Data labeling method and device, storage medium, and terminal

Abstract

The present invention discloses a data annotation method and device, a storage medium, and a terminal, which relate to the field of data processing technology, and the main purpose is to solve the problem of low efficiency of existing data annotation. It includes: obtaining target data to be annotated; calling data annotation rules matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data; annotating the target data based on the data annotation rules to obtain the annotation results of the target data. It is mainly used for data annotation.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data labeling method and device, a storage medium, and a terminal. Background Art

[0002] With the advent of the big data era, data security issues have become the focus of attention of all parties. In order to determine whether the data is safe, the data will be labeled with security attributes according to specific conditions for security analysis.

[0003] At present, the existing security attribute labeling of data is usually based on the if-else method for matching and labeling. However, since the if-else method is used as the matching condition, the number of update logics is relatively large, which will lead to maintenance difficulties for a large number of complex if-else logics, resulting in high maintenance costs, low efficiency, and high error rates for matching conditions. For complex matching conditions, the code layer needs to be modified and added, resulting in poor scalability of the labeled matching conditions. At the same time, directly labeling data in an if-else manner will also cause a large consumption of system resources, which cannot meet the business needs for flexible data labeling, thereby affecting data labeling efficiency. Summary of the invention

[0004] In view of this, the present invention provides a data labeling method and device, a storage medium, and a terminal, the main purpose of which is to solve the problem of low efficiency of existing data labeling.

[0005] According to one aspect of the present invention, a data labeling method is provided, comprising:

[0006] Obtain target data to be labeled;

[0007] Retrieving data annotation rules matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data;

[0008] The target data is labeled based on the data labeling rule to obtain a labeling result of the target data.

[0009] Furthermore, the rule engine includes: a rule manager and a rule executor. Before the rule engine retrieves the data annotation rule matching the target data, the method further includes:

[0010] The rule manager in the rule engine loads the entered rule execution requirement information through the front-end interface. The rule execution requirement information is used to characterize the logical conditions when executing the annotation operation, as well as the rule operation mode and rule operation attributes;

[0011] Generating, in the rule manager, a data annotation rule that matches the rule execution requirement information;

[0012] The rule manager submits the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; and / or,

[0013] The rule manager in the rule engine receives the initial data annotation rule through the front-end interface, and processes the initial data annotation rule to obtain the data annotation rule;

[0014] The rule manager submits the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule.

[0015] Furthermore, generating a data annotation rule matching the rule execution requirement information in the rule manager includes:

[0016] The rule manager performs syntax analysis on the rule execution requirement information to obtain rule logic information;

[0017] The rule manager performs contextual logic identification on the rule logic information to generate data annotation rules.

[0018] Furthermore, after the rule manager submits the data annotation rule to the rule executor in the rule engine, the method further includes:

[0019] After receiving the data annotation rules submitted by the rule manager, the rule executor determines the data information that can be annotated by each data annotation rule, wherein the data information includes: a data identifier;

[0020] The data annotation rules and corresponding data identifiers are stored in a preset storage location in association with each other;

[0021] The step of retrieving a data annotation rule matching the target data through a rule engine includes:

[0022] When the rule executor receives the labeling task, it parses the labeling task to obtain the target data to be labeled;

[0023] Based on the data identifier of the target data to be labeled, the matching data labeling rules in the preset storage location are retrieved, and the labeling task and the corresponding data labeling rules are sent to the task cluster to perform data labeling through the task cluster.

[0024] Furthermore, the method further comprises:

[0025] After the target data to be labeled is acquired, a corresponding labeling task is created for the target data, and the labeling task is sent to a task manager in the task cluster, so that the task manager determines an execution node of the task to be labeled from the task cluster.

[0026] Furthermore, after sending the labeling task and the corresponding data labeling rule to the task cluster, the method further includes:

[0027] Determine an execution node in the task cluster that executes the labeling task of the target data;

[0028] The labeling task and the corresponding data labeling rule are sent to the execution node through the rule executor in the rule engine, so that the target data corresponding to the labeling task is labeled by the execution node based on the data labeling rule.

[0029] Furthermore, the determining of the execution node in the task cluster that executes the target data labeling task includes:

[0030] Acquire the rule execution configuration information of the rule execution requirement information from the rule manager, and / or acquire the data attribute of the target data from the distributed data source;

[0031] The rule executor determines an execution node that executes the labeling task corresponding to the rule execution configuration information and / or the data attribute.

[0032] Further, the determining, by the rule executor, an execution node that executes the labeling task corresponding to the rule execution configuration information includes:

[0033] Obtaining node resource information of execution nodes in the task cluster, wherein the node resource information includes at least one of node resource occupancy, node resource execution language, and node resource execution environment;

[0034] Based on the node resource occupancy, node resource execution language, and node resource execution environment in the node resource information and matching with the rule execution configuration information, the execution node for executing the labeling task is determined. The rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task.

