Content |
How software robots are smarter
The complication of bots and their intellectualization is an obvious path of development. Vendors and suppliers of RPA technologies have already embarked on this path. "Classic RPA automation is developing towards intelligent automation," said Anton Sergeyev, RPA Development Manager of TietoEVRY Russia. "Initially, when the RPA industry was just emerging, it was enough to automate actions with applications, following a simple predefined logic. And today, most simple scenarios of trivial interaction with applications have already been implemented. There are complex interactions not only with applications, but also with people and relevant artifacts, such as documents, signatures and seals. "
What smart robots know
Victoria Babankina, head of NFP's RPA department, agrees with her colleague: "Companies that have been in step with RPA have already robotic" classic "processes where the robot is involved as a" clicker. " And now they are looking at other processes that could be robotized with advanced functionality, including AI. Demand gives birth to supply - now there are requests and problems, the solution of which pushes us, as an integrator, to use more complex technologies. " She cites as an example a robotic process of verifying the authenticity of a signature in documents based on a reference one: the robot transmits the sample to a previously trained neural network, and it produces the result in the form of a percentage of the authenticity of the signature.
 | This direction continues to develop thanks to new achievements in the field of artificial intelligence and machine learning. Synergy between disciplines leads to the emergence of smart and knowledgeable digital employees. |  |
For example, those talked about by Pavel Borchenko, Director General of ROBI: {{quote 'Robots with AI components are actively used to check documents: the robot will collect the necessary information, using a component for analyzing and classifying text will form comments and provide them to the user in a convenient format. Previously, such a task was solved by difficult and long-term integration of systems, and today, increasingly, companies solve this problem using RPA - it is faster and cheaper. }}
| | Speech synthesis and recognition have stepped far forward, which allows the robot to speak realistically with a person, "he explains. - Voice commands are one of the important elements from which the algorithm for communicating with a software robot is formed, the use of a voice component is an important step towards the formation of a full-fledged digital employee. | |
| | However, this concept is fully disclosed only now, with the advent of elements of artificial intelligence in RPA solutions. | |
It is AI in 2022 that will become one of the main drivers of the growth of the robotic automation market, he believes, since there are many scenarios of RPA and AI synergy: you can use conversational AI and RPA to automatically respond to customer requests, for intelligent document management (IDP), etc.
Outline of the new IPA market
According to Dmitry Smykalov, two types of robots will be presented on the market: ordinary software robots, to which we are already accustomed, and robots that imitate elements of human behavior. The second got a new name - Intelligent Process Automation (IPA).
IPA is the result of the evolution of RPA, when intellectual capabilities of machine learning and artificial intelligence are added to ordinary non-intelligent bots. This makes the robot smart and eliminates several drawbacks of RPA technology. So, robots use elements of machine learning and artificial intelligence, which reduces the requirements for structuring and uniformity of data, and also makes it possible to start working with unstructured data.
| | This no longer sounds fantastic, because we are already used to communicating with chat bots and voice assistants. So intelligent robots are just a further step in automation, "says Dmitry Smykalov. | |
At the same time, we can expect the emergence of a new wave of suppliers of RPA solutions specializing in AI, he said.
True, Alexander Chernikov, project manager of the RPA "First Bit. Office Sportivnaya, "doubts that the intelligent component of RPA is already ready to become a business element of the new segment of the RPA market :
| | A robot is a program that is excellent at interacting with other programs. That is, for solving problems of text recognition, it is more logical for him to use a specialized OCR solution, and in order to "become smarter" to contact an external neural network, which has a deliberately wider range of capabilities than its own built-in ML. | |
But the built-in additional capabilities of RPA platforms of this kind are still at the beginning of their development, the expert believes, and in this regard seem insufficient to solve most of the problems facing business users.
| | And in order to scale and develop the built-in machine learning mechanisms to the level of a full-fledged neural network, the developers of the RPA platform will have to retrain to the developers of the ML platform. Here it will be necessary to decide where the RPA project ends and ML begins, "Alexander Chernikov reflects and adds: However, it can be assumed that the built-in components will help create a quick simple prototype and understand whether the game is worth the candle. Nevertheless, in general, Machine Learning is of interest as an independent technology and can be useful in a number of RPA projects, said Alexander Chernikov: These are, first of all, projects where classification, information analysis, forecasting are required. But to say that ML synergy with RPA is the most promising area for the development of corporate automation, in my opinion, is premature due to the high cost of implementation. | |
This thesis is to some extent supported by recent market research. So, according to the analytical company Statista, in 2021, $10.9 billion will be spent on intelligent process automation (IPA). At the same time, less will be spent on the classical robotization of business processes and the introduction of artificial intelligence: $5.4 and 4.0 billion, respectively. Statista analysts expect RPA + AI joint business operations to grow significantly by 2023.
