Microsoft has unveiled a range of new machine learning models for its Cognitive Services platform, reports TechCrunch. These include an API for building personalization functions, a form recognizer for automating data entry and an API for recognizing manuscripts.
First of all, the Personalizer is new. This model helps to build personalization functions. This is difficult, because it often involves building models based on data in different silos. With Personalizer, Microsoft focuses on reinforcement learning, which is a technique that does not require the use of labelled training data. Instead, the agent is constantly trying to find the best way to achieve a predefined goal, based on what users do.
In addition, there is an API to recognize manuscripts, called Ink Recognizer. This API can automatically recognize manuscripts, common shapes and documents. The service is already used in Microsoft Office 365 and Windows. The API allows developers to bring these features to their own applications.
Microsoft also adds Conversation Transcription to its Cognitive Services. This transcribes conversations and is part of the existing speech-to-text functions in Cognitive Services. The tool can label different speakers, write out the conversation in real time and deal with changing speakers.
Form Recognizer is a whole new API that makes it easier to extract text and data from business forms and documents. The service only needs five samples to understand how to collect data and users do not need to add labels by hand.
Form Recognizer also comes to cognitive services containers, allowing developers to use the models outside of Azure on their edge devices. This also applies to the existing voice-to-speech and text-to-speech services, as well as the existing anomaly detector.
Finally, the company announced that its Neural Text-to-Speech, Computer Vision Read and Text Analytics Named Entity Recognition APIs are generally available.This news article was automatically translated from Dutch to give Techzine.eu a head start. All news articles after September 1, 2019 are written in native English and NOT translated. All our background stories are written in native English as well. For more information read our launch article.