KR20150112441A - Knowledge acquisition system based on un-structured data for never-ending and self-evolving - Google Patents

Abstract

Provided is a knowledge acquisition system capable of self-learning and self-evolving. The knowledge acquisition system may include: a data collecting unit which collects unstructured data through a network; an analysis processor managing unit which selects an analysis module by analyzing the type of the data with respect to the unstructured data collected by the data collecting unit and configures the workflow which is an order of progressing the selected analysis modules; a knowledge analyzing unit which analyzes the unstructured data collected by the data collecting unit based on the configured workflow and the analysis modules selected by the analysis processor managing unit and maps the knowledge data; a knowledge verifying unit which performs a verification with respect to the knowledge data mapped by the knowledge analyzing unit; and a knowledge model managing unit which generates a knowledge model by storing the knowledge data verified by the knowledge verifying unit, wherein the knowledge analyzing unit maps the knowledge data by referring to the knowledge model generated by the knowledge verifying unit.

Description

Technical Field [0002] The present invention relates to a knowledge acquisition system based on unstructured data,

The present invention relates to an infinite knowledge-based evolutionary self-learning knowledge acquisition system based on unstructured data, and more particularly, to a knowledge acquisition system capable of self-learning by understanding self-knowledge and learning infinite knowledge in order to acquire knowledge by analyzing unstructured data .

The present invention is derived from research conducted by Korea Electronics and Telecommunications Research Institute and Saltlux Co., Ltd. as a part of SW Computing Industry Source Technology Development Project (SW) of Future Creation Science Department.

[Research period: 2013. 05. 01 ~ 2014. 04. 30, Research institute: Korea Industrial Technology Evaluation & Management Service, Research title: Development of Intelligent Evolutionary Wise QA Platform Technology for Human Knowledge Enhancement Service, Assignment No.: 10044577 ]

In order to generate knowledge from unstructured data, it is important to collect vast natural words, to extract and acquire knowledge from the contents, and to find the solution of the optimal knowledge learning method. This knowledge enhancement process requires a large amount of unstructured data and is limited in finding optimal solutions due to knowledge mapping from natural language through managers, selection of optimal knowledge learning algorithms, and natural language processing analysis techniques. And it is difficult to access the public.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and it is an object of the present invention to provide a knowledge acquisition system capable of self-learning and evolving while minimizing human intervention in a process of generating knowledge from unstructured data .

In order to solve the above technical problem, a knowledge acquisition system capable of self-learning and evolution is provided.

A knowledge acquisition system according to the present invention includes: a data collection unit for collecting irregular data through a network; An analysis processor managing unit for analyzing types of data of irregular data collected by the data collecting unit to select an analysis module and constructing a work flow that is a progress order of the selected analysis modules; A knowledge analysis unit for analyzing atypical data collected by the data collection unit and mapping knowledge data based on a workflow configured with the analysis module selected by the analysis processor management unit; A knowledge verification unit that verifies the knowledge data mapped by the knowledge analysis unit; And a knowledge model management unit for storing the knowledge data, which has been verified by the knowledge verification unit, to generate a knowledge model, wherein the knowledge analysis unit refers to the knowledge model generated by the knowledge verification unit, Mapping.

Wherein the knowledge verification unit comprises: an administrator verification unit for providing a user interface so that an administrator can perform verification on the knowledge data mapped by the knowledge analysis unit; And an error automatic verification unit for automatically verifying whether or not an error has occurred in the knowledge data mapped by the knowledge analysis unit based on the knowledge model.

Wherein the data collecting unit selects a part of the collected atypical data as a test set and the remainder as a learning set and the knowledge verifying unit analyzes the atypical data selected by the test set and verifies the mapped knowledge data The manager verification unit may be configured to perform the verification of the mapped knowledge data by analyzing the unstructured data selected by the learning set and to perform the verification of the mapped knowledge data in the error automatic verification unit.

A knowledge learning unit that generates and updates a learning model based on the knowledge model; And a learning model manager for storing the generated and updated learning models, wherein the knowledge analysis unit can map the knowledge data based on the learning models stored in the learning model management unit.

And an analysis module management unit that manages a plurality of analysis modules, wherein the analysis module management unit includes an analysis module used for analysis of the knowledge model for generating and updating the learning model among the plurality of analysis modules, - > to the knowledge learning unit.

And an analysis module management unit for managing the plurality of analysis modules to be selected by the analysis processor management unit, wherein the analysis module management unit is configured to analyze the atypical data collected by the analysis processor management unit, The analysis module can be provided to the knowledge analysis unit in a plug-in form.