RPA vendors are preparing for these changes. Gartner Research Director Stephanie Stoute-Hansen notes that RPA vendor roadmaps for the development of their products reflect the desire to use more sophisticated technologies: computer vision, built-in automation, etc. For example, the company UiPath has a module UiPath AI Center that organizes and implements AI in business processes.
| | This Artificial Intelligence Center not only simplifies the deployment of models, but also provides the necessary control, continuously improving machine learning models, "explains Dmitry Smykalov. - Users can independently" drag "machine learning models into the workflow and begin to create reliable cognitive automation. At the same time, the developer of the AI process itself does not need to be a data specialist. | |
PIX Robotics this year also released an enhanced version of the innovative PIX RPA platform with the ML module. Advanced functionality allows robots to use analytics tools and other technologies to independently search, process and remember information.
Gartner analysts talk about the prospects for combining RPA and AI:
| | By 2022, 80% of organizations that have deployed RPA will implement AI. | |
Colleagues from Deloitte cite from the reasons for this development: "Intelligent automation reduces the cost of business processes."
Sergey Lozhkin, CEO of PIX Robotics, notes a more general trend:
| | So-called "smart" robots, which have "brains," begin to perform more complex processes. This is not only a trend in process robotization. In fact, now everything is becoming "smart." | |
This powerful trend, according to the expert, makes it possible to use the emerging "intelligence" of information systems in various ways:
| | Where there is a data set for training neural networks, or where you need to make any decisions. Software robots here help not only generate and receive data for training the neural network, but also can then be controlled by this neural network. | |
Moreover, in which direction do not look, everywhere neural networks will be actively introduced, the expert is sure:
| | This applies to the recognition of texts, and smart databases, and labeling in ERP systems, and segmentation of various sources, and communication with users using voice bots. The growth of cognitive functionality will continue. | |
According to Sergei Lozhkin, today the main challenge is the creation of these cases and the training of neural networks.
Moving towards IPA
In general, Deloitte analysts identify three stages in the development of RPA intellectual abilities that correspond to the level of realized intellectual capabilities:
- Basic RPAs: Can automate multiple processes, but with relatively low business impact.
- RPA with AI elements, which may include, for example, machine learning elements. They are often called IPA today, although, by and large, they are not yet full-fledged products of intelligent automation. This layer provides intelligent document processing.
Since, according to expert estimates, more than 80% of these organizations are in documents and forms, the digitization and processing of the totality of this information is the key to any transformation. Process mining solutions and analytical capabilities to implement non-trivial RPAs are also emerging at this level.
All leading RPA vendors now have a basic platform - the "operating system" for bots, which provides a mechanism for creating and managing bots. You can connect additional components to the bots, provided by both the vendors themselves and their partners.
- Full-fledged IPA mechanisms: they are characterized by the capabilities of end-to-end automation of processes and the transformation of a set of autonomous RPAs into a new entity - the "digital workforce." This implements intelligent automation of complex tasks, implying cognitive automation.
Cognitive automation techniques assume that information systems capable of working with unstructured data and implicit knowledge are used for decision making.
Schematically, the evolution of automation of robotization processes can be seen as a series of steps towards ever higher levels of automation and artificial intelligence:
- RPA. Software robots that automate repetitive, rule-based processes using structured data
- Cognitive automation. Automate more complex processes using unstructured data with machine learning capabilities
- Digital assistants. Robots equipped with speech and text user interfaces using natural language processing capabilities
- Stand-alone agents. Comprehensive software systems capable of leveraging deep learning algorithms for self-decision and process initiation to automate critical business functions
Clearly, the number and value of RPA/IPA-derived business opportunities is growing as we move from underlying RPAs to AI-based RPAs and then to digital assistants.
