3

I am trying to determine how a name 'might' sounded based on text values how a name is pronounced audibly.

I am aware of SoundEx and this calculator for it, the improved MetaPhone, Double Metaphone, MetaPhone3, the alternative Caverphone, and the proprietary NameX algorithms to describe similar names as many of which were mentioned in this previous post. For example my last name in SoundEx is S260 but there are a variety of last names that have the same score that are only vaguely similar.

These numerical values largely seem like a one way translation as I have not been able to find a way to turn these values into an audible sound to see if they are related.

I am aware of the text symbols used in English Pronunciation Guides, I have also become a huge fan or Forvo.com when used in conjunction with GoogleTranslate, but it is not oriented towards names but can still be very helpful. I have turned towards pronunciation guides, for example old Pennsylvania dutch as well.

I am wanting to do this to try to determine names I come across might actually be pronounced similarly or not with a qualitative audible check as well. Some names using the above mentioned tools have taken hours to work through so I am looking for additional or better resources.

The root of my question: I am looking for a resource that can take text and coded values (Soundex, MetaPhone), and pronunciation symbols and pronounce them audibly.

  • I'm also wondering whether, to make this less broad, you could focus your question on finding a "resource that can take these text, coded values (Soundex, MetaPhone), and pronunciation symbols and pronounce audibly in different dialects / languages?" for just the dialect/language that is of most interest to you first. The answer to that may point to something more generic anyway. Also, I'm guessing that even within the same language/dialect the pronunciation may drift over time so it may be important to specify the period of interest too. – PolyGeo Jan 1 '15 at 3:55
  • @PolyGeo I think that is what I was really trying to answer but was trying to not make it so specific to a period as starting point. I revised again. – CRSouser Jan 1 '15 at 3:59
  • 3
    I am not sure your desire is possible. The Soundex and MetaPhone algorithms intend to produce a value such that similar sounding words end up with the same value. Thus Pack, Peck, Pick, Pock and Puck would, I think, all have the same value. Yet they do not sound the same. Put another way, they are "a one way translation" which is lossy. Without checking up on all the algorithms, I suspect they all have the same intention - just to concoct a value to detect similar sounds, not encode actual sounds. Dictionary phonetic symbols are complex because they aim to encode actual sounds. – AdrianB38 Jan 1 '15 at 20:33
3

One site I use is Russian website translate.Yandex.ru which is a website, and that has a language "Translator" function ("Переводчик") on the main page's function bar. That translator is by comparison in my opinion overwhelmingly superior to Google's rinky-dink language translator.

That being said, the Пекеводчик will provide audible pronunciations of some languages, but not all of them. For example, the audible pronunciation of the English surnames "Smith" and "McCorriston" are pronounced very correctly by a female computer voice, and the German name "Schmidt" is also pronounced quite correctly by a male computer voice. Portuguese and French are other languages for which audible pronunciations of surnames and text is available, and of course Russian surnames and text as well.

I too make relatively extensive use of Forvo.com for audible word and phrase pronunciations by users (currently for Bulgarian), and that website is really an important tool in my language learning toolbox.

And oh, I'll make one further point about Yandex; you can copy large swatches of text of interest from foreign language news (or other) websites, then paste that text into the source text window and read out the almost-instantaneous translation in the target text window. If the target text is one of the languages that Yandex provides audible translations, only the first sentence or two of that translated text will be spoken. But that's better than nothing, eh?

(You can also cut and past to Google, but although Google is able to pronounce some languages well (e.g. Hawaiian), it can't handle large batches of cut and past input or output text input for either printed or audible translation.)

| improve this answer | |
  • 1
    While not what I was looking for exactly I think the additional resource and explanation was useful so I left it as an answer after some edits. – CRSouser Nov 29 '16 at 16:32
  • Thanks very much for your edits; I really appreciated them. So one turn deserves another, as they say: Given your background, re: your original question, have you tried asking that question in the "Artificial Intelligence" Q&A forum, perhaps under the tag "NLP" (natural language processing)? – К. Келлогг Смиф Dec 1 '16 at 8:19

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.