[0035] Further, the determining, by the rule executor, an execution node corresponding to the data attribute to execute the labeling task includes:

[0036] Acquire source data of a to-be-executed labeling task and node resource information of at least one execution node from a distributed data source;

[0037] Parsing data attributes of the source data, wherein the data attributes include field content and business data source;

[0038] Execution nodes corresponding to the node resource information matching the data attributes are allocated to different data in the source data, so as to determine execution nodes for executing annotations of different target data.

[0039] Furthermore, the determining of the execution node in the task cluster that executes the target data labeling task includes:

[0040] If it is determined based on the rule execution configuration information that there are multiple execution nodes for executing the labeling task, the execution node for executing the labeling task is determined based on matching the data attributes of the target data with the node resource information of the execution node.

[0041] According to another aspect of the present invention, a data labeling device is provided, comprising:

[0042] An acquisition module is used to acquire target data to be labeled;

[0043] A calling module, used to call a data annotation rule matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data;

[0044] The labeling module is used to label the target data based on the data labeling rules to obtain the labeling results of the target data.

[0045] Furthermore, the rule engine includes: a rule manager and a rule executor, and the device also includes:

[0046] A loading module is used for the rule manager in the rule engine to load the entered rule execution requirement information through the front-end interface. The rule execution requirement information is used to characterize the logical conditions when performing the annotation operation, as well as the rule operation mode and rule operation attributes;

[0047] A generation module, used to generate a data annotation rule matching the rule execution requirement information in the rule manager;

[0048] A first submission module is used for the rule manager to submit the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; and / or,

[0049] A receiving module, used for a rule manager in a rule engine to receive an initial data annotation rule through a front-end interface, and to process the initial data annotation rule to obtain the data annotation rule;

[0050] The second submission module is used for the rule manager to submit the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule.

[0051] Furthermore, the generating module comprises:

[0052] A parsing unit, used for the rule manager to perform syntax parsing on the rule execution requirement information to obtain rule logic information;

[0053] The identification unit is used for the rule manager to perform contextual logic identification on the rule logic information and generate data annotation rules.

[0054] Furthermore, the device also includes:

[0055] A determination module, configured to determine, after the rule executor receives the data annotation rule submitted by the rule manager, data information that can be annotated by each data annotation rule, wherein the data information includes: a data identifier;

[0056] A storage module, used to associate and store the data labeling rules and corresponding data identifiers in a preset storage location;

[0057] The calling module includes:

[0058] A parsing unit, configured to parse the labeling task when the rule executor receives the labeling task, and obtain target data to be labeled;

[0059] A retrieving unit is used to retrieve the matching data labeling rules in the preset storage location based on the data identifier of the target data to be labeled, and send the labeling task and the corresponding data labeling rules to the task cluster to perform data labeling through the task cluster.

[0060] Furthermore, the device also includes:

[0061] A creation module is used to create a corresponding labeling task for the target data after obtaining the target data to be labeled, and send the labeling task to the task manager in the task cluster so that the task manager can determine the execution node of the task to be labeled from the task cluster.

[0062] Furthermore, the device further comprises: a sending module,

[0063] The determination module is further used to determine an execution node in the task cluster that executes the labeling task of the target data;

[0064] The sending module is also used to send the labeling task and the corresponding data labeling rule to the execution node through the rule executor in the rule engine, so that the target data corresponding to the labeling task can be labeled based on the data labeling rule by the execution node.

[0065] Furthermore, the determining module includes:

[0066] an acquisition unit, configured to acquire rule execution configuration information of the rule execution requirement information from the rule manager, and / or acquire data attributes of the target data from a distributed data source;

[0067] A determination unit is used to determine, through the rule executor, an execution node that executes the labeling task corresponding to the rule execution configuration information and / or the data attribute.

[0068] Furthermore, the determining unit includes:

[0069] A first acquisition subunit is used to acquire node resource information of execution nodes in the task cluster, wherein the node resource information includes at least one of node resource occupancy, node resource execution language, and node resource execution environment;

[0070] A determination subunit is used to determine the execution node for executing the labeling task by matching the node resource occupancy, node resource execution language, and node resource execution environment in the node resource information with the rule execution configuration information, wherein the rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task.

[0071] Furthermore, the determining unit further includes:

[0072] A second acquisition subunit is used to acquire source data of the to-be-executed labeling task and node resource information of at least one execution node from a distributed data source;

[0073] A parsing subunit, used to parse data attributes of the source data, wherein the data attributes include field content and business data source;

[0074] The allocation subunit is used to allocate execution nodes corresponding to the node resource information matching the data attributes to different data in the source data, so as to determine the execution nodes for executing the annotations of different target data.

[0075] Furthermore, the determination unit is specifically used to determine the execution node for executing the labeling task by matching the data attributes of the target data with the node resource information of the execution node if it is determined based on the rule execution configuration information that there are multiple execution nodes for executing the labeling task.

[0076] According to another aspect of the present invention, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned data labeling method.

