I'm finding stuff like this in every language, but it's all written in sentence form scattered all over the place. Is there a central database of this sort of stuff for each language, or a book of some kind?

In most Russian dialects, о is pronounced as in hello only when it is stressed. If it comes immediately before the stressed syllable, it is pronounced 'ah', as in father. Otherwise, it is pronounced 'uh', as in money.

Basically, the variations in pronunciations of the written letters based on the context.

For example, is this EVERYTHING for Gaelic, or is there a lot more (type of thing). I would like to see:

the a in {x}gh{v:vowel} pronounced [ɤ]
the a in {x}gh$ pronounced [ɤ]
the a in {x}gh$ pronounced [ɤ]
The a in {x}ll{c:consonant} is [au]
The a in {x}nn{c:consonant} is [au]
The a in {x}m{c:consonant} is [au]
The a in {x}ll$ is [au]
The a in {x}nn$ is [au]
The a in {x}m$ is [au]
The a in {x}rr$ is [aː]
The a in {x}rr{c:consonant} is [aː]
The a elsewhere is [a]
  • I don't quite understand the {x}gh$ notation, etc. What exactly does that mean? – matt Sep 16 '20 at 15:26

eSpeak, or its more current fork eSpeakNG, are formant-based speech synthesizers that cover a decently large number of languages, although not all with the same quality. Their grapheme-to-phoneme conversion is rule-based (or dictionary-based when rules fail, as is often the case for English), which means the distribution contains a number of rule files of the type you seem to want, although, of course, they are expressed using the language of eSpeak's own rule engine.

Here is the rules files for English for example, or for something less complicated and closer to your example, the one for Irish.

Although the rules files aren't natively written using IPA, they do contain a lookup table between their internal format and IPA, as eSpeak can be made to output IPA with a command line like

espeak --IPA "Text goes here"

If eSpeak is not satisfactory for you, I believe you can be on the right track if you search for "graphemes-to-phoneme" systems, although such systems may not necessarily include language files, if they are merely an engine.

  • Excellent information! This is great, thank you!!!! – Lance Pollard Sep 28 '19 at 17:03

If you mean literally all "rules of pronunciation", you are extremely out of luck. That's a call for "knowledge of all phonological and phonetic systems of all languages", which we don't have, not even close. If you mean "rules for orthographic interpretation such that I can pronounce any word in that standardized written language" (e.g. what does "ł" stand for), you're also out of luck. First, not every language has an ideal phonemic orthography (e.g. English, Russian – stress isn't marked, stress affects pronunciation; Arabic – I have no idea what vowel to use in that word in that dialect). Second, there isn't necessarily a pronunciation for a word, for example tomato. Or, in North Saami, spelling isn't chaotic but is abstracts away from dialect differences, so you need to learn specific rules to map spelling to phonetic output in a given dialect. The wiki article does not tell you that "Dan áiggi Jesus lei oktii vázzimin gordnebealdduid čađa sábbáhin" is pronounced (one dialect, broad transcription) [dan aajkki jesus lej okti vaaddzimin goreʔnebealttuih tʃɑðaa saabbaahin] (I think: I haven't verified that).

There is no central database for languages w.r.t. any linguistic fact.

  • Where would I begin to find even any information like the other answer shared, and what I posted from the Wikipedia table. Do I go to grammar books, or to "pronunciation guides" (searching for that leads to spam mostly), or is there something else? From your answer I gather there is nothing robust, but where can I find some pieces efficiently? – Lance Pollard Sep 28 '19 at 17:05
  • Google "X orthography wikipedia" where X is the name of a language. If you get a hit, look for a table; otherwise (1) give up or (2) look for a grammar, getting hints from the Glottolog entry. – user6726 Sep 28 '19 at 18:21

Wikipedia articles titled "[language name] orthography" or "[language name] alphabet" typically have a list of grapheme-to-phoneme correspondences.

A site called Omniglot also has summaries of writing systems.

  • I have been to all of these pages but I don't know if they are thorough / complete. For all I know they are just sampling the pronunciation system 10%. – Lance Pollard Sep 28 '19 at 17:06

Basically, the variations in pronunciations of the written letters based on the context.

There are two separate steps here.

The first is going from the orthography (written language) to the underlying form (list of phonemes). In some languages, this is pretty easy: the Russian letter о (according to most analyses) represents a single phoneme /o/. In some languages, it's horrendously complex: for English and Mandarin, there's no way to avoid a long list of hardcoded special cases.

The second is going from the underlying form (list of phonemes) to the surface form (list of phones). This is one of the main things phonologists study, and any good grammar should include the most important points of it.

For this second step, there are various different theories for how this process works, which can be somewhat cleanly divided into rule-based theories (apply a series of transformation rules one by one to the underlying form, eventually you end up with the surface form) and constraint-based theories (various candidate forms are generated, then compared against a ranked set of constraints, which eliminates the worst offenders until only one remains). In real-world grammars you're going to find mostly rule-based explanations, since constraint-based theories are significantly newer and less widely accepted.

  • Do you have any constraint research you could share that would be interesting? – Lance Pollard Sep 28 '19 at 17:09
  • @LancePollard That would make a reasonable question! ("What's a good introduction to constraint-based phonology?") – Draconis Sep 28 '19 at 17:11
  • Non-theoretical descriptive grammars do tend to give rules rather than positing constraints and ranking. Wile OT (the dominant constraint-based theory) is newer, is is also the received standard theory. I say this as a rule-based curmudgeon who wishes it weren't so. But the fact is that constraint-based theories are the most widely accepted. – user6726 Sep 28 '19 at 18:17
  • @user6726 Most widely accepted among theorists, maybe, but I'm a descriptivist, and it seems rules are more widely used among L1 speakers of linguistics-paper-ese—therefore rules are grammatical and constraints are not! (In seriousness, I agree, but it'll be much easier for a newcomer to find specific lists of rules than specific lists of constraints for whatever language.) – Draconis Sep 28 '19 at 18:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.