Hello again, Stefan:
   Since your enquiry, found some Internet sites that may be useful to you in your search. I applaud you in your interest. There are extremely few people in the world who would like to make communications easier. And will actually spend the necessary time to do that. Such as your phonetics-based idea of creating a computer software program to transliterate between selected languages. I had the desire to make that process simpler, too. However, I discovered that no one in the world had ever analysed, and listed out, the different ways people spelled the same, or similar, speech sounds of various languages that used the 'same' alphabet. Say, the so-called 'Roman alphabet'.
   You remember the IPA (International Phonetic Association) started more than a hundred years ago. A few people in Europe wanted to provide one symbol for each speech sound (vowel, or else consonant) for a few of the languages spoken in Europe. The IPA people ignored the way the foreigners across the Atlantic Ocean spoke in the United States. To this day, each publisher of a so-called "pronouncing dictionary" who relys on, or modifies, the phonetic alphabet of the IPA ignores specific, common, actually used, American phonemes. That is, each phoneme contrasts against another phoneme to convey meaning in speech.
  Examples of difficulties between a speaker of a European language speaking American 'English' are these: Some Europeans assume they can say the word written as 'ship' in English the same as 'sheep,' and can say the word written as 'lamp' the same as the spoken word for a baby sheep-- lamb. A scribing system must accurately provide a one-for-one correspondence between the letters of its alphabet, say, and the phonemes of the spoken language the scribing system represents.
   The contributors to the IPA never intended that that phonetic alphabet be adopted for use in scribing any particular spoken language. Thank God. Yet, dictionary publishers, and persons interested in computerized analysis and transliteration of main languages, fall into the trap of relying on that phonetic alphabet.
   Or, there are other extremes. One transcribing system I discovered today (urged by your request) comes out of the Carnegie Mellon University. The expert(s) at 'Speech at CMU' came up with a "Phoneme Set." We assume it's for the American form of spoken 'English.' The list contains 39 phonemes, four of which contradict the definition. The four are sets of joined vowels. The first vowel set is AW for the two vowels in the word [cow] they spell as KAW. The next set is AY for the i...e in the word [hide] they spell as HHAYD (sic). The third set is EY for the two vowels a...e in the word [ate] they spell as EYT. The fourth set is OY for the two vowels in the word [toy] they spell as TOY. That's nice. Except the vowel spelled O, above, they also spell as AO in AOT for the single vowel spelled as ough- in the word [ought]. Quite some confusion.
   Thus, they use the letter 'Y' for two separate vowel phonemes as well as to represent the typical American consonant. The list contains 16 single letters from the 'Roman-French-English' alphabet used in America for 16 phonemes. The list also contains 20 sets of two letters for 20 single phonemes. As mentioned in the previous letter, there are 42 provable, actually used, necessary phonemes in the speech of Americans. So, this CMU compilation is quite limited in accounting for American speech sounds. I avoided downloading their free 125,000 word dictionary. We must remember the caveat, GIGO-- garbage in = garbage out.
  You can reach this University source at http://www.speech.cs.cmu.edu/speech/ and http://www.speech.cs.cmu.edu/cgi-bin/cmudict/ in America. For Europe an excellent source for more research, is at http://tn-speech.essex.ac.uk/tn-speech/ This e-mail address is the Socrates Thematic Network in Phonetics and Speech Communication.
  As stated, it is extremely unlikely that anyone has compiled the knowledge you seek, Stefan, and has it in a form that is commercially available. I spent the equilivant of fourteen years full time just on the basic research for American English. I also spent over two full years designing the layout of a 114-key computer keyboard necessary to accommodate the keys needed for an extra 22 letters missing from the Roman-French-English alphabet. This is after I invented symbols easily recognizable as belonging with the current, familiar alphabet. Prior to that I scrutinized more than 10,000 different letters of existing alphabets and scribing systems. I rejected the existing symbols as inapplicable. And next I compiled the quantity of the top 2000 or so most-often occurring words in running text, and transliterated them into this first and only scribing system designed exclusively for American English. then I counted the oftenness of occurrence of the letters that represented the individual speech sounds of the language. And finally I arranged the complete, consistent, 43 letters (yes, including one for a common American and British speech sound that is not a phoneme) into the most ergonomic and efficient layout practicable.
  The only thing now necessary is to come up with $200,000 American to complete the basic patents and manufacture the prototype keyboard to be able to enter my tens of thousands of pages of words in manuscript into a computer. Then, we easily could come up with the answer to your request, Stefan. As you might guess, this will also be the solution to the greatest problem in the world. To have a simple, phonetic, consistent scribing system for one of the world's main languages. For the first time in recorded history.
  Is the above any help to you in your quest?