|
|
|
|
We are currently working on a speech recognition engine for Turkish. We already have developed the algorithms, tools, knowledge base, and data that can be used by a speech recognition engine to successfully recognize spoken Turkish. It took us six years to develop and perfect these tools. The problem was that the grammar of Turkish language is VERY different from English and a classical approach (e.g., the one used in Speech API by Microsoft) does not work well on Turkish. For example, in Turkish, by adding suffixes to a verb you can generate approximately 55,000,000 words. However, despite to this problem, we realized that the development of a successful dictation program in Turkish is easier compared to English because there is a one-to-one mapping between written text and its pronunciation (in layman's terms, you read what you type and you can type what you hear without using any pronunciation dictionary or grammar). However, this was only on the surface and subtle problems hindered the development of successful speech recognition engines for Turkish Speech Recognition engines. Since we think that the speech recognition engines developed for English is better than the speech recognition engine we started to develop but abandoned or any other "academic" engines, and it makes sense for us to work with a company that already has a successful speech recognition engine. Thus currently we are looking for a partner to develop a Turkish Speech Recognition Engine and other dictation tools. It also makes sense for the partner to work with us because we already have extensive knowledge about Turkish Language and completed most of the work you would have to do if you wanted to develop a Turkish Speech Recognition Engine. Also, there is definitely a big market for Turkish speech recognition engines that can justify the short/long term investment from our potential partner's part (e.g., Turkish is currently spoken by more than 200 million people in more than six countries) Thus, we would like to form some kind of alliance with a company that has demonstrated the capability to develop and market a speech recognition engine for another language. We also have experience in text-to-speech for Turkish since we successfully localized the text-to-speech engine used in Amiga. You can download the hidden markov model of the phonemes in Turkish. To setup the demo, run the setup program and run the program called "sozcuk uzayi.exe" DOWNLOAD THE HIDDEN MARKOV MODEL NOW |
__________________________
|
