KR20190021877A - Method and system for ontology-based big data harness for accelerated machine learning and big data analytics - Google Patents

Abstract

The present invention relates to a method which dynamically generates and manages metadata based on a data classification system and ontology in order to automatically represent semantic features of the metadata. Moreover, the method provides a smart data framework or an intelligent data framework for accelerated machine learning or deep learning, which are not necessary to develop each connection and each pre-processing software for utilizing big data.

Description

TECHNICAL FIELD [0001] The present invention relates to ontology-based large data access and utilization methods and systems for accelerating machine learning and big data analysis,

Field of the Invention The present invention relates to information technology and related fields, and relates to a method and apparatus for accessing and acquiring big data utilized in machine learning or " deep learning ", cognitive computing or cognitive analytics and to methods and systems for data management and utilization for accelerating machine learning and cognitive analysis by effectively exploiting and ingesting data.

Due to the lack of standardization of data handling the syntactic aspects of data as well as the insufficiency or absence of standardization for the semantic (content and semantic) nature of Big Data, machine learning, which utilizes vast multidimensional and heterogeneous data, And has remained a technical void.

The present invention solves the problems associated with connecting multidimensional and heterogeneous big data by systematically presenting semantic properties of data through dynamic generation management of metadata based on taxonomy and ontology I want to.

The present invention allows machine learning algorithms to access, acquire and analyze large data without the need to develop software for a variety of different data adapters and preprocessors, and without prior knowledge of individual data. I can do it. Data analysis software or machine learning algorithms can systematically discover the semantic and syntactic characteristics of the Big Data at the time of connection and acquisition without prior knowledge of Big Data using a " Smart Data Interface ".

The present invention provides a framework for accelerating machine learning or deep learning and improving efficiency by eliminating the need to develop software for individual data adapters and preprocessors to utilize the Big Data do.

The present invention is used to automatically generate and manage metadata describing the semantic properties of the data based on a data classification scheme and an ontology at the time of data collection. In other words, at the time of data collection, metadata for the semantic and syntactic properties of the data is automatically generated, and the metadata is stored using the big data and the universal resource locator (URL) Loosely coupled logically. A " smart data interface " is provided so that analysis software or machine learning algorithms can discover the semantic and syntactic characteristics of the data without prior knowledge of their respective data.

The present invention relates to a " smart data framework " that automatically detects semantic and syntactic characteristics of various big data at the time of data connection, through a predefined " smart data interface " Will realize the concept of " intelligent data framework ".

The method and system of the present invention can be independently implemented and operated independent of various hardware systems, operating systems, network infrastructure, or system implementation language, regardless of their development and operating system.

A number of prior art techniques have been published for automatically generating and managing metadata to address the technical problems associated with machine learning or big data analysis described above. A number of prior arts are also referred to as methods and systems for " Self-Describing Data " or " Smart Data ", and solve technical problems related to the connection and utilization of various types of large data .

However, such prior art solves some of the syntactic aspects of the data (the organization and format of the data) and does not provide a solution to the problems with the semantic properties of the data (content and meaning of the data).

SUMMARY OF THE INVENTION The present invention solves the above-mentioned problems of not only the syntactic as well as the semantic aspects of data, but also enables data analysis software or machine learning algorithms to be implemented through predefined " smart data interfaces " To a novel method and system for discovering semantic properties of data.

In conclusion, the present invention relates to a method and system that is entirely different from any prior art that includes or describes "Self-Describing Data", "smart data", or similar names, Is a new method and system for solving the problems of semantic properties (data content and meaning) of big data.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention to those skilled in the art Other drawings can be obtained based on these figures without an inventive task being performed.
Figure 1 illustrates an overall system architecture for ontology-based big data utilization that accelerates progressive machine learning or deep learning.
FIG. 2 illustrates a method for automatically generating metadata that can be accessed and utilized through a predefined " smart data interface " without the need to develop individual data access and preprocessing software to access the data, And the overall process are illustrated.
Figure 3 shows process flows and sequences for systematically generating data that can be accessed and utilized through a predefined " smart data interface " without prior knowledge of the data.

The following detailed description of the invention refers to the accompanying drawings, which illustrate, by way of example, specific embodiments in which the invention may be practiced in order to clarify the objects, technical solutions and advantages of the invention. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention.

Throughout the detailed description and claims of the present invention, "learning" is a term referring to performing machine learning, for example, deep learning, etc., in accordance with procedures, and is not intended to refer to mental actions such as human educational activities It will be understood by those of ordinary skill in the art.

And throughout the description and claims of the present invention, "data" refers to data, raw data, filtered data, preprocessed data, curated data, analyzed data, manipulated data, Refers to all sorts of data, such as data, synthesized data, fabricated data, etc., and a 'data entity' is a term that is useful, filtered through a minimum of filtering and preprocessing, It is a term referring to a collection of data.

Next, throughout the description and claims of the present invention, the term " capture data " and variations thereof indicate that the collected data is initial raw data generated on a particular machine While the term 'data payload' refers to the formatting of a data object created by initial filtering and eliminating noise to a specific specification Which may include a data header or a trailer and may also be encrypted or compressed. Here, the filtering is different from the preprocessing in the data ingestion of the present invention.

Also, 'metadata' used in the present invention is a term referring to 'semantic and syntactic information about data payload of a data entity' as will be described in more detail below.

And throughout the specification and claims of this invention, the term " comprises " and variations thereof are not intended to exclude other technical features, additions, elements, or steps. Other objects, advantages and features of the present invention will become apparent to those skilled in the art from this description, and in part from the practice of the invention. The following examples and figures are provided by way of illustration and are not intended to limit the invention.

Moreover, the present invention encompasses all possible combinations of embodiments shown herein. It should be understood that the various embodiments of the present invention are different, but need not be mutually exclusive. For example, certain features, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in connection with one embodiment. It should also be understood that the position or arrangement of individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the present invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is to be limited only by the appended claims, along with the full scope of equivalents to which such claims are entitled, if properly explained.

Unless otherwise indicated herein or clearly contradicted by context, items referred to in the singular are intended to encompass a plurality unless otherwise specified in the context. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The present invention provides a new method and system for automatically generating and managing metadata of big data for machine learning based on a data classification system and an ontology. By allowing data analysis software or machine learning algorithms to discover the content and meaning of various types of big data that are diverse, heterogeneous, distributed, and low veracity, And potency. Here, 'low reliability' means that there is a doubt about the quality and / or authenticity of the big data obtained from the Internet, Internet of Things ("IoT") or CPS equipment (Cyber Physical System device) , Additional data preprocessing, validation, and curation procedures are needed.

It should be understood that the present invention can be practiced in a variety of different embodiments, and the " computing entity " referred to in such embodiments is intended to encompass a wide variety of data, Quot; is a term that refers to a data processing device that is a machine that is capable of processing data.

For example, the computing entity may include existing von Neumann machines and associated software platforms, and may be a general purpose computer, workstation, portable or mobile computing device. Such computing entities may be connected to an Internet connection or other networks (e.g., a local area network (LAN), or a wide area network (WAN)). The connection of the computing entities to the Internet, intranet or internal network may be a wireless connection or a wired connection. LANs, WANs, intranets, internal networks, and Internet connections need not be described in detail because they are well known to those of ordinary skill in the art. It will be appreciated that a computing entity may be referred to as a node that is part of a data network, such a data network may include a plurality of nodes coupled by a communication path.

In other words, the computing entity may be a computer platform that includes a processor that runs software and acts as a node that can exchange data with other nodes in the data network over a wireless or wired communication link. The computing entity may be a mobile device or a static device.

Computational entities are not limited to the examples described above, and ordinary technicians may use neuromorphic processors, molecular DNA computers, photonic computers, quantum computers, And the like can be used in the present invention.

In addition, the computing entities referred to herein and their networks not only exchange data between other nodes, but also allow software to migrate to various nodes for performing specific purposes assigned to the software (mobile software migrating to various system node based on its itinerary) You can perform specific tasks on each node.

In some other embodiments of a computing entity, the computing entity may be implemented by at least one portable or non-portable computer having one or more databases stored in a computer-readable memory, A portable or non-portable computer can also have at least one computing unit or processor programmed with software, and when the software is executed, a series of steps included in the software are executed.

It will be appreciated by those of ordinary skill in the art that such a " computing entity " is not limited to a von Neumann type of computer, and any device capable of executing all or at least some of the software according to the present invention may be included in the & , And such a " at least one computing entity ", as a whole, will be included as an embodiment of the present invention. It is also evident that such interworking will be accomplished through the wireless or wired connection described above.

The present invention implements " smart data " or " intelligent data " that can automatically detect semantic properties as well as syntactic characteristics when a machine learning algorithm or data analysis software acquires data without prior knowledge.

The present invention augments prior art describing the syntactic properties of various data using metadata.

The present invention provides a method for allowing machine learning algorithms or data analysis software to enable transparent connection and acquisition without having to know all the big data, its contents or semantics, or the data structure and format.

Semantic properties of data (eg, protein structure, human gene sequence, protein structure, 3-D structure of rock taken from the Moon, brain image of the brain) Based on the data classification system (taxonomy) and ontology is established, and ontology based on the metadata to automatically create and manage a framework to implement.

