Speechmatics’ Auto-Auto framework has just completed its first build of Italian. Our Auto-Auto framework allows us to build almost any language automatically and on minimal resources.
Initial results for Italian show an impressive accuracy of >80% on good quality audio. Accuracy levels depend on many factors including audio quality, compression rates, background noise etc. Unlike conventional systems, Speechmatics builds very robust systems, which can deal well with most common accents and dialects.
Speechmatics look forward to adding a new language every two weeks to our portfolio, a feat which is unrivalled in the ASR industry and only possible because of our continuing research and development of leading edge Artificial Intelligence and Deep Machine Learning techniques.