TECHNOLOGY
Unique combination of technology
artificial intelligence and business intelligence
How we create technology
1
Important task
We see an important task, the solution of which is necessary for business, society or the scientific community.
2
Method set
We decompose the needs of the market to a basic set of problem statements that are solved by scientific methods.
3
Technology implementation
The staff of our laboratory create an intelligent system, relying on scientific and technological background
4
Testing and marketing
Our technology is tested on real data, confirming the possibility of solving the assigned tasks. Next, it integrates with our partners.
Image & Video Data Analysis
Image Analysis Direction
The laboratory team for image recognition solves various problems: classification, detection of objects and many others.
The main focus of research is recognized structures and information in photographs of printed documents. The urgency of solving this problem is confirmed by many customers of the applied area.

The team successfully implements the OCR and Text Detection methods, which deliver the best quality in the Cyrillic alphabet compared to competitors.
iDog library
Optical Character Recognition
Character recognition system in a limited image area
iDog library
Text Detection
Technology of detecting texts in the image
Open Source Dataset
TD & OCR Learning Dataset
Data sampling for training TD and OCR modules, open for use
IDog library
Data Generator
Data format generator for learning TD & OCR technology
Text & Transaction Data Analysis
Direction of text and log analysis
The laboratory team deals with three coherent tracks on thematic modeling, textual information analysis and user activity logs.


Logs user activity is easy to imagine a coherent text with a unique specific vocabulary, so our team has adapted the text analysis techniques to solve problems and personalization pattern recognition in this field.

Thematic modeling tool allows you to isolate the interpreted hidden information, or meaning, from large-sized text collections. The method allows you to create multiple technologies, from semantic markup dialogues contact center to the intelligent thematic search large collections of corporate documents.
Open Source
BigARTM
Open-source library for Topic Modelling
Open Source
Topic Net
Open library of learning hierarchical thematic models
iTexta Library
Preprocessing Pipeline
Single customizable text information pre-processing module
iTexta Library
Text Clustering Model Analysis
Library for analyzing the results of clustering large volumes of texts
Time Series & IoT Data Analysis
Time Series Analysis Direction
The laboratory team is developing a time series analysis of various nature.
The tasks of both forecasting and classification of multidimensional objects, and the detection of anomalies and differences in industrial systems
Subsequence Searching Engine
Search module for the closest time series subsequence
Made on
Tilda