The workflow execution unit may stop the configured workflow and send a notification to the administrator when there is an analysis module in which an error occurs during the analysis process in the knowledge analysis unit through the analysis module provided in a plug-in form .

Wherein the type of data analyzed by the analysis processor management unit is text, image, sound source, or image, and the plurality of analysis modules managed by the analysis module management unit include a language analysis module, an image analysis module, a sound source analysis module, Module.

The language analysis module, the image analysis module, the sound source analysis module, and the image analysis module may each be composed of one or a plurality of sub-analysis modules.

The analysis processor management unit may configure the workflow by selecting sub-analysis modules from one or more of the language analysis module, the image analysis module, the sound source analysis module, and the image analysis module.

The knowledge acquisition system according to the present invention can select analysis modules operating independently of each other, provide them in the form of plug-ins, configure a workflow for combining and arranging them, The same effect can be achieved by providing a specific analyzer having individual functions.

Further, the knowledge acquisition system according to the present invention uses a knowledge model and a learning model generated through analyzing atypical data selected by a set of test sets of a predetermined number or more, so that even if the manager does not intervene or minimizes the intervention, Infinite self-learning and evolution of the model may be possible.

1 is a schematic diagram showing a configuration of a knowledge acquisition system according to an embodiment of the present invention.
FIG. 2 is a schematic diagram illustrating a configuration and operation of a knowledge acquisition system according to an embodiment of the present invention for explaining knowledge acquirement for unstructured data selected by a test set. FIG.
FIG. 3 is a schematic diagram illustrating the construction and operation of a knowledge acquisition system according to an embodiment of the present invention for explaining knowledge acquirement for unstructured data selected by a learning set.
4 is a schematic diagram showing a main part of an analysis process management unit of a knowledge acquisition system according to an embodiment of the present invention.
5 is a schematic diagram showing a main part of an analysis module management unit of a knowledge acquisition system according to an embodiment of the present invention.
6 is a conceptual diagram for explaining a function of a workflow execution unit of a knowledge acquisition system according to an embodiment of the present invention.
7 is a schematic diagram showing a workflow for acquiring knowledge from unstructured data that is text in a knowledge acquisition system according to an embodiment of the present invention;
8 is a schematic diagram illustrating a workflow for acquiring knowledge from unstructured data that is an image in a knowledge acquisition system according to an embodiment of the present invention.
9 is a schematic diagram showing a workflow for acquiring knowledge from unstructured data that is a sound source in a knowledge acquisition system according to an embodiment of the present invention.

Hereinafter, a knowledge acquisition system according to embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to the following embodiments, The present invention may be embodied in various other forms without departing from the spirit of technical thought. That is, it is to be understood that the specific structural or functional descriptions are merely illustrative of the embodiments of the present invention, and that the embodiments of the present invention may be embodied in various forms and are construed as being limited to the embodiments described herein No. It is to be understood that the invention is intended to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention, as defined by the following claims.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In the present application, the terms "comprises ", or" comprising ", etc. are intended to specify the presence of stated features, integers, steps, operations, elements, or combinations thereof, But do not preclude the presence or addition of steps, operations, elements, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the meaning in the context of the relevant art and are not to be construed as ideal or overly formal in meaning unless expressly defined in the present application .

Unless otherwise specified in the specification of the present invention, a manager is a person who, through a knowledge acquisition system, constructs an initial knowledge model and learning models, verifies the mapped knowledge, May refer to both administrators and users who have access and use rights to the knowledge acquisition system.

In the specification of the present invention, the term "knowledge data" refers to entities required to implement "knowledge" or "knowledge" accumulated in a knowledge base which is a component of an expert system, it means. The term "knowledge" means that a plurality of unspecified information can be selected and combined to be used as a solution suitable for the time and space of a given problem. "Knowledge acquisition" means mapping "knowledge data" from unstructured data.

In the specification of the present invention, for example, a case having a relatively large analysis function, for example, a case having an analysis function for a specific data type is referred to as an "analysis module", a case having a relatively small analysis function, For example, a case where the detailed analysis function for the specific data type is called a "sub-analysis module", and the "sub-analysis module" also functions as a separate analysis module. "And" sub-analysis module "may all be referred to as " analysis module ".

1 is a schematic diagram showing a configuration of a knowledge acquisition system according to an embodiment of the present invention.

1, the knowledge acquisition system 10 includes a data collection unit 1000, a data learning / analysis unit 2000, a knowledge verification unit 3000, and a model management unit 4000.

