The last several years brought exciting innovations in the field of Artificial Intelligence, especially when it comes to advancements in speech and language processing. Processing speech and making text and audio information searchable enables a diverse set of innovative applications, including helping researchers in searching for related papers, or building information graphs for predicting the best new drug candidates, or uncovering issues with products and services in near real time. For region like Central and Eastern Europe, which includes 30+ countries, most speaking their own language, support for local languages is a critical condition for implementing innovation. That’s why the recent (September 2020) Azure Speech services update has opened a whole new area of opportunity for our region.
With updated language support,most of the EU languages are now supported in Azure Speech services. For region which I am covering in my current role, it means that we now have support for all of our CEE EU languages(Polish, Bulgarian, Czech, Greek, Croatian, Hungarian, Romanian, Slovak, Slovenian, Estonian, Lithuanian, Latvian, Maltese)andRussianin Azure Speech and Translator services. Additionally, our speech generation models have also been updated, now leveraging the Neural TTS - a powerful speech synthesis capability, which enables to convert text to lifelike speech which is close to human-parity. Below you will find3 benefits, how this might help you advance your products and services today.
Automatic generation of medical summary from spoken conversations between doctors and patients
First, analyzing speech data or generating speech enables you to extract insights from audio or video information, which otherwise would be unreachable for analytical systems. This might include data like customer support conversations or employee speech in videos or transcribing speech for field employees or doctors. Communicating with your customers with natural-sounding generated speech in your own language is another area of innovation, which enables scenarios from voice announcements to supporting people with visual impairments to building voice assistants. Is information the new currency? If you answer “yes” to this — why then would you have terabytes of currency sitting without you getting use of it? Now you can turn it into tangible cash-flow.
Azure Speechservices are a sub-set of pre-built (but customizable) APIs for working with Speech. This includes transcribing spoken language into text for further analysis (Speech-to-Text) and generating naturally sounding speech form text input (Text-to-Speech). Azure Translator is another piece in the puzzle, which has also received major update for the languages, now translating text between 70+ languages.
Second, there are new scenarios enabled now by these pre-built AI models. Do you have that innovative idea for analysing customer conversations or augmenting your service with spoken messages in your local language? Often, these ideas were not realized due to the associated challenges like finding the right skilled people within your organization and investing into a project with unknown development cycle and returns. Now it is possible to build a realistic prototype app quickly to extract insights from your speech data, by calling the service through the API — in days, if not hours.
Analysing customer support conversations brings insights from priceless data, which is untapped without applying Speech processing
Third, this is one of those cloud services, which may work without sending your data to the cloud! Many of Azure Cognitive Services today may be deployed right within your own data center as containers. This means, that none of the actual data will be sent to the cloud, as even processing will happen locally. In this case, only billing information will be exchanged with Azure.