Google Voice Search is already pretty awesome. Most of the time it understands your question without any issues and serves up a relevant answer quickly. Now the service is set to take a huge leap forward.
Google announced this week that it has upgraded voice search to new neural network acoustic models. The company is switching to a system it calls Connectionist Temporal Classification (CTC). This marks the first big shift since 2012, when the search giant adopted Deep Neural Networks (DNNs) as the core technology powering voice search.
The new system should be more accurate overall, and promises to drastically improve voice recognition in noisy rooms. Google’s upgraded voice search is also a lot faster. It’s capable of recognizing the difference between similar sounding letters and groups of letter using fewer computations.
If you’re interested in diving into the specifics, hit the source link below for a pretty detailed explanation of the technology behind Google’s new voice search. If not, you can start testing out the upgraded service on Android and iOS now to experience it for yourself.