The data collecting unit 1000 can collect the unstructured data 30 through the network 10. [

The network 10 may be a wired Internet service, a local area network (LAN), a wide area network (WAN), an intranet, a wireless Internet service, a mobile computing service, a wireless data communication service, a wireless Internet access service, And can transmit / receive data through wire / wireless. When the network 10 is connected to a smart phone or a tablet, the network 10 may be a wireless data communication service such as 3G, long term evolution (LTE), wireless LAN such as Wi-Fi, Bluetooth,

Un-structured data (30) refers to unstructured data that is different in form and structure, such as image, sound source, image, and text (text), unlike numeric data having a certain standard or form. The unstructured data 30 may be, for example, data generated on-line with mobile devices, such as e-mail, Twitter, or blog, as well as traditional data such as books, magazines, medical records, voice information and video information.

In particular, the unstructured data 30 refers to data that is irregular in type from the viewpoint of information and is ambiguous to grasp the meaning, and thus it is difficult to apply the conventional computer processing method. The unstructured data 30 may include semi-structured data.

The data learning / analyzing unit 2000 can analyze the irregular data 30 collected by the data collecting unit 1000, map knowledge data, and learn from analysis results. The data learning / analysis unit 200 may include an analysis management unit 2100 and a data processing unit 2200.

The analysis management unit 2100 can manage the process of mapping the knowledge data and learning from the analysis result. The analysis management unit 2100 can analyze the types of data on the collected irregular data 30 to select an analysis module, configure a work flow, which is a procedure sequence of the analysis modules, and select a plurality of analysis modules And provide the selected analysis module to the data processing unit 2200 in the form of a plug-in.

The analysis management unit 2100 may include an analysis process management unit 2120 and an analysis module management unit 2140. The analysis processor management unit 2120 can analyze the types of data on the collected irregular data 30 to select the analysis module, and configure the workflow, which is a procedure sequence of the analysis modules. The analysis module management unit 2140 manages a plurality of analysis modules to be selected and can provide the analysis modules selected for use in the data processing unit 2200 in a plug-in form. The analysis module management unit 2140 may directly store and manage a plurality of analysis modules, or may call and provide analysis modules held by one or more external systems.

The data processing unit 2200 may analyze the irregular data 30 collected by the data collecting unit 1000 based on the workflow configured with the analysis module selected by the analysis processor managing unit 2120 to map the knowledge data. The data processing unit 2200 can generate and update a learning model based on the knowledge model generated by the model management unit 4000 after the knowledge verification unit 3000 verifies the mapped knowledge data.

The data processing unit 2200 may include a knowledge learning unit 2220 and a knowledge analysis unit 2240. The knowledge analysis unit 2240 analyzes the irregular data 30 collected by the data collection unit 1000 based on the workflow configured with the analysis module selected by the analysis processor management unit 2120 and maps the knowledge data. The knowledge analysis unit 2240 can map the knowledge data with reference to the knowledge model generated by the knowledge verification unit 3000. [ The knowledge learning unit 2220 can generate and update a learning model based on the knowledge model generated by the model management unit 4000 after the knowledge verification unit 3000 verifies the mapped knowledge data.

The knowledge verification unit (3000) can perform verification of the mapped knowledge data in the knowledge analysis unit (2240). The knowledge verification unit 3000 may include an administrator verification unit 3100 and an error automatic verification unit 3200.

The manager verification unit 3100 may provide a user interface for the manager to perform verification on the mapped knowledge data in the knowledge analysis unit 2240. [ The administrator 20 can perform verification of the mapped knowledge data, via the user interface provided by the administrator verification unit 3100, on-line or off-line. The error automatic verification unit 3200 can automatically verify whether there is an error in the knowledge data mapped by the knowledge analysis unit 2240. Specifically, the error automatic verification unit 3200 performs verification on the knowledge data that is pre-built or mapped through the administrator verification unit 3100, and then, based on the knowledge model generated in the model management unit 4000, The analyzer 2240 can automatically check whether or not the mapped knowledge data is erroneous. Verification in the administrator verification unit 3100 and verification in the error automatic verification unit 3200 will be described in detail with reference to FIG. 2 and FIG.

The model management unit 4000 can manage the learning model and the knowledge model, which are necessary for mapping knowledge data and learning from analysis results. The model management unit 4000 may include a learning model management unit 4100 and a knowledge model management unit 4200. The model management unit 4000 may further include a database or may include a database in the learning model management unit 4100 and the knowledge model management unit 4200, respectively. The database may be a space capable of storing data in any form such as NoSQL, relational database, file system, or the like. The database may be logically divided into one storage unit, or may be a division unit for logically dividing one or a plurality of storage units, or a storage unit physically divided or a logically divided division unit.

The learning model management unit 4100 can create and update a learning model, and store and manage the learning model. The learning model is a machine learning result of an analysis module such as a language analyzer used in the knowledge learning unit 2220. In the knowledge analysis unit 2240, the knowledge data can be mapped by analyzing the irregular data 30 based on the learning model have. The knowledge model management unit 4200 can store the knowledge data that has been verified by the knowledge verification unit 3000 and generate a knowledge model. The knowledge model may be a semantic dictionary for analyzing modules such as a language analyzer used in the knowledge analyzer 2240 to give meaning to unstructured data. The knowledge model may be an entire R-DB (Relational Data Base) such as a knowledge base, or a table schema constituting a knowledge base. For example, if the knowledge model is a table schema that constitutes a knowledge base, any queries to the knowledge base may be inferred according to the knowledge model structure, or data may be searched to return data as answers.

The knowledge acquisition system 1 analyzes and processes the irregular data 30 collected using a learning model and a knowledge model that have been constructed or using a learning model and a knowledge model generated by the verification by the manager 20 The knowledge model and the learning model can evolve themselves by learning infinitely. This will be described in detail with reference to FIG. 2 and FIG.

FIG. 2 is a schematic diagram illustrating a configuration and operation of a knowledge acquisition system according to an embodiment of the present invention to explain a knowledge acquisition process for unstructured data selected as a test set.

Referring to FIG. 2, the data collection unit 1000 of the knowledge acquisition system 1 may select a part of the irregular data 30 collected through the network 10 as a test set. If there are no learning models and / or knowledge models created or built in the learning model management unit 4100 and / or the knowledge model management unit 4200, the collected irregular data 30 are all set as a test set Can be selected. Hereinafter, the operation of the knowledge acquisition system 1 using the unstructured data 30 selected as the test set will be described.

The type of data is analyzed in the analysis process management unit 2120 of the analysis management unit 2100 in the unstructured data 30 selected as the test set. For example, the type of data may be text, image, sound source, or image, but is not limited thereto.

The analysis process management unit 2120 may select an analysis module managed by the analysis module management unit 2140 according to the type of analyzed data, and may configure a workflow that is a process sequence of the selected analysis modules. The analysis processor management unit 2120 analyzes the irregular data 30 selected by the test set based on the workflow configured to the knowledge analysis unit 2240 so that the knowledge analysis unit 2240 can map the knowledge data have. At this time, the analysis module management unit 2140 can provide the selected analysis module to the knowledge analysis unit 2240 in a plug-in form.

The reason why the analysis module management unit 2140 provides the selected analysis module to the knowledge analysis unit 2240 in a plug-in form is that the analysis module selected in the knowledge analysis unit 2240 is combined into a plug- Which can function as a customized analyzer for the analyzer 30.

Therefore, the knowledge acquisition system 1 does not provide a separate analyzer according to the type of data of the collected unstructured data 30, but rather provides the knowledge analysis unit 2240 with an analysis module as a plug- . ≪ / RTI > Therefore, the knowledge analysis unit 2240 can simultaneously perform the function of the customized analyzer for each irregular data 30, in accordance with the analysis module that is plugged in as a single work frame for the general purpose analyzer.

For example, if the type of data of the collected irregular data 30 is text, the analysis process management unit 2120 selects analysis modules for analyzing the language, Configure your workflow. Thereafter, the knowledge analysis unit 2240 combines the analysis modules for analyzing the language provided in the plug-in form in the analysis module management unit 2140, and generates the unstructured data, which is text, according to the workflow configured in the analysis process management unit 2120 (30). ≪ / RTI >

The knowledge analysis unit 2240 refers to the knowledge model managed by the knowledge model management unit 4200 on the basis of the learning model managed by the learning model management unit 4100 and analyzes the unstructured data 30, Can be mapped. If the learning model and / or the knowledge model managed by the learning model managing unit 4100 and / or the knowledge model managing unit 4200 is not generated, the knowledge analyzing unit 2240 analyzes the irregular data 30 ) To map the knowledge data. Or knowledge model managed by the learning model management unit 4100 and / or the knowledge model management unit 4200 is relatively poor, the knowledge analysis unit 2240 analyzes the unstructured data 30 The knowledge data can be mapped to a level at which the accuracy is lowered.

Thereafter, the manager 20 performs verification of the knowledge data mapped by the knowledge analysis unit 2240 through the user interface provided by the administrator verification unit 3100 of the knowledge verification unit 3000, and if necessary, The knowledge data can be supplemented.

Knowledge model manager 4200 may generate a knowledge model from mapped knowledge data that has been verified and / or supplemented. The knowledge learning unit 2220 of the data processing unit 2200 can create or update the learning model based on the knowledge model generated by the knowledge model management unit 4200. [ The knowledge learning unit 2220 can use a learning algorithm such as CRF (Conditional Random Field), SVM (Support Vector Machine), or Patterns to create or update a learning model.

By repeating the above process, the knowledge model and the learning model are generated without analyzing the intervention of the manager 20 by using the knowledge model and the learning model generated through analysis of the irregular data 30 selected by a predetermined number or more of the test sets. It is possible to learn and evolve infinite self, which will be described in detail in FIG.

FIG. 3 is a schematic diagram illustrating a configuration and operation of a knowledge acquisition system according to an embodiment of the present invention for explaining a knowledge acquisition process for unstructured data selected as a learning set. In the description of FIG. 3, the same portions as those of FIG. 2 can be omitted.

3, the data collecting unit 1000 of the knowledge obtaining system 1 may select the remaining unregistered data 30 among the unstructured data 30 collected through the network 10 as a learning set. If the learning model and the knowledge model built in the learning model management unit 4100 and / or the knowledge model management unit 4200 are sufficiently evolved, all the collected irregular data 30 can be selected as the learning set. Hereinafter, the operation of the knowledge acquisition system 1 using the unstructured data 30 selected as the learning set will be described.

The type of data is analyzed in the analysis process management unit 2120 of the analysis management unit 2100 in the atypical data 30 selected as the learning set. The analysis process management unit 2120 may select an analysis module managed by the analysis module management unit 2140 according to the type of analyzed data, and may configure a workflow that is a process sequence of the selected analysis modules. The analysis processor management unit 2120 analyzes the irregular data 30 selected by the test set based on the workflow configured to the knowledge analysis unit 2240 so that the knowledge analysis unit 2240 can map the knowledge data have. At this time, the analysis module management unit 2140 can provide the selected analysis module to the knowledge analysis unit 2240 in a plug-in form.

The knowledge analysis unit 2240 refers to the knowledge model managed by the knowledge model management unit 4200 on the basis of the learning model managed by the learning model management unit 4100 and analyzes the unstructured data 30, Can be mapped.

Thereafter, the error automatic verification unit 3200 of the knowledge verification unit 3000 automatically verifies the mapped knowledge data in the knowledge analysis unit 2240, thereby automatically verifying whether or not an error has occurred. Although not shown, the error auto-verification unit 3200 can verify whether there is an error in the mapped knowledge data by referring to a previously constructed knowledge base.

The knowledge model management unit 4200 can create and update a knowledge model from the verified mapped knowledge data. The knowledge learning unit 2220 of the data processing unit 2200 can update the learning model based on the knowledge model generated and updated by the knowledge model management unit 4200. [

If the error automatic verification unit 3200 determines that there is an error in the mapped knowledge data, the manager 20, through the user interface provided by the administrator verification unit 3100, Verification of the knowledge data can be performed, and the mapped knowledge data can be supplemented if necessary.

In FIGS. 2 and 3, after the analysis of the atypical data 30 selected by the test set is performed, the analysis is performed as to the atypical data 30 selected by the learning set. However, 1000 may be selected as a test set, and the remainder may be selected as a learning set, so that the process described in FIG. 2 and the process illustrated in FIG. 3 may be continuously performed together.

4 is a schematic diagram showing a main part of an analysis process management unit of a knowledge acquisition system according to an embodiment of the present invention. In the description of FIG. 4, the contents overlapping with those described in FIGS. 1 to 3 may be omitted.

4, the analysis process management unit 2120 may include a data type recognition unit 2122, an analysis module selection unit 2124, a workflow configuration unit 2126, and a workflow execution unit 2128.

The data type recognizing unit 2122 can recognize the type of data of the irregular data collected by the data collecting unit 1000. For example, the type of data may be text, image, sound source, or image, but is not limited thereto.

The analysis module selection unit 2124 can select an analysis module managed by the analysis module management unit 2140 according to the type of analyzed data. The analysis module selection unit 2124 can select the analysis module in consideration of the intention of the analysis set by the administrator or the user of the knowledge acquisition system 1. [

The workflow configuration unit 2126 may configure a workflow that is a progression order of the selected analysis modules. The proceeding order of the selected analysis modules constituting the workflow can be determined in consideration of the analytical intent set by the manager or the user of the knowledge acquisition system 1. [ The user's intended analytical intent can be defined in standard formats such as XML (eXtensible Markup Language), JSON (JavaScript Object Notation), and the like.

