15 Sep '21
TechnoIntellekt, a member of the ATOLLis group of companies, which innovates artificial intelligence services (robots) for data management (DatiUm software) in Skolkovo, has released the first in a series of software robots DatiUm.Smart File for digitalization of unstructured data sources.
The purpose of a robot of the RPA (Robotic process automation) class is automatic processing, identification and classification of file archives of different industries by elements of the ontological model under control of a trained neural network.
The result of the robot's work is a digitized source of unstructured data in the form of a register of digital passports of documents with filled attributes from the contents of files and distributed files over the structures of digital passports without changing the physical file storage scheme.
According to research by the Microfocus corporation, within 3 - 5 years, companies reduce the proportion of hits in relation to the volume of data from 65% to 5% due to the impossibility of quickly extracting business-critical information. The source of unstructured data digitized by the robot becomes available for use by data buses, enriching the generated queries according to the ontological model with information stored in the digitized file archive.
In routine use, the robot minimizes the time of specialists in classifying newly appearing files and searching for business-relevant data in file archives.
Together with DatiUm.Smart File, a trained neural network for fuel and energy enterprises was released, which allows to automatically process more than 90% of standard documents in file archives. The remaining 10% are performed in an automated mode, retraining the model for classification options specific to each enterprise. For other industries, the initial training of a neural network can take several months.
The training of the neural network was carried out by business experts from Ot-Oil, a member of the ATOLLis group, using more than 1TB of subsoil use files from publicly available sources of Rosgefond. During 23 iterations, 2348 objects were processed at 188 nodes of the ontological model. The use of a high-performance bench provided by IBM (within the framework of a technology partnership) allowed us to reduce the time of each iteration of the full training of the model from 2.5 months to 10 days.
The first applications of the robot were successfully carried out by Ot-Oil in the implementation of project management projects in the oil industry.
Currently TechnoIntellect is developing several robots from the data management series at once:
• robot cognitive search for structured and unstructured data sources based on an ontological model;
• a data unification robot that automatically generates conversion tables of reference values that have different codes in different systems;
• a robot assistant that supports the execution of routine tasks of employees when working in process IT solutions.
The DatiUm module with any set of robots to be created is based on openSource technologies and can be integrated into any data management product. In particular, in OT-OIL it is used in the PLATFORM ATOLL and ELDOCA application lines for solving the problems of maintaining a register of digital passports of documents and objects, searching for data and viewing them.