2019-09-13: The text limits become tighter
I didn't write here for a very long time. And there was a reason for that. For the last three months I was working non-stop on English phonetic translator.
I spent about two-thirds of that time manually proofreading 30,000 most popular English words. The rest of the time I was improving the algorithm and testing the converter. The updates mostly concern the American English, but there are improvements for British English as well.
So, what's new?
- There is a significant increase in the speed and the number of words in the dictionary, including the modified forms of words.
- I added the support for homographs and pronunciation variants for American English dictionary compiled from different sources.
- Now there is an option for adding the aspiration symbol [ʰ] after the consonants [p], [t] and [k] in cases when they are aspirated. Examples:
- talk [ˈtʰɔk],
- put [ˈpʰʊt],
- cat [ˈkʰæt].
- As you may know, in American English there's a phenomenon called cot–caught merger when speakers pronounce these two words in the same way. So instead of [ˈkʰɔt] for "caught" both words sound like [ˈkʰɑt]. Some dictionaries such as Merriam-Webster's Advanced Learner's Dictionary officially adapted this approach and use only the symbol [ɑ] for both sounds. Now you have an option: either differentiate between [ɔ] and [ɑ] for such words or display only [ɑ].
- I introduced the algorithm for word combinations. Compare:
- I have an apple [ˈaɪ ˈhæv ən ˈæpəl].
- I have to read [ˈaɪ ˈhæf tə ˈriːd].
- The unknown acronyms—both with (U.S.) and without (USA) dots—are now converted letter by letter.
- The common endings are localized for British and American English. Compare:
- wanted [ˈwɔntəd US, ˈwɒntıd GB],
- services [ˈsərvəsəz US, ˈsɜːvısız GB],
- weakness [ˈwiːknəs US, ˈwiːknıs GB],
- biggest [ˈbɪgəst US, ˈbıgıst GB].
- The algorithm for words with a hyphen is improved. Now even if one part of the word is in its modified form, the word is converted correctly. Example:
- two-thirds [ˈtuː-ˈθərdz].
In order to deliver you a high-quality product I tested the converter with ten official TOEFL tests – listening and speaking sections (almost 60,000 words). I used Adobe Audition to decrease the playback speed of audio tracks by 50%. This software allows to keep the natural quality of the voice. Then I compared what I was hearing with the phonetic transcription obtained with the phonetic converter. The decreased speed allowed me to discern little nuances in pronunciation. With the exception of some rare scientific terms and certain proper nouns, the transcription practically always matched the audio.
I also started to invest income in proofreading the database for rare English words. So eventually this tool will become simply the best.
If you like it, please share it with your friends!