[0077] According to another aspect of the present invention, there is provided a terminal, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0078] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned data labeling method.

[0079] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0080] The present invention provides a data labeling method and device, a storage medium, and a terminal. Compared with the prior art, the embodiments of the present invention obtain target data to be labeled; retrieve data labeling rules matching the target data through a rule engine, wherein the rule engine stores data labeling rules matching different data; label the target data based on the data labeling rules to obtain labeling results of the target data, thereby realizing automatic matching of data identification rules based on the rule engine to simplify the generation steps of complex labeling rules, so that the labeling matching conditions can be highly expanded, and at the same time, rule matching based on the rule engine meets the business needs for flexible data labeling, realizes convenient processing of data labeling, and thus improves the efficiency of data labeling.

[0081] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0083] Figure 1 A flow chart of a first data labeling method provided by an embodiment of the present invention is shown;

[0084] Figure 2 A schematic diagram of a rule engine processing flow provided by an embodiment of the present invention is shown;

[0085] Figure 3 A structural diagram of a data annotation execution system provided by an embodiment of the present invention is shown;

[0086] Figure 4 A flow chart of a second data labeling method provided by an embodiment of the present invention is shown;

[0087] Figure 5 A schematic diagram of the structure of a rule engine provided by an embodiment of the present invention is shown;

[0088] Figure 6 A flow chart of a third data labeling method provided by an embodiment of the present invention is shown;

[0089] Figure 7 A flow chart of a fourth data labeling method provided by an embodiment of the present invention is shown;

[0090] Figure 8 A fifth data labeling method flow chart provided by an embodiment of the present invention is shown;

[0091] Fig. 9 A flow chart of a sixth data labeling method provided by an embodiment of the present invention is shown;

[0092] Fig.10 A block diagram of a data labeling device provided by an embodiment of the present invention is shown;

[0093] Fig.11 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0094] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0095] Security attribute labeling for data is usually based on the if-else method for matching and labeling. However, since the if-else method is used as the matching condition, the number of update logics is relatively large, which will lead to a large number of complex if-else logics that are difficult to maintain, resulting in high maintenance costs, low efficiency, and high error rates for matching conditions. For complex matching conditions, the code layer needs to be modified and added, resulting in poor scalability of matching conditions. At the same time, simply labeling according to the set rules cannot meet the business needs of labeling the security attributes of data, thereby affecting the efficiency of data labeling based on security attributes. An embodiment of the present invention provides a data labeling method, such as Figure 1As shown, the method includes:

[0096] 101. Obtain target data to be labeled.

[0097] Among them, the target data to be labeled is the data content that needs to perform the labeling task, and the labels generated by the labeling task can be used to mark the security attributes of the data, including but not limited to network behavior data generated by users, system data generated by devices, etc., which are not limited in the embodiments of the present invention.

[0098] It should be noted that the target data to be labeled can be obtained by calling the target data through the execution node scheduling processor (such as the Yarn cluster), or it can be directly extracted from a preset database. The embodiment of the present invention does not make specific limitations, and the labeling task is executed by calling the data labeling rules in the rule engine.

[0099] 102. Retrieve data annotation rules matching the target data through a rule engine.

[0100] In the embodiment of the present invention, Figure 2 As shown, the rule engine stores data annotation rules that match different data. After obtaining the target data to be annotated, the data annotation rules that match the target data in the rule engine are retrieved to annotate the target data according to the data annotation rules. The data annotation rules are the rule content that matches the data annotation task, including but not limited to different target data can be annotated according to the same data annotation rules or different data annotation rules. When configuring the data annotation rules, the annotation rule content can be determined based on the specific judgment conditions of the data security attributes. For example, if the target data is all user IPs, the data annotation rule can be the user IPs in the black IP list, so that the data can be annotated according to this conditional content.

[0101] It should be noted that since the rule engine stores data annotation rules that match different data, when it is retrieved, the target data to be annotated is matched with each data annotation rule to determine the data annotation rule for annotating the target data. At the same time, the stored different data annotation rules are pre-generated, which speeds up the execution efficiency of data annotation, simplifies the configuration and operation difficulty of data annotation rules, and improves the reusability of data annotation rules.

[0102] 103. Label the target data based on the data labeling rule to obtain a labeling result of the target data.

[0103] In an embodiment of the present invention, after the data annotation rules are retrieved from the rule engine, the target data is annotated based on the matched data annotation rules, thereby completing the annotation operation and obtaining the annotation results of the target data. Among them, the form of the annotation results can be annotating the target data with different symbols and marks, or it can include screening the target data that meets the conditions of the data annotation rules and marking them as meeting the data annotation rules, thereby completing the output or storage of the annotation of the target data, which is not specifically limited in the embodiment of the present invention. In addition, the annotation of the target data can be performed directly based on the current rule engine, or based on the execution node scheduling processor that has a data transmission relationship with the rule engine, such as sending the matched data annotation rules to the execution node scheduling processor for execution, which is not specifically limited in the embodiment of the present invention.

