10

There have been a couple questions lately with requests for some kind of web site where you can somehow search by pronunciation to find words:

Since there don't appear to be any such tools available on the web, I have begun building such a site, using the CMU Pronouncing Dictionary as a starting point.

Some things that I'm probably going to include:

  • Search by pronunciation of a whole word to find all words with that pronunciation (e.g. searching for /B EH1 R/ gives you bear, bare, Behr, Bair, Bahr, Baehr, and Baer)
  • Search by partial pronunciation to find all words containing that sound sequence (e.g. searching for D AH1 B AH0 L gives you double, doubles, doubled, redouble, Doubleday, Doubletree, doublespeak, doublethink, double-decker, double-quote, and double entendre)
  • Do the above searches with wildcards for one (?) or zero or many (*) sounds.
  • Hear pronunciations (using some kind of TTS system that can produce utterances with an arbitrary phoneme string)
  • See pronunciations using a variety of pronunciation notation schemes (the above is ARPAbet, but also support IPA and various dictionary schemes)

The proof-of-concept prototype I am working on has, so far, the first two bullet points. Since the target audience of the site is users of English.StackExchange, I'd like to ask users what other functionality they would expect or desire from such a site.

5
  • Out of curiosity, what is the ETA on this?
    – MrHen
    Jun 16, 2011 at 12:59
  • @MrHen I'd say a couple weeks before I have anything deployed to a publicly-accessible place. In the meantime you are welcome to follow my progress at GitHub github.com/nohat/pron
    – nohat Mod
    Jun 16, 2011 at 17:19
  • 1
    The project name is pron? I don't know if I am proud of that or not...
    – MrHen
    Jun 16, 2011 at 17:28
  • 1
    @MrHen I know, it's kind of cheeky, but it's also a standard abbreviation in the speech software community for "pronunciation". The final project will probably have a more descriptive/discoverable name.
    – nohat Mod
    Jun 16, 2011 at 17:38
  • What happened to this project?
    – Laurel Mod
    Dec 14, 2017 at 3:59

1 Answer 1

Reset to default
4

Well, here is a short list of fantastic features (as opposed to practically useful features):

  1. Searches along the lines of "contains the string th but does not have the sound θ"
  2. Ability to merge vowel sounds into "similar sound" buckets (e.g. ɑː + ɒ + ɔː) before searching
  3. As much support for varying dialects as possible
  4. Searches that can allow for multi-word rhymes including emphasis on specific syllables
  5. Pronunciation results for specific bigrams (i.e. searching for ea returns all valid pronunciations those two vowels)

Feature #1 is really the only one I would use often.

2
  • Thanks, I will keep these in mind as development proceeds. 1 and 2 are pretty straightforward and I should be able to implement those soon. 3 is basically impossible unless I can get access to freely-licensed pronunciation data. I am currently using the CMU Pronouncing Dictionary which is American English-only. I'm not sure what you mean by 4. Feature 5 is tricky, because I'd have to build a letter-to-sound mapping—that data is not included in the dictionary.
    – nohat Mod
    Jun 14, 2011 at 20:05
  • No worries on 4. I wasn't clear and it isn't that useful anyway. I understand the difficulty with 3; I guess I was just poking to stimulate thought about having access to different dictionaries? Don't know how plausible that is. 5 is tricky, yes, but it sure would be cool. ;)
    – MrHen
    Jun 14, 2011 at 20:35

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .