According to reports, Google could be working on making Google it is speech recognition-based helper work for Android users if they’re not online.

Based on a report by Gizmodo, speech recognition greatly relies on cloud computing to work optimally. It is crucial as memory and the processing power needed is enormous. The report discloses that in a recent paper, Google engineers have shared how they used deep machine learning methods to run a not too significant speech recognition application which lives on the smartphone.

The paper describes the research project as ‘a big vocabulary speech recognition system that’s precise, has low latency, and has a modest enough memory and computational footprint to run faster than realtime on a Nexus 5 Android smartphone.’

The newspaper adds the system reaches 13.5% word error rate on an open ended dictation job, running with a median speed that’s seven times faster than real time.

Technology Times reports that Google used 2,000 hours of anonymised Google search traffic encompassing nearly that is voice 100 million requests, and added YouTube background to better that is noise model real life states.