[0104] Furthermore, the rule engine in the embodiment of the present invention includes: a rule manager and a rule executor, such as Figure 3 As shown, the rule manager and the rule executor are connected through a data interface. The rule manager manages the data labeling rules that match different data and submits them to the rule executor for storage through the data interface. The rule executor is used to match the data labeling rules corresponding to the target data from the database storing multiple data labeling rules after determining the target data, and then send them to the task cluster for executing the labeling task, namely the Yarn cluster, so as to determine the execution nodes that can execute each labeling task and execute the labeling task.

[0105] In one embodiment of the present invention, in order to further define and illustrate, as Figure 4 As shown, before step 102 uses a rule engine to call a data annotation rule that matches the target data, the method further includes:

[0106] 201a. The rule manager in the rule engine loads the entered rule execution requirement information through the front-end interface;

[0107] 202a. Generate, in the rule manager, a data annotation rule that matches the rule execution requirement information;

[0108] 203a. The rule manager submits the data annotation rule to the rule executor in the rule engine.

[0109] In order to label data according to the user's security requirements, realize the editability of data labeling, and simplify the tediousness of updating and executing labeling, the rule engine provides users with a front-end interface so that the rule manager in the rule engine can interact with the front-end interface (such as the front-end Web interface) to load the rule execution requirement information entered by the user, so that the rule manager can generate data labeling rules that match the rule execution requirement information. Among them, the rule execution requirement information is used to characterize the logical conditions when performing the annotation operation, as well as the rule operation mode and rule operation attributes. Specifically, the logical conditions are used to characterize the annotation conditions when performing the annotation operation, that is, including attribute objects, logical comparison operators, conditional comparison values, etc., so as to establish annotation rules based on logical conditions. For example, if Xip is not 12432423, then the attribute object is Xip, the logical comparison operator is the inequality symbol ≠, and the conditional comparison value is 12432423, so that when annotating, annotation is performed according to this rule; the rule operation mode is a rule mode for executing annotations for different logical conditions, for example, including but not limited to execution priority, whether to jump out of the rule sequence, the format of the annotation result, etc.; the rule operation attributes are the content that defines the annotation rule mode, including but not limited to data source, annotation tag name, rule description, etc., so as to determine a rule mode in which the user expects to annotate the data, so as to generate annotation execution instructions, store them, and realize fast reading, etc., which is not specifically limited in the embodiments of the present invention.

[0110] It should be noted that when the rule manager is loaded with the above-mentioned execution requirement information entered by the user based on the front-end interface, it can directly generate matching data annotation rules, so that after the rule manager generates the data annotation rules, it submits the data annotation rules to the rule executor in the rule engine, and the rule executor can perform data annotation or matching of data annotation rules based on the data annotation rules. Figure 3 As shown, the rule executor with controller function can also monitor, feedback, execute logic, check rules and other operations on the data labeling rules submitted by the rule manager, so as to ensure that the generated data labeling rules correctly label the data.

[0111] In addition, in an application scenario of the embodiment of the present invention, Figure 5The schematic diagram of the rule engine structure shown in the figure provides the user with the content of defining rules by configuring the UI interface of the front end. The user can generate matching data annotation rules by custom editing the rule execution requirement information. Among them, the user obtains the rule content through custom editing, including attributes, comparison operators, values, data sources, tag names, rule descriptions, priorities, jump rules, result formats and other contents. The rule manager generates data annotation rules by loading the rule content edited by the user, performing syntax parsing and context recognition. After the data annotation rules are generated by the rule manager, they are sent to the rule executor of the rule engine to determine the storage of the data annotation engine. At the same time, the rule executor in the rule engine can perform rule verification, rule matching, logic execution, result output and other operations on the generated data annotation rules to determine whether the rule content edited by the user matches the generated data annotation rules, that is, whether the annotation operation can be performed according to the rule content, thereby outputting. Among them, the rule executor can be a Noah controller, etc., and the embodiment of the present invention does not make specific limitations. Specifically, after the user edits the rule content based on the front-end, the rule manager can record the rule content entered by the user according to the editing id, description information and other contents, for example, "id":"P0f1402d834014c76af3affe9c8afae5a", "desc":"description information", and store the comparison conditions in the description information (such as "condition":"or", indicating the comparison relationship between the objects in the filter), rule types (such as "type":1), nested structures (such as "filterChildrenList":[]), left comparison attributes (such as "attributeLeft":"dev_id"), right comparison attributes (such as "attributeRight":"ip_addr"), operation functions (such as "operator":"equalIgnoreCase"), operation types (such as "operatorAttType":"string"), etc., so that when generating data annotation rules, the corresponding description content can be retrieved for syntax parsing and context recognition, which is not specifically limited in the embodiments of the present invention.

