Now you may be thinking, why would you want a delay between when someone talks and when the machine outputs the translation? The answer is simply due to the complexity of the language involved. When dealing with languages that are similar (between English and Spanish, or Mandarin and Korean for example), instantaneous translation can kick in immediately with relatively smaller margin of error. But when going from languages with significantly different grammar and syntax (say between Malay and Japanese), the translation system needs to take in more of the sentence so that it’s not missing out on important context.
So instead of waiting for a person to finish speaking, the translation system can start translating before the end of the sentence. And you can adjust, if need be, the delay before the translation system kicks in. It’s something that Google already does with Translate, but Baidu’s system grants a little bit more flexibility in terms of the delay before translation begins. And since Google isn’t in China (yet), this can be a useful alternative if you need to head into the mainland. This is assuming, of course, the company has plans to release it for commercial use. (Source: Baidu)