The workflow execution unit 2128 analyzes the unstructured data to map the knowledge data to the knowledge analysis unit 2240 based on the configured workflow. At this time, the analysis module management unit 2140 may provide the selected analysis module to the knowledge analysis unit 2240 in a plug-in form. The unstructured data collected by the data collecting unit 1000 may be provided to the knowledge analyzing unit 2240 through the workflow executing unit 2128 or may be directly provided to the knowledge analyzing unit 2240 by the data collecting unit 1000 .

The analysis module management unit 2140 may provide all or some of the selected analysis modules to the knowledge learning unit 2220 in a plug-in form. The knowledge learning unit 2220 can generate and update a learning model based on the knowledge model by referring to the analysis module selected by the analysis module selection unit 2124 and the workflow configured in the workflow organization unit 2126. [

For example, the knowledge learning unit 2220 uses a part of the analysis modules selected for the unstructured data in the analysis module selection unit 2124 and the workflow configuration unit 2126, The learning model can be created and updated based on the knowledge model. Or knowledge learning unit 2220 selects an analysis module for atypical data in the analysis module selection unit 2124 and the workflow configuration unit 2126 and selects an analysis module selected separately for generation of a learning model The learning model can be generated and updated based on the knowledge model for the corresponding unstructured data.

5 is a schematic diagram showing a main part of an analysis module management unit of a knowledge acquisition system according to an embodiment of the present invention.

5, the main module managing unit 2140 of the analysis managing unit 2100 includes an analyzing module recognizing unit 2142 and an analyzing module storing / calling unit 2144.

The analysis module recognition unit 2142 selects an analysis module in the analysis module storage / call unit 2144 and requests the knowledge learning unit 2220 and / or the knowledge analysis unit 2240, according to a request from the analysis process management unit 2120. [ And provides the selected analysis module in a plug-in form.

The analysis module storing / calling unit 2144 may store a plurality of analysis modules or may call an analysis module in an external system. The analysis module that can be stored or called by the analysis module storing / calling unit 2144 may be a language analysis module (LAM), an image analysis module (MAM), a sound source analysis module (SAM) However, the present invention is not limited thereto, and various analysis modules can be stored or called according to the type of data of the unstructured data to be analyzed. For example, a language analysis module (LAM), an image analysis module (MAM), a sound source analysis module (SAM), or an image analysis module (IAM), which can be stored or called by the analysis module storage / May each be comprised of one or a plurality of sub-analysis modules.

The analysis process management unit 2120 selects the entire language analysis module (LAM), the image analysis module (MAM), the sound source analysis module (SAM), or the image analysis module (IAM) (MAM), a sound source analysis module (SAM), and an image analysis module (IAM), and configure one or more sub analysis modules constituting them and configure their workflows.

For example, the workflow constituted by the analysis process management unit 2120 may be a language analysis module (LAM) for analyzing text or a sub analysis module constituting a language analysis module (LAM) Or a sub-analysis module constituting a language analysis module (LAM) and sub-analysis modules constituting an image analysis module (MAM) or an image analysis module (MAM). An exemplary embodiment of this will be described in detail with reference to FIGS. 7 to 9. FIG.

6 is a conceptual diagram for explaining a function of a workflow execution unit of a knowledge acquisition system according to an embodiment of the present invention.

Referring to FIG. 6, the analysis process management unit 2120 can monitor the operation state of the knowledge analysis unit 2240 through the workflow execution unit 2128. When there is an analysis module in which an error occurs in the analysis process in the knowledge analysis unit 2240 through the analysis module provided in the form of a plug-in, the workflow execution unit 2128 analyzes the knowledge analysis unit 2240 through the configured workflow, And can automatically notify the manager 20 of the notification.

When the analysis modules specific to the knowledge analysis module 2240 are provided in a plug-in form and are combined and operate according to the workflow, the knowledge analysis module 2240 can function as one specific analyzer. Therefore, although an error may occur in the individual analysis module itself, it may be an error of the individual analysis module itself. However, if the knowledge analysis module 2240 determines that a particular analysis module is provided in a plug- When functioning as an analyzer, that is, there is no error in the individual analysis module itself, but it may be an error due to the workflow configuration. Accordingly, the manager 20 can select an analysis module in the analysis process management unit 2120 and confirm the operation for configuring the workflow, as well as confirm the individual analysis modules in which an error occurs. Or if the type of data is erroneously recognized, an error may occur in the selected analysis module, so that the manager 20 can also confirm the type.

