From what I have read in this article and from other websites defining neural networks, neural networks are systems that are modelled to be in similar structure to the brain and spinal cord. When neural networks decipher input transmitted through voice (Google Voice using it for this function), it divides the word said into many different sounds until it is as basic as a vowel sound using the first layer of the structure that contains input neurons. Once this sound has been found the other sounds in all the words spoken into Google Voice, it is sent to the second layer where the output shall be sent to the third layer. These neural networks have one very useful feature which is that it learns from the data presented to it. Thus, neural networks have been used to interpret voice input to then perform a command or other function.