For example, a study of Smart RPA implementation projects, which Everest Group analysts carried out in 2018, showed that Smart RPA investments bring benefits to enterprises. Their businesses that previously deployed RPA, about 20% implemented AI in combination with existing RPAs. On average, enterprises have achieved 30-40% improvement in operational performance, including process accuracy, cycle time, staff productivity, and SLA compliance in the RPA/AI direction. The report also indicates that, on average, companies achieved ROI in RPA/AI during the first 12 months after the introduction of smart robots. The average cost savings was 30%, and more than half of the enterprises noted a strong or very strong impact on improving the quality of service to their customers.
It should be noted that in parallel with the "intelligence" of bots, the overhead costs of implementing Smart RPA increase sharply. This issue requires a detailed study in each company that decides to develop the intellectual component of its business processes. Experts urge to remember that technological progress inexorably makes equipment cheaper and more expensive for people, and RPA and especially IPA are not related to hardware and software complexes, but to human activity. Robotization saves the work of ordinary employees: accountants, lawyers, operators, etc., but the cost of IT employees at the same time only increases in proportion to the complexity of IT projects.
| | It is already clear that there are areas in which expectations from the use of AI are clearly overestimated, but there are areas where, on the contrary, it has shown itself successfully, and it is possible to move to replication. | |
For example, if the accuracy of the receivables/payables clearing task model is 60%, it means that up to 40% of transactions are processed manually, but the risk of robot error remains.
| | This is a typical example of overstated expectations from smart robots. It seems that if we are talking about accounting and tax accounting, it is better to reinsert and entrust this work to a person, "says Gimranov. | |
An intelligent information robot will be functional, the expert believes, only if there is a well-marked, prepared data model for training this robot.
| | And if there is no such model, or the conditions have changed, or the business context has changed, the model must be retrained again, generate a flow of training data, "says Rinat Gimranov. - And this is a separate serious activity that needs to be done, and which also reduces payback. | |
Based on considerations of economic expediency, he says, it is often easier to make an traditimic robot in those areas that are clearly described, and leave people in areas with a fuzzy context.
In other words, the problem of scripting descriptions inherent in basic software robots moves to the level of intelligent robotization, taking the form of a model problem. These considerations define the specifics of the further evolution of corporate automation at the business process level.
Towards Hyperautomatization
Gartner experts believe that RPA technologies will move from automating individual tasks to more, automating increasingly complex tasks, gradually choosing automation at the level of complex end-to-end processes - a state that is called hyperautomatization.
| | Streaming of standard documents, for example, primary accounting payments, receipts, etc., is an excellent platform for maximum automation, the so-called DPA (Digital Process Automation), "says Konstantin Istomin, executive director of Directum. - Hyperautomatization is recognized by world analytical agencies as one of the main IT trends for 2022. | |
Digital Transformation and Digital Employees
Market experts note that one of the significant barriers to the development of RPA systems in Russia is the lack of a corporate strategy for digital business transformation. The system approach to digital transformation will obviously help to move from the state of point testing of RPA solutions to the phase of systematic scaling of automation, because automation of smart processes allows you to create chains from software modules that are launched at specific points in business processes. At the same time, RPA technologies allow you to "glue" internal and external applications, and microservice architecture allows you to get away from monolithic applications in IT architecture, notes Andrei Koptelov, vice president of ABPMP Russia.
Obviously, as you move from underlying RPAs to IPAs, the lines between RPAs, business process management (BPM), and hyperautomatization are erased. These concepts are essentially the three main aspects of digital transformation. Moreover, as Comindware says, at the stage of transition to hyperautomatization, the basis of digital transformation is formed, in which traditional RPA and BPM play the role of instruments "orchestrating" digital processes.
Accordingly, the set of autonomous RPAs is transformed into a new entity - the "digital workforce."
| | Here we are actually talking about digital employees: there is a transition from separate terms: RPA, AI, etc. - to the concept of "digital employees," which solves the problems that form a certain business process, "says Pavel Borchenko. | |
Indeed, the software robot is becoming more and more like a real employee, as it smarts and gains the ability to solve more complex tasks, and, therefore, its value for business increases.
| | The ability of the RPA platform to manage the creation of digital employees and monitor their work is the value that companies receive, the expert emphasizes. | |
A key parameter of the digital workforce is the flexibility to create and modernize smart digital processes. It is hyperautomatization that helps CIO in these matters, says Brian Burke, Gartner's vice president of research: helps identify and execute processes to automate. The pandemic in this case played a driving force, he notes: more than 70% of commercial ones revised their digitalization plans and gave the green light to initiatives of business units in the field of hyperautomatisation. Today, hyperautomatization is inevitable and irreversible, Burke believes.