Through this process, the knowledge acquisition system 1 according to the present invention shown in FIG. 1 can select analytic modules that operate independently, combine and arrange the analytic modules and configure the workflow, The same effect as having a specific analyzer having individual functions close to infinite can be obtained.

7 is a schematic diagram showing a workflow for acquiring knowledge from unstructured data that is text in a knowledge acquisition system according to an embodiment of the present invention;

Referring to FIG. 7, the analysis modules (LAM1, LAM2, and LAM3) selected when the type of data of the irregular data is text, and the language analysis flow (LAM-F), which is a workflow including them. The language analysis flow (LAM-F) includes, for example, a morphological analyzer (MA), a Named Entity Recognizer (LAM2), a dependency analyzer Analyzer (DP: Dependency Parser, LAM3). The language analysis flow (LAM-F) analyzes the morpheme of the text in the morpheme analyzer (LAM1), passes the result to the object name recognizer (LAM2), and transmits the object name recognition result to the dependency analyzer (LAM3) . Therefore, the knowledge analysis unit 2240 shown in FIGS. 1 to 3 analyzes the morpheme in the text, which is the unstructured data, and recognizes the object name in the morpheme, after the language analysis flow (LAM-F) You can perform dependency analysis to analyze relationships. Dependency analysis can analyze relationships and relationships between subject, object, modifier, and so on.

8 is a schematic diagram illustrating a workflow for acquiring knowledge from unstructured data that is an image in a knowledge acquisition system according to an embodiment of the present invention.

Referring to FIG. 8, there is shown an analysis module (IAM1, IAM2, IAM3) selected when the type of data of the irregular data is an image and an image analysis flow (IAM-F) which is a workflow including them. The image analysis flow (IAM-F) includes, for example, an image frequency analyzer IAM1, an edge analyzer IAM2, an image object recognizer IAM3, an object analyzer IAM1, Analyzer (IAM4). The image analysis flow (IAM-F) analyzes the frequency of the image in the image frequency analyzer (IAM1), passes the result to the edge analyzer (IAM2), analyzes the edge included in the image, (IAM3), recognizes the object included in the image, and transmits the result of the recognized object to the object / character analyzer IAM4 to acquire information about the object included in the image, that is, You can configure a workflow to get information. In addition, although not shown, an analysis module for analyzing the dependency of objects and characters may be further selected, and a workflow including the selected analysis modules may be configured.

When the image object recognizer IAM3 obtains a result that there is a character in the object included in the image, it additionally delivers the result to the character recognizer IAM5 and converts it to a language analysis flow LAM-F ) Can be configured.

Therefore, the knowledge analysis unit 2240 shown in FIGS. 1 to 3 can combine the image analysis flow (IAM-F) to recognize / analyze the object in the image, which is the unstructured data, have.

9 is a schematic diagram showing a workflow for acquiring knowledge from unstructured data that is a sound source in a knowledge acquisition system according to an embodiment of the present invention.

9, the analysis modules SAM1, SAM2, and SAM3 selected when the type of the data of the atypical data is a sound source, and the sound source analysis flow SAM-F, which is a work flow including the analysis modules SAM1, SAM2, and SAM3. The sound source analysis flow (SAM-F) may have a sound source frequency analyzer (SAM1), sound source information detector (SAM2), and voice recognizer (IAM3), for example, as selected analysis modules (SAM1, SAM2, SAM3). After analyzing the frequency of the sound source in the sound source frequency analyzer (SAM1), the sound source analyzer (SAM2) analyzes the sound source information and transmits the result to the sound source information detector (SAM2).

If the sound source frequency analyzer (SAM1) analyzes the frequency of a sound source and analyzes the sound source such as a voice having a language, the result is transmitted to the voice recognizer (SAM3) to recognize the sound included in the sound source. The workflow for delivering the result to a language analysis flow (LAM-F) similar to or similar to that shown in Fig. 7 can be constructed.

For example, when the analysis module such as the sound source information searcher (SAM2) is implemented in an external system having a separate database, the analysis module store / call unit 2144 shown in FIG. 5 may include a sound source information searcher SAM2).

Therefore, the knowledge analysis unit 2240 shown in FIGS. 1 to 3 may combine the sound source analysis flow (SAM-F) to identify / search sound source information such as music in a sound source which is an unstructured data, Language analysis can be performed.