[0112] In one embodiment of the present invention, for further limitation and explanation, step 202a generates in the rule manager a data annotation rule that matches the rule execution requirement information, including: the rule manager performs grammatical parsing on the rule execution requirement information to obtain rule logic information; the rule manager performs contextual logic identification on the rule logic information to generate a data annotation rule.

[0113] Specifically, in order to automatically generate data annotation rules in the rule manager, meet the annotation requirements of different users, and realize flexible configuration of rule content, after loading the above-mentioned execution requirement information entered in the rule engine, the rule logic information is obtained through instruction syntax parsing. Since the syntax parsing at this time is only to translate the rule content configured by the user into a specific variant language expression, or dynamically generate the code of the calculation logic, that is, to generate the rule logic information of the data annotation rule, it can be implemented based on the CGLib library of the Java language, so what is obtained is the logic of the annotation rule. The data annotation task requires the introduction of external data sets. Therefore, context logic recognition is performed on the basis of the rule logic information, so as to generate data annotation rules that accurately annotate the data according to the conditional content. For example, a certain IP is not in the black IP list as a annotation rule. On the basis of the generated rule logic information, the "black IP list" is identified as the context of the condition of this annotation rule, so that the data annotation rule can be accurately operated.

[0114] For further definition and explanation, in an embodiment of the present invention parallel to 201-203, as Figure 6 As shown, before step 102 uses a rule engine to call a data annotation rule that matches the target data, the method further includes:

[0115] 201b. The rule manager in the rule engine receives the initial data labeling rule through the front-end interface, and processes the initial data labeling rule to obtain the data labeling rule;

[0116] 202b. The rule manager submits the data annotation rule to the rule executor in the rule engine.

[0117] Specifically, in order to meet the user's rule configuration requirements, the rule manager can directly receive the initial data annotation rules entered by the user through the front-end interface, that is, process it as an initial condition with logical content, so as to obtain a data annotation rule that can annotate the target data. Among them, the initial data annotation rule includes but is not limited to the content of the logical code deleted or omitted in the complete data annotation rule, and the text content containing the complete data annotation rule logic, so that the initial data annotation rule is processed in the rule manager to obtain a data annotation rule that can be read and executed in the rule executor. For example, if the user enters "the ip address belongs to whitelist 1" based on the front-end interface, the initial data annotation rule is processed, and the text "belongs to" is converted into a computer execution language, "=" in the Java language, so as to obtain a data annotation rule in an expression form that can be used for computer execution, so that the rule manager submits the data annotation rule to the rule executor to execute the data annotation, which is not specifically limited in the embodiment of the present invention.

[0118] In one embodiment of the present invention, in order to further define and illustrate, as Figure 7 As shown, after the rule manager submits the data annotation rule to the rule executor in the rule engine in step 203a or 202b, the method further includes:

[0119] 301. After receiving the data annotation rules submitted by the rule manager, the rule executor determines the data information that can be annotated by each data annotation rule;

[0120] 302. The data labeling rule and the corresponding data identifier are stored in a preset storage location in association with each other.

[0121] In an embodiment of the present invention, in order for the rule executor to match the data content to be annotated with the corresponding data annotation rule after the rule manager submits the data annotation rule, it is necessary to determine the data information that can be annotated by each data annotation rule. The data information includes a data identifier, that is, the data identifier can be used to characterize the unique identity of different data, so as to determine the data annotation rules that different data can match and execute. At the same time, in order to facilitate the accurate determination of the data annotation rule that matches the target data when matching the data annotation rule, the data annotation rule and the corresponding data identifier are associated and stored in a preset storage location. At this time, the associated storage can be achieved by binding or establishing a relationship between the data annotation rule and the data identifier, so as to be stored in the preset storage location. Among them, the preset storage location can be a cache area or a fixed storage area, so that when requesting to retrieve the data annotation rule that matches the target data, the data annotation rule corresponding to the data identifier of the target data is searched from the preset storage location, which greatly improves the rule matching efficiency of the annotation task execution.

[0122] Further, for the method implemented in steps 301-302, as Figure 8 As shown, step 201 retrieves data annotation rules matching the target data through a rule engine, including:

[0123] 401. When the rule executor receives a labeling task, it parses the labeling task to obtain target data to be labeled;

[0124] 402. Based on the data identifier of the target data to be labeled, retrieve the matching data labeling rule in the preset storage location, and send the labeling task and the corresponding data labeling rule to the task cluster.

[0125] During the labeling operation, the rule executor can receive the labeling task sent from the task cluster. The labeling task contains the target data to be labeled. At this time, the rule executor parses the labeling task to obtain the target data to be labeled, so as to match the data labeling rules corresponding to the target data. Among them, the target data to be labeled has a data identifier. After parsing the target data, the rule executor retrieves the data labeling rules in the preset storage location based on the data identifier of the target data. At this time, in order to enable the target data to be labeled in the execution node of the task cluster, the rule executor sends the matched data labeling rules and the labeling task to the task cluster together, so that the task cluster can perform data labeling based on each execution node, which greatly improves the execution efficiency of data labeling.