Technological implementation of hyperautomatization
Gartner analysts introduce the term DigitalOps. It implies a management approach that combines management and digital technology. This means that business planning, management and control is based on the analysis of data that comes from all departments and departments, and through AI mechanisms that are able to work with all the many accumulated corporate data. For example, AI processes scanned documents or telephone records, generates analysis results of analysis data in the form of reports intended for decision making and further actions
Source: Move Beyond RPA to Deliver Hyperautomation, Gartner, December 2019
The DigitalOps model uses basic RPA mechanisms to automate routine operations and integrate data. In this sense, the RPA "non-invasiveness" property is important, that is, the ability to automate processes without changing the systems they work with and almost without affecting the processes themselves. RPAs organize interactions with legacy applications, where traditional integration will be complex, time-consuming, and expensive, and migrate, consolidate, and validate data from disparate sources as part of a large-scale ERP migration project. In addition, RPA is useful for quick experimentation and proof of concept before launching a new service or process.
Data transfer can also occur between machine and machine due to [[Internet of Things (IoT)|[[Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|[[Internet of things of Internet of Things (IoT)|Internet of Things (IoT)]]]]]]]]]]]]]]]]]]]]]]]]]]].]] This approach allows the business to easily rebuild due to any changes and adapt to new customer requests.
Along with RPA, the DigitalOps toolkit also includes intelligent business process management (iBPMS) systems, integration platforms as a service (iPaaS), process analytics, and policy-based business decision management systems.
Then we can talk about the evolution of the technological architecture, in the center of which is RPA and its environment (it is called in Gartner "augmented RPA" - Complemented RPA, CoRPA) with simple systems for integrating business processes (Task-Based systems) into the tools DigitalOps Toolbox.
This means that the state of DigitalOps Toolbox corresponds to an integrated platform of higher order, which will provide the ability to freely combine a wide range of digital technologies: BPM, RPA, ML, chat bots, Process Mining, etc.
It should be noted that on a full scale, not a single vendor has yet implemented such a platform, but everyone is actively experimenting with integration. It is possible that the role of the core of such a digital business platform in the hyperautomatization phase is best suited to the corresponding BPM platforms, say Comindware. Moreover, RPA during its smart evolution is rapidly turning into BPMS subsystems.
According to Sergey Lozhkin, today all the main RPA platforms declare the creation of an ecosystem when the robot can perform customer service from beginning to end:
| | They strive for this, they develop it, and this will continue. Now there is an increase in expertise in industry specifics, in the specifics of specific business functions, creating and honing robots that can work out of the box. Accordingly, robots perform both the old function of communicating one system with another and are an interface and end-to-end digitalization. | |
| | You need to be careful when they say that the robot can do everything. It is very important that the robot be embedded in the current IT infrastructure and balanced with classic systems, which the robotization platform complements. | |
Dmitry Smykalov says that RPA, artificial intelligence, machine learning, automation of cognitive processes, process mining (Process Mining) are just some of those advanced technologies that are used in hyperautomatization (iBPMS).
| | The architecture and technological filling of robots, of course, is constantly improving, "he notes." But key changes are associated with the possibility of creating robots without programming skills or attracting external developers, with simplifying interactions between people and robots, analyzing the effectiveness of implementing RPA solutions, and, of course, integrating artificial intelligence into automated processes. | |
Anatoly Belaychuk focuses on the fact that today professionals operate on terms that often have fuzzy definitions.