Although not shown, when the data type of the irregular data is a video, the workflow can be configured by combining the analysis modules shown in FIG. 7 to FIG. 9 or by selecting a few additional analysis modules for analyzing an image.

Further, the embodiments of the present invention can be made into a program executable in a computer system. In addition, the program read from the computer-readable recording medium containing the program can be executed in the digital computer system.

Examples of the computer-readable recording medium include a ROM, a RAM, a CD-ROM, a DVD-ROM, a magnetic tape, a floppy disk, an optical data storage device and the like. Further, a carrier wave (for example, And the like. The computer readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner. And functional programs, codes, and code segments for implementing the present invention can be easily inferred by programmers skilled in the art to which the present invention pertains.

2120: an analysis process management unit, 2122: a data type recognition unit, 217: a data acquisition unit, 217: a data acquisition unit, The analysis module selection unit 2124 includes a workflow configuration unit 2128 a workflow execution unit 2140 an analysis module management unit 2142 an analysis module recognition unit 2144 an analysis module storage and call unit 2200 a data processing unit 2220 : Knowledge Learning Section, 2240: Knowledge Analysis Section, 3000: Knowledge Verification Section, 3100: Manager Verification Section, 3200: Error Auto Verification Section, 4000: Model Management Section, 4100: Learning Model Management Section, 4200: Knowledge Model Management Section

Claims (10)

A data collector for collecting unstructured data through a network;
An analysis processor managing unit for analyzing types of data of irregular data collected by the data collecting unit to select an analysis module and constructing a work flow that is a progress order of the selected analysis modules;
A knowledge analysis unit for analyzing atypical data collected by the data collection unit and mapping knowledge data based on a workflow configured with the analysis module selected by the analysis processor management unit;
A knowledge verification unit that verifies the knowledge data mapped by the knowledge analysis unit; And
And a knowledge model management unit for storing the knowledge data, which is verified by the knowledge verification unit, to generate a knowledge model,
Wherein the knowledge analysis unit maps the knowledge data with reference to the knowledge model generated by the knowledge verification unit. The method according to claim 1,
Wherein the knowledge verification unit comprises:
An administrator verification unit for providing a user interface so that an administrator can perform verification on the knowledge data mapped by the knowledge analysis unit; And
And an error automatic verification unit for automatically verifying whether or not an error has occurred in the knowledge data mapped by the knowledge analysis unit based on the knowledge model. 3. The method of claim 2,
Wherein the data collection unit selects one of the collected atypical data as a test set and the other as a learning set,
Wherein the knowledge verification unit analyzes unstructured data selected by the test set and performs verification of the mapped knowledge data in an administrator verification unit, analyzes unstructured data selected by the learning set, And the verification is performed by the error automatic verification unit. The method according to claim 1,
A knowledge learning unit that generates and updates a learning model based on the knowledge model; And
And a learning model manager for storing the generated and updated learning models,
Wherein the knowledge analysis unit maps the knowledge data based on the learning model stored in the learning model management unit. 5. The method of claim 4,
And an analysis module management unit for managing the plurality of analysis modules,
Wherein the analysis module management unit provides the analysis module used for analyzing the knowledge model for creating and updating the learning model among the plurality of analysis modules to the knowledge learning unit in the form of a plug- Wherein the knowledge acquisition system comprises: The method according to claim 1,
And an analysis module management unit for managing the plurality of analysis modules to be selected by the analysis processor management unit,
Wherein the analysis module management unit provides the analysis module selected by the analysis processor management unit of the plurality of analysis modules to the knowledge analysis unit in a plug-in form so as to analyze the collected atypical data. The method according to claim 6,
The workflow execution unit,
Wherein when there is an analysis module in which an error occurs during the analysis process in the knowledge analysis module through the analysis module provided in a plug-in form, the configured workflow is stopped and a notification is sent to the manager Acquisition system. The method according to claim 6,
The type of data analyzed by the analysis processor management unit is text, image, sound source or image,
Wherein the plurality of analysis modules managed by the analysis module management unit are a language analysis module, an image analysis module, a sound source analysis module, or an image analysis module. 9. The method of claim 8,
Wherein the language analysis module, the image analysis module, the sound source analysis module, and the image analysis module each comprise one or a plurality of sub-analysis modules. 10. The method of claim 9,
Wherein the analysis processor management unit configures the workflow by selecting sub-analysis modules from one or more of the language analysis module, the image analysis module, the sound source analysis module, and the image analysis module Knowledge acquisition system.

Images