The term " hyper-ontology " referred to in the present invention refers not only to terminology and synonyms for a specific domain, but also to integrative terms and synonyms associated with other domains related to the specific domain, ontology. For example, a medical-related hyper-ontology is not only a jargon about medical and pathology, but also related fields such as biotechnology, genetics, biology, biochemistry, food and nutrition, physics, psychology, philosophy, environmental engineering, theology, sociology, And terminology and synonyms.

The semantic category of data refers to the content and meaning of data such as chest x-ray, EEG, heart sound, gene expression results, protein structure, human gene sequence, protein structure, The three-dimensional structure of the rock, the meaning of the data such as the diagnostic image of the brain, the data owner and version, the data creation date and time, and the access restrictions for data security.

Syntactic properties of data are data standards such as binary, string, sequence, image, XML, tabular, CSV, BSML, MAGE-ML, CDA / HL7, PSI, Notation is included. For data protection and security, encryption can specify software for decryption as part of the syntactic property. In addition, when compressing data, software for decompression can be specified as part of the syntactic property.

In this case, software for data acquisition and preprocessing, data acquisition software according to data standard notation, software for data decoding, and software for data decompression can be dynamically used as needed according to the syntactic characteristics of the big data (Dynamic software loading and binding on demand).

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily carry out the present invention.

As depicted in FIG. 1, application programs that utilize large amounts of heterogeneous data, such as machine learning algorithms 101 or big data analysis software 102, use predefined " smart data interfaces " It is possible to access and utilize various data without having to develop software for data connection and preprocessing separately for data.

The metadata generation and management framework 107 that allows the machine learning algorithm 101 or the big data analysis software 102 to access various heterogeneous data through a predefined " smart data interface " Specific controlled vocabularies (105) and a synonym collection (106) that complement the hyper-ontology.

The metadata management framework 107 collects data entered by a person or generated by a machine, and at the same time automatically generates the metadata 108. Metadata 108 includes the semantic and syntactic attributes of the data 110.

Examples of semantic properties of data and examples of syntactic properties of data entity 110 are depicted in table 109. [

A data scientist may participate in the definition of the hypertextoligon 104, the related vocabulary 105 and the synonym collection 106, and the metadata attributes 109 ) May be involved in the definition of domain experts.

In actual implementation, much more information than the table 109 illustrated in FIG. 1 may be included in the metadata 108.

The data path or URL for connecting to the real data entity 110 is included in the metadata 108 as shown at the end of the table 109. [ The connection of the data object automatically finds its actual storage system and its location according to the URL so that the data object 110 can be stored everywhere.

As depicted in FIG. 2, the method and process for systematically creating and managing metadata for the Big Data, which may be accessed and obtained via the " Smart Data Interface " 216, Process. One is data 201 automatically generated from the machine and the other is data 207 manually entered by a person.

In the case of the data 201 automatically generated from the machine, the configuration of the machine and the runtime parameters are provided through the man-machine interface 202. Examples of machine configuration and operating parameters are depicted in table 109 of FIG.

As an example, the configuration regarding the configuration of the machine may be performed by a domain expert, and the operating parameters may be performed by an operator of the machine. For example, a functional MRI used in a hospital can be used by a radiology specialist in the radiology department to perform its initial setup and configuration (i.e., to set up the machine configuration) The operating parameters (e.g., information about the patient, date and time of the scan, information about the physician requesting the scan, purpose of the scan, operator's ID, etc.) can be performed by the MRI machine operator.

When new data is generated (203) from such a machine (e.g., functional MRI), metadata is automatically generated for the big data generated from the machine, depending on the machine configuration and operating parameters 204).

The data 205 generated from the machine is passed through the big data memory container 212 after the noise removal and formatting operation and then the storage system is designated 213 and the data path or URL of the storage system is assigned to the metadata And the completed metadata 214 is generated.

The completed metadata 214 is stored 215 by specifying a place in the data hierarchy according to the data classification scheme (taxonomy) of the corresponding domain. The hierarchical structure of the metadata and data objects is defined by data scientists and data modelers based on the classification system of each domain. Based on the hierarchical structure, data modelers and database experts database specialist.

If the newly created data is a new version of the already existing data, the past version data and related metadata are maintained in different versions in the system, and the latest version is placed at the top of the cascade in the data hierarchy . If there are multiple versions of data, via the smart data interface 216, they are connected to the latest version of the data. When using the smart data interface 216, the old version of data can be connected to a specific version by specifying a specific version number or generation time zone.