| | The most terrible examples of the use of RPA robots that I had to see are the automation of end-to-end business processes, "he says." It looks like this: the robot opens e-mail, extracts an Excel file from the letter, reads data, performs some kind of manipulation with them, then sends output data by mail to the next participant in the process. | |
To name these actions Robotic Process Automation means to mislead people, the expert is sure:
| | In fact, RPA robots do not automate the process, but the tasks performed during the process. | |
Robots have also been invented to coordinate tasks within the framework of the end-to-end process, says Anatoly Belaychuk:
| | But these are not RPA, but process engines in BPMS systems. It turns out that RPA and BPMS are an ideal combination: the BPMS robot as a conductor, the RPA robots as musicians, and all together - the optimal process, which is carried out with minimal cost. | |
On the one hand, this confirms that the trend towards integration of these two technologies - BPM and RPA - is objective. On the other hand, this does not mean that vendors who offer all the tools of DigitalOps on their platform will definitely win the market struggle. Moreover, today the trend of easy and flexible integration of digital tools and assets looks more preferable. But along the way there are their "pitfalls" and "silver bullets."
How will the digital workforce be put into practice?
All comparisons of the current RPA and IPA technologies with employees are correct to a certain extent: they focus on the processing of data integration from various sources, etc., but the question of the actual knowledge of a digital employee remains open.
Rinat Gimranov, head of the information technology department of Surgutneftegaz, believes that the path of evolution that software robots are taking today does not provide the opportunity to "act like a person": not to press buttons like a person, but to "understand" what actions are being taken and make decisions on their own in a situation of uncertainty. Indeed, it is important for both the robot and the person to understand the context in which one or another activity takes place. But for a robot, this context must be spelled out in much more detail than for a person, because he does not have any initial "background" knowledge about his activities.
| | A good platform for RPA should have some tools that allow you to describe in detail the entire context of production activities, eliminating any uncertainty. For example, to take the necessary information from some systems, "Rinat Gimranov reflects. | |
In principle, the industry already has suitable tools.
Semantic Data Layer
A fundamental barrier to improving the efficiency of enterprise data processing is the fact that valuable digital information is stored in various places: enterprise data warehouses, on local servers, in data centers or cloud environments. At the same time, business users do not understand the language in which IT professionals describe this data. Until now, this internal data fragmentation, both structural and mental, has not affected business processes. But in the era of the transition to hyperautomatization, the previous way of managing corporate data is becoming a brake on the evolution of automation.
Overcoming this barrier, the transition to semantic descriptions of the structure of corporate data is logical: it assumes that the basic element of IT systems is not the data itself, but its semantic representation. The second important thing is the transition to a single enterprise data warehouse that supports a semantic description of data.
Gartner analysts point to the importance of the semantic level in their report "How to Use Semantics to Drive the Business Value of Your Data" published in 2018: "An unprecedented level of data scaling and distribution makes it almost impossible for organizations to effectively use their information resources. Data leaders and analysts should take a semantic approach to their corporate information resources, otherwise they will lose the battle for competitive advantage. "
Today, the semantic representation of data is actively used by vendors of analytical systems, and the trend of supporting the sematic layer of data in enterprise repositories is also beginning to unfold.
Source: Kyligence, 2018
For example, the "Data Cloud" of Sberbank (Data Factory) - a Big Data cluster based on Apache Hadoop - for effective management of data downloads from highly loaded banking systems at the level of several tens of terabytes per day contains a Single Semantic Layer (ESS) of data. The main task of the ESS is to provide business users with a single way to access consistent and high-quality data, moreover, in the language of business users. The semantic description of ECC data simplifies the integrated processing of data, ensures the transparency and understandability of all data operations.
Source: Sberbank, 2018
The semantic data layer is essentially a business abstraction that is above technical IT concepts and therefore is able to uniformly support business logic, hierarchical relationships, calculations, etc.
Implementation options for the semantic data layer
Graph of Knowledge
Gartner analysts recognized this well-known method of semantic description of objects and relationships in 2018 as a key new technology in Hype Cycle of artificial intelligence: "The growing role of content and context for providing information using AI technologies, as well as recent proposals for the use of knowledge graphs for AI applications, brought knowledge graphs to the forefront of AI technologies."