[0126] In one embodiment of the present invention, for further limitation and explanation, the method of the embodiment of the present invention also includes: after acquiring the target data to be labeled, creating a corresponding labeling task for the target data, and sending the labeling task to the task manager in the task cluster, so that the task manager determines the execution node of the task to be labeled from the task cluster.

[0127] Since the labeling task contains target data to be labeled, in order to enable different target data to execute their own labeling tasks, after obtaining the target data to be labeled in the rule executor, a corresponding labeling task can be created for the target data, and this labeling task can be sent to the task manager in the task cluster so that the task manager can determine the execution node for executing the labeling task from the task cluster. At this time, the creation of the labeling task can be generated based on the data identifier matching of different target data, so that one target data corresponds to one data identifier and one labeling task, so that a matching data labeling rule is retrieved for matching, which greatly improves the accuracy of data labeling execution. In addition, after receiving the labeling task, the task manager determines the execution node for executing data labeling, thereby achieving the purpose of determining the distribution of the labeling task to the corresponding execution node in the task manager.

[0128] In one embodiment of the present invention, in order to further define and illustrate, as Figure 8 As shown, after sending the labeling task and the corresponding data labeling rule to the task cluster in step 402, the method further includes:

[0129] 403. Determine an execution node in the task cluster that executes the labeling task of the target data;

[0130] 404. Send the labeling task and the corresponding data labeling rule to the execution node through the rule executor in the rule engine.

[0131] After the labeling task and the data labeling rules are sent to the task cluster, in order to accurately find the optimal execution node for the labeling task of the target data from the execution nodes that contain multiple task executions, thereby saving the execution efficiency of each labeling task, it is necessary to first determine the execution node for the labeling task of the target data in the task cluster. Among them, the execution node for determining the labeling task of the target data can be one or more. The task manager in the task cluster can determine the execution node corresponding to the labeling task that can execute the target data in the task cluster, and then send the labeling task and the corresponding data labeling rules to the determined execution node, so that the target data corresponding to the labeling task is labeled in the execution node based on the data labeling rules. At this time, since the execution node for determining the labeling task of the target data in the task cluster can be performed in the task manager, the matching efficiency of the execution node is greatly accelerated, thereby achieving the processing efficiency of data labeling.

[0132] In one embodiment of the present invention, in order to further define and illustrate, as Fig. 9 As shown, step 403 determines the execution node of the task cluster that executes the target data labeling task, including:

[0133] 4031. Obtain rule execution configuration information of the rule execution requirement information from the rule manager, and / or obtain data attributes of the target data from a distributed data source;

[0134] 4032. Determine, through the rule executor, an execution node that executes the labeling task corresponding to the rule execution configuration information and / or the data attribute.

[0135] In order to further achieve the matching of the execution nodes and realize the labeling purpose of the optimal resources of the labeling task, the execution node for performing the labeling task can be determined based on the rule execution configuration information and / or data attributes of the rule executor. Among them, since the rule execution configuration information is the content in the rule execution requirement information, when determining the execution node based on the rule executor, the rule execution configuration information in the rule execution requirement information can be obtained from the rule manager. At the same time, since the target data is stored in the distributed data source, when the rule executor determines the execution node, the data attributes of the target data can be obtained from the distributed data source, so that the rule executor determines the execution node for performing the labeling task based on the rule execution configuration information and / or data attributes. Among them, the rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task, and the data attributes include field content and business data source, so as to match the execution node for the labeling task according to the rule execution configuration information and data attributes, realize the task execution of data labeling with the optimal resources, and greatly improve the efficiency of the execution node in labeling the target data.

[0136] In one embodiment of the present invention, in order to further limit and explain, step 4032 determines, through the rule executor, the execution node that executes the labeling task corresponding to the rule execution configuration information, specifically including: obtaining the node resource information of the execution node in the task cluster; based on the node resource occupancy, node resource execution language, and node resource execution environment in the node resource information and matching with the rule execution configuration information, determining the execution node that executes the labeling task.

[0137] In order to accurately determine the execution node and improve the execution efficiency of the execution node for the labeling task, the node resource information of each execution node in the task cluster Yarn cluster is obtained so as to match it with the rule execution configuration information and determine the execution node. Among them, the rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task, and the node resource information includes at least one of the node resource occupancy, node resource execution language, and node resource execution environment, that is, each execution node in the Yarn cluster, whether it is executing the labeling task or not, pre-defines the node resource space size, node resource execution language, and node resource execution environment. For example, the node resource space size is defined as 256 megabytes, the node resource execution language is Java language, and the node resource execution environment is Windows xp. At the same time, since the rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task, the execution node can be confirmed based on the node resource information of each execution node and the rule execution configuration information. During this process, the rule executor can determine the execution node for executing the labeling task based on a one-to-one match between the node resource space size, the node resource execution language, the node resource execution environment and the system environment, the data space size, and the operation language, and finally can execute at least one execution node of the data labeling rule.

