How to make speech modification by voice input method of flying message

The ambient, accent, and speed of speech can have an impact on the accuracy of speech recognition when choosing voice input, once the identification of the error may change the meaning of the whole sentence, want to modify it, but still need to cooperate with the hands, in the car or other special scenes to achieve it is not so easy, this should be an intelligent service, The user experience does not seem to have improved much.

In order to solve this pain point, the voice of the launch of the exclusive voice modification function, the user can use pure natural language to correct the error, you only need to open the message to fly Input method, click the "Voice modification" function, you can talk, and then modify the text entered.

For example, add "eight o'clock in the evening" before "Go eat together";

For example, will Xiao Wen's "Xiao" word changed to the size of the "small";

For example, by deleting a word;

For example, "Tianjin" changed to "Beijing";

For example, before the word "good", add "you";

For example, after eating, add a "question mark",

This kind of intelligent interaction can be realized, the key is not in speech recognition, more in semantic understanding. Accurate semantic understanding, requires the application of in-depth learning based artificial intelligence technology, and artificial intelligence is based on large data, cloud computing and machine learning on the basis of the search engine is born in these three areas of the most core technology carrier.