photo credit: Jamiesrabbits One of the key elements in text to speech online is to have the web site or mobile app render the reading out of the text to a near faultless experience. We work with default dictionaries as well as dictionaries that are customer specific. In some cases, industry jargon (think the pharmaceutical industry here) can have very specific words that the default dictionary will not always correctly read out. That is where we provide a service which helps our customers enrich their own dictionaries with the corrected pronunciations. If the term is generic, the corrections then also get added to the default dictionary. Here is how the process typically works. The customer reports mispronounced words/phrases using our pronunciation sheet, where they:
- fill in how and what it is that is mispronounced;
- a description of how it should be pronounced;
- the context where the mispronunciation occurs;
- and an URL to where the mispronunciation occurs (quite often the mispronunciation are due to the context or the HTML code).
We then use the information we get to make corrections to the HTML code or to a specific word. If it is indeed a word itself that is mispronounced we make phonetic transcriptions using different phonetic notations that represent the International Phonetic Alphabet. The notations can for example be X-SAMPA or Kirschenbaum. The corrections are made in our default or customer specific dictionaries were we make search and replace patterns using regular expressions.