[0138] In addition, in the embodiment of the present invention, for a Yarn cluster configured with multiple nodes that perform labeling tasks, in order to avoid the situation where the execution node for executing the data labeling rule cannot be matched, when configuring the execution node, it can be expected that the node resource space size, node resource execution language, and node resource execution environment can be configured in a compatible manner. For example, the node resource execution environment can be configured as the entire system environment of the Windows series, etc., so that when matching using the rule execution configuration information, at least one execution node can be matched. The embodiment of the present invention does not make specific limitations.

[0139] In one embodiment of the present invention, for further limitation and explanation, step 4032 determines, through the rule executor, an execution node that executes the labeling task corresponding to the rule execution configuration information, and specifically also includes: obtaining source data of the labeling task to be executed and node resource information of at least one execution node from a distributed data source; parsing data attributes of the source data; and allocating execution nodes corresponding to node resource information matching the data attributes to different data in the source data to determine execution nodes that execute labeling of different target data.

[0140] In order to achieve the purpose of matching execution nodes based on data attributes and improve the execution efficiency of labeling tasks, the rule executor can obtain source data and node resource information corresponding to each execution node from distributed data sources, such as Kafka. Among them, the source data is the data corresponding to different data sources stored in the source database. At this time, since the rule executor can allocate labeling tasks through the Yarn cluster, the data source can be directly connected to the Kafka database, that is, the target data to be labeled is transferred to the Kafka database for effective retrieval. In addition, the node resource information includes at least one of the node resource space size, the node resource execution language, and the node resource execution environment. According to the data attributes of different data of the parsed source data, that is, including field content and business data source, the source data is assigned execution nodes that match the node resource information according to the data attributes to determine the data applicable to the execution of data labeling rules in different execution nodes. Execution nodes corresponding to node resource information that matches data attributes are allocated to different data in the source data. Specifically, this can be determined based on whether the field content matches the node resource space size, the node resource execution language, and whether the business data source matches the node resource execution environment. If they match, an allocation relationship is established between the data of the node resource information that matches the execution node in the source data and the execution node. After determining the target data, the rule executor searches for matching execution nodes based on the data attributes of the target data. The embodiments of the present invention do not make specific limitations.

[0141] In one embodiment of the present invention, for further limitation and explanation, step 2013 determines the execution node in the task cluster that executes the labeling task of the target data, including: if it is determined based on the rule execution configuration information that there are multiple execution nodes for executing the labeling task, then the execution node for executing the labeling task is determined based on matching the data attributes of the target data with the node resource information of the execution node.

[0142] Since the rule executor may match multiple execution nodes based on the rule execution configuration information or data attributes, for a scenario in which the execution node is determined based on the rule execution configuration information and the data attributes, in order to improve the efficiency of the execution node in executing the data labeling rule, the execution node that can execute the data labeling rule is first determined based on the rule execution configuration information. If there are multiple execution nodes determined, although the Spark task executor in the Yarn cluster can determine a target execution node from multiple execution nodes in a pre-set random manner or sorting manner, the execution node determined by the random manner or sorting manner is not necessarily the most resource-optimal. Therefore, on the basis of the rule executor determining multiple execution nodes based on the rule execution configuration information, the data attributes of the target data are further matched with the node resource information of the multiple confirmed execution nodes to determine the execution node with the best resources for executing the labeling task. In the embodiment of the present invention, how the data attributes are matched with the node resource information has been explained and will not be repeated.

[0143] An embodiment of the present invention provides a data labeling method. Compared with the prior art, the embodiment of the present invention obtains target data of a labeling task to be executed, and calls a labeling execution instruction that matches the target data, wherein the labeling execution instruction is generated based on input labeling logic information and labeling requirement information; determines an execution node for executing the labeling task based on execution configuration information of the labeling execution instruction and / or data attributes of the target data; executes the labeling execution instruction on the target data in the execution node to complete data labeling of the target data, and simplifies the generation steps of complex labeling rules by automatically matching data identification rules based on a rule engine, so that the labeling matching conditions can be highly expanded. At the same time, rule matching based on the rule engine meets the business needs for flexible data labeling, and convenient processing of data labeling is achieved, thereby improving the efficiency of data labeling.

[0144] Furthermore, as a response to the above Figure 1 In order to realize the method shown in FIG. 1 , an embodiment of the present invention provides a data labeling device, such as Fig.10 As shown, the device comprises:

[0145] An acquisition module 51 is used to acquire target data to be labeled;

[0146] A retrieval module 52, configured to retrieve data annotation rules matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data;

[0147] The labeling module 53 is used to label the target data based on the data labeling rule to obtain the labeling result of the target data.