In the case of human-input data 207, when the data entry 210 is started, the system requests information about the semantic properties and syntax attributes of the input data through the human-machine interface 208, (209). Metadata 212 for manually entered data is automatically generated based on information 209 about semantic properties and syntax attributes entered by a person.

The data 210 input by a person is subjected to a formatting operation according to the input device and is then collected as input data 211.

The flow of subsequent processes 212 to 219 is the same as the process flow related to the data processing generated from the above-mentioned machine, and a description thereof will be omitted.

The data analysis software or machine learning algorithm 219 makes a data connection using the smart data interface 216. At this time, the data analysis software or the machine learning algorithm 219 can search the metadata by a key word search using a conventional method. However, in the system according to the present invention, taxonomy-based hierarchical tree structure (215).

The data consuming entity, such as data analysis software or machine learning algorithms, analyzes the metadata to retrieve the semantic properties of the data (217).

The syntax attributes of the data are retrieved from the metadata and the parser 218 and the data parser of the data are automatically loaded in the library to access and acquire the necessary data 213.

The detailed process flow and the individual steps and sequences for systematically generating and managing metadata are depicted in FIG. The process flow and the individual process steps have already been described above with reference to FIG. 2, and the description thereof will be omitted. Instead, the key technical innovations of the present invention are summarized as follows.

The essence of the present invention is that data analysis software or machine learning algorithms, without prior knowledge of the data, can discover semantic properties as well as syntactic attributes for data at the time of data access.

In the present invention, a unified indexing mechanism is used to provide transparent access to various, heterogeneous, and high-dimensional big data. This integrated indexing mechanism, coupled with data classification schemes and ontology - based dynamic metadata management, enables transparent access and utilization of multidimensional heterogeneous big data as well as dynamic metadata creation and management.

In the present invention, the semantic categories (e.g., protein structure, human gene sequence, protein structure, crystallographic properties of moonstone, brain PET scan, chest X-ray, EEG, heart sound audio, gene expression data) And version (e.g., date and time of creation, agency) by generating ontology-based metadata. Furthermore, the present invention discovers syntactic characteristics such as syntactic characteristics (e.g., binary, string, image, XML) and data formats (e.g., BSML, MAGE-ML, CDA / HL7, PSI, DFDL) To dynamically utilize various data parsers as needed to increase the efficiency and effectiveness of data analysis software or machine learning algorithms by eliminating the need for individual data access and preprocessing software.

The metadata may use XML notation or other similar notation for human readability as well as data analysis software or machine learning algorithms.

It is apparent that the technical idea of the present invention described above should not be construed as being limited to practical examples such as the above-mentioned examples.

Claims (8)

At least one computing entity may automatically and systematically generate and manage metadata for the semantic properties and syntactic attributes of the big data based on a data classification scheme (taxonomy) and ontology at the time of data generation ,
(a) metadata for machine-generated data is automatically generated and managed based on machine configuration and operating parameters,
(b) metadata about human-generated data is generated and managed based on human input through human-machine interaction. The method according to claim 1,
(i) jargon and synonyms for a particular domain; And
(ii) jargon and synonyms in other fields related to the particular domain;
Characterized in that an integrated holistic hyper-ontology is used by fusing the hyper-ontology. The method of claim 2,
The metadata includes:
(a) metadata defining the semantic properties of the big data based on the hyper-on -tology associated with the domain; And
(b) metadata defining a syntactic attribute of the big data;
≪ / RTI > The computing entity stores and manages the big data and the corresponding metadata according to a hierarchical structure based on the data classification scheme according to claim 1,
(a) the computing entity searches for the metadata according to a conventional search mechanism based on a key word or phrase, and a process for directly searching for the metadata along a hierarchical tree structure Performing at least one;
(b) if there are multiple versions of data, the computing entity manages all versions of the data and associated metadata in a common stack, placing the latest version of the data and associated metadata at the top of the stack Thereby automatically connecting to the latest version of the data if there is no specific request; And
(c) directing, by the computing entity, the data directly via a universal resource locator, regardless of the actual physical location of the data;
≪ / RTI > The method according to claim 1,
Machine learning or data analysis software implements a smart data interface that is provided to connect big data without development of individual data access and preprocessing software for each specific data and without prior knowledge of the data How to. The method according to claim 1,
A method for implementing a human-machine interface used when connecting to various big data without prior knowledge. The method according to claim 1,
A method for implementing an integrated, independent software system for accessing and utilizing big data, regardless of the hardware platform, operating system, or programming model used as the computing entity. A program recorded on a machine-readable non-volatile storage medium, comprising instructions that direct the computing entity to perform the method of any one of claims 1-7.

Images