frequency of the word.

How is pronunciation written?

It is X-SAMPA. Information about French pronunciation can be found on the French Wiktionary. Be aware that because of Lexique limitations, the pronunciation indicated is a colloquial one and not the one that you would use to versify (for instance, "placement" is reported as "plasmA~" but would be read as "plas@mA~"). Heuristics are used to work around this when filtering on the number of syllables.

Is there an API?

Sort of. Provide a "json" GET parameter and you'll get the page in JSON.

What about the name?

It's pronounced [dRim], as a pun with "dream" (the dream rhyme dictionary) but also because it is short for "dictionnaire de rimes" in French.