Translation. Region: Russian Federation –
Source: Novosibirsk State University –
An important disclaimer is at the bottom of this article.
As part of the scientific and industrial forum "Golden Valley", which is held by Novosibirsk State University, employees NSU Center for Artificial Intelligence presented a framework for managing artificial intelligence models.
"Our framework is for managing and testing datasets and AI models, which must be done before they are included in any workflow. To make this more clear, we also showed examples of the framework's operation with several datasets we already have," said Evgeny Pavlovsky, PhD, a leading researcher at the NSU Center for Artificial Intelligence.
One such example is a project implemented with the State Public Scientific and Technical Library.
The framework was loaded with digital library card data and its markup, and a model for card recognition and bibliographic information structure recognition, created by the center's own staff, was tested.
"We are currently in the process of handing over the framework to the client. Its implementation will significantly automate the work of the State Public Scientific and Technical Library staff in creating new cards. In the future, we plan to add an algorithm to the framework that will allow it to scan a new book and automatically extract the necessary bibliographic information from it, without the involvement of a library employee," explained Evgeny Pavlovsky.
There will be some “advantages” for readers, first of all, the process of finding the necessary books published before 2000 will be significantly simplified.
"Currently, this task requires a rather complex approach. For example, it took me about half an hour to find a book about Al-Khwarizmi, published in Tashkent in 1968. But I was very motivated to search. Modern users don't always spend 30 minutes searching for a single book. Therefore, with our framework, we've significantly simplified this process; you can find the book or article you need much faster, without resorting to specialist help," Pavlovsky emphasized.
There are other examples of the framework's use, all related in one way or another to smart city technologies, which are the specialty of the NSU Center for Innovative Research. This primarily applies to developers of artificial intelligence models, those who have a good understanding of their clients' needs and are currently in the model development stage. For example, when modeling heat distribution in a city, the client clearly understands the objectives, structure, and capabilities of the city's heating networks and the entire housing and utilities infrastructure, but also needs a model to optimize certain processes and parameters.
As is well known, some AI models can sometimes produce incorrect data, or, as they say, hallucinate. In such cases, a framework will be needed that allows developers to test models and understand which ones are usable and which are not.
But, as the developers emphasize, the range of its potential applications is much broader. This was confirmed by the interest the development generated among visitors to the exhibition organized as part of the Golden Valley forum.
Please note: This information is raw content obtained directly from the source. It represents an accurate account of the source's assertions and does not necessarily reflect the position of MIL-OSI or its clients.