In 2020, the company MicroStrategy in its study "10 Enterprise analytics trends to watch in 2020," added knowledge graphs to the Top 10 key trends in corporate data analytics:
| | The semantic graph becomes paramount to ensure business value. It captures, organizes and enriches metadata using a graph representation, and uses graph analysis methods to obtain information. | |
| | The semantic graph combines the capabilities of graphics and the presentation of metadata in order to provide an expanded semantic representation of the data landscape in order to support analytical workflows - this is how Gartner explained why knowledge graph technology is placed in the Top 10 most relevant trends in data processing and data and analytics for 2019. | |
Analysts MicroStrategy emphasize that the semantic graph will become the basis supporting data and analytics in an ever-changing data environment. So, it can use the data and the results of their processing in different systems, supplement them with AI and ML, etc. In particular, the semantic knowledge graph provides transparency in the use of data.
Unlike relational structures, the knowledge graph allows you to combine data, even if their properties are heterogeneous or the definitions of entities differ depending on the use case. For example, it is possible to unite data from relational databases, NoSQL DBMS, from documents and to attach to them geospatial data. You can simultaneously use data from structured, semi-structured, and unstructured sources in a single graph to help create a single data layer enriched with the full context of the data from each source. Stardog goes beyond simply supporting graph-based integration by offering a unique combination of graphics, virtualization, and output.
I do not see anything so special, different from others. If we talk about promising tasks, then just here is a description of the context and setting the task to a clear business process that the robot should do, it seems to me that ontological models and an ontological approach would be a more correct tool for this.
Unfortunately, I have not yet seen mature solutions in which this would be implemented. Here, say, Comindware contains just inside the graph DBMS and the semantic engine. That is, they could potentially take on this, but so far I don't see such functionality there either. Although it is purely theoretically possible to assume that I use BPMN 2.0, which supports Comindware, describe the entire process that is done by the robot, and launch it into work. That is, it is a system that appropriately generates the desired context, depending on the description I give.
Ontologies
Ontologies - a variant of the knowledge graph - can be called the current leader in the formal methods of descriptions of various subject areas. The Ontology Web Language (OWL) is used to describe the semantic Semantic Web and has already become the international standard for ontological descriptions.
Using ontologies, you can create a single description of all objects, properties, and relationships that cover all business processes and aspects of a company's business. We can say that the language of descriptions of ontologies allows you to create a semantic digital twin of real business.
An important feature of ontological models is that they are easy to expand and modernize, including by quickly connecting a wide variety of enterprise-wide information systems, including accounting transaction systems, analytical systems, management decision support systems, technological production management and dispatching systems (SCADA).
The ontological approach to describing business processes is used, for example, in the Comindware Business Application Platform system.
For hyperautomatization purposes, it is important that ontological models provide a uniform approach to the implementation of both RPA and BPM mechanisms, as well as various data integrations in terms of processes and objects based on semantic descriptions.
The only nuance, Rinat Gimranov notes, is that in order to implement informobots and hyperautomatization in this style, it is necessary to significantly change the current platforms.
Two options for implementing the digital workforce
Depending on the chosen path of evolution of robots towards hyperautomatization: current or semantically oriented, the resulting digital workers will differ significantly. Rinat Gimranov explains the difference by example: in order to get a self-driving car, you can put an "iron" robot in a regular car cabin and teach it to behave like a professional driver: twist the steering wheel, press the pedals, etc. However, a real unmanned vehicle does not need a cabin with its traditional elements at all. "Such a software robot is literally built into the body of the machine, it copes perfectly with its work without the controls inherent in a person," the expert says.
By the way, both options are different implementations of the idea of a digital workforce. Only in the first case it is placed in an existing car. And this has its own plus - it is not necessary to radically rebuild the vehicle, as well as its minus - extra parts that create points of failure of the equipment. In the second case, the most efficient robot is obtained, which optimally works in the area of work entrusted to it. But the fee for this is a radical processing of the infrastructure.
It will be interesting to see how these two approaches will evolve over time, and what the technological competition between them will look like.
See also
- RPA (Robotic process automation)
- Russian RPA Systems Market
- RPA Solutions and Projects Catalog
- Why is RPA becoming technology of the year?
- New RPA Manifesto
- Differences in development on RPA platforms UiPath and Kofax RPA
- Overview of the BPM system in Russia
- Chat-bots Virtual interlocutors
| The site content is translated by machine translation software powered by PROMT. The machine-translated articles are not always perfect and may contain errors in vocabulary, syntax or grammar. Read original article If you find inaccuracies or errors in the results of machine translation, please write to editor@tadviser.ru. We will make every effort to correct them as soon as possible. |