One of the areas of activity of the group of companies is the development of intelligent solutions for the automation of processes related to audit and consulting activities, compliance procedures at enterprises.
Since December 2019, a team of specialists has been allocated to develop fundamentally new technologies for parsing the meaning of documents "with one button", designed to reduce the user's labor costs from several hours / days to several seconds.
The solution uses its own machine learning algorithms for image recognition, image pre-processing, document classification using clustering, automated learning by receiving a sample set of labeled documents with representative data sets and machine learning through a continuous feedback loop during document processing operations.
To extract information from documents, an approach with pre-trained language models in the Transformer architecture is used (https://ai.googleblog.com/2017/08/transformer-novel-neural-network.html). This architecture was first presented by Google, but in this project it was supplemented with layers, and the language model used was adapted to the specifics of the recognized documents.
The development is supervised by the scientific director Irina Anatolyevna Arekhina, a candidate of economic sciences with many years of experience in management activities, audit, as well as research and practical work on setting up a budgeting system, management accounting, development of financial management, in the field of developing practical recommendations for managing regional and municipal finances, developing methodological and practical recommendations for credit organizations of all levels, practical experience in training and consulting employees of financial services.
