Usually people use their native language + english for conversation. I need google api to recognise both languages in a speech sample.
For example: "aaj ka weather kaisa hai" this sentence contains both Hindi(hi-IN) and English (en-IN) languages
How to set api parameters to recognise the code-switching mixed or multilingual speech?
You cannot mix language.
Speech Recognition roughly contains 3 part -> Accoustic model, Language model, and dictionary.
Accoustic model is the result of data training contains relationship between audio signal and phonetic
Dictionary contains words and how they pronounced, for e.g, word TOP are pronounced "T AH P" on the general speech recognition dictionary.
Language model is the connection between words to create sentences, for e.g. the word "I" is connected with "am", so the speech recognizer will very rarely (or never) give the result of "I are" or "I is".
Every Language have their own Accoustic Model (phonetic), Dictionary (words), and Language Model (sentences), so we can just mix them up.
The Question is : Is it still possible?
The Answer is : YES!
You can build your own language (in this case Hindi + English) using many tools, one I already tried called CMU Sphinx / Pocket Sphinx. You can build your own model, train it, and make a dictionary out of it. It will be alot work to do, but you can configure anything you will need for speech recognition.
Link for any platform implementation : https://github.com/cmusphinx
Google speech API does not work this way and it was not designed for mixed language. There are specialized APIs developed by few companies in India for Hindi+English cases, they recognize such mixed language just fine.
