Retail Crowd

Complete British News World

It is an important advance in IT processing of spoken and written language

A language model used in informatics, which enables processing of spoken and written Hungarian texts, has been created at the University of Pécs (PTE) with the help of Microsoft technology, Microsoft Hungary told MTI on Thursday.

Among the reasons for the development, the communication said that everyone prefers to use their native language in chat and other automated applications, but since relatively few people speak Hungarian, it is often not feasible for companies to develop the necessary software to process it. The Beck University Applied Data Science and Artificial Intelligence team recognized this problem and began looking for natural language processing methods to make it easier to work with large amounts of Hungarian language data. The solution was to create a so-called “PERT” model in Hungarian.

BERT is an open source technology from the Google Watch Company designed to help you process your natural language.

The new model, which was created by the PTE team with less than 200 man-hours and an investment of 1,000 euros, helps the computer to understand text that can be interpreted in many ways by building context from context.

A stream of at least 3.5 billion words is required for the model to work. This database was compiled by the Linguistics Research Center, another participant in the project, from the national Hungarian dictionary, online media libraries and Hungarian language materials for the free translation database opensubtitles.org.

They noted that the team used company-developed Microsoft Azure artificial intelligence and a high-performance inference engine for machine learning models ONNX Runtime. AI and cloud-based education have been reported to have become a key area of ​​PTE since the organization partnered with the IT company in 2019 under the Microsoft Artificial Intelligence Knowledge Center program.

READ  Index - Tech-Science - The Beatles and Elvis Presley will be back again soon