[0148] Furthermore, the rule engine includes: a rule manager and a rule executor, and the device also includes:

[0149] A loading module is used for the rule manager in the rule engine to load the entered rule execution requirement information through the front-end interface. The rule execution requirement information is used to characterize the logical conditions when performing the annotation operation, as well as the rule operation mode and rule operation attributes;

[0150] A generation module, used to generate a data annotation rule matching the rule execution requirement information in the rule manager;

[0151] A first submission module is used for the rule manager to submit the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; and / or,

[0152] A receiving module, used for a rule manager in a rule engine to receive an initial data annotation rule through a front-end interface, and to process the initial data annotation rule to obtain the data annotation rule;

[0153] The second submission module is used for the rule manager to submit the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule.

[0154] Furthermore, the generating module comprises:

[0155] A parsing unit, used for the rule manager to perform syntax parsing on the rule execution requirement information to obtain rule logic information;

[0156] The identification unit is used for the rule manager to perform contextual logic identification on the rule logic information and generate data annotation rules.

[0157] Furthermore, the device also includes:

[0158] A determination module, configured to determine, after the rule executor receives the data annotation rule submitted by the rule manager, data information that can be annotated by each data annotation rule, wherein the data information includes: a data identifier;

[0159] A storage module, used to associate and store the data labeling rules and corresponding data identifiers in a preset storage location;

[0160] The calling module includes:

[0161] A parsing unit, configured to parse the labeling task when the rule executor receives the labeling task, and obtain the target data to be labeled;

[0162] A retrieving unit is used to retrieve the matching data labeling rules in the preset storage location based on the data identifier of the target data to be labeled, and send the labeling task and the corresponding data labeling rules to the task cluster to perform data labeling through the task cluster.

[0163] Furthermore, the device also includes:

[0164] A creation module is used to create a corresponding labeling task for the target data after obtaining the target data to be labeled, and send the labeling task to the task manager in the task cluster so that the task manager can determine the execution node of the task to be labeled from the task cluster.

[0165] Furthermore, the device further comprises: a sending module,

[0166] The determination module is further used to determine an execution node in the task cluster that executes the labeling task of the target data;

[0167] The sending module is also used to send the labeling task and the corresponding data labeling rule to the execution node through the rule executor in the rule engine, so that the target data corresponding to the labeling task can be labeled based on the data labeling rule by the execution node.

[0168] Furthermore, the determining module includes:

[0169] an acquisition unit, configured to acquire rule execution configuration information of the rule execution requirement information from the rule manager, and / or acquire data attributes of the target data from a distributed data source;

[0170] A determination unit is used to determine, through the rule executor, an execution node that executes the labeling task corresponding to the rule execution configuration information and / or the data attribute.

[0171] Furthermore, the determining unit includes:

[0172] A first acquisition subunit is used to acquire node resource information of execution nodes in the task cluster, wherein the node resource information includes at least one of node resource occupancy, node resource execution language, and node resource execution environment;

[0173] A determination subunit is used to determine the execution node for executing the labeling task by matching the node resource occupancy, node resource execution language, and node resource execution environment in the node resource information with the rule execution configuration information, wherein the rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task.

[0174] Furthermore, the determining unit further includes:

[0175] A second acquisition subunit is used to acquire source data of the to-be-executed labeling task and node resource information of at least one execution node from a distributed data source;

[0176] A parsing subunit, used to parse data attributes of the source data, wherein the data attributes include field content and business data source;

[0177] The allocation subunit is used to allocate execution nodes corresponding to the node resource information matching the data attributes to different data in the source data, so as to determine the execution nodes for executing the annotations of different target data.

[0178] Furthermore, the determination unit is specifically used to determine the execution node for executing the labeling task by matching the data attributes of the target data with the node resource information of the execution node if it is determined based on the rule execution configuration information that there are multiple execution nodes for executing the labeling task.

[0179] The embodiment of the present invention provides a data labeling device. Compared with the prior art, the embodiment of the present invention obtains the target data of the labeling task to be executed, and calls the labeling execution instruction matching the target data, wherein the labeling execution instruction is generated based on the input labeling logic information and labeling requirement information; determines the execution node for executing the labeling task based on the execution configuration information of the labeling execution instruction and / or the data attribute of the target data; executes the labeling execution instruction on the target data in the execution node to complete the data labeling of the target data, and realizes automatic matching of data identification rules based on the rule engine to simplify the generation steps of complex labeling rules, so that the labeling matching conditions can be highly expanded. At the same time, the rule matching based on the rule engine meets the business needs for flexible data labeling, realizes convenient processing of data labeling, and thus improves the efficiency of data labeling.

[0180] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the data labeling method in any of the above method embodiments.

[0181] Fig.11 A schematic diagram of the structure of a terminal provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the terminal.

[0182] like Fig.11 As shown, the terminal may include: a processor (processor) 602 , a communication interface (Communications Interface) 604 , a memory (memory) 606 , and a communication bus 608 .

[0183] The processor 602 , the communication interface 604 , and the memory 606 communicate with each other via a communication bus 608 .

[0184] The communication interface 604 is used to communicate with other devices such as clients or other servers.

[0185] The processor 602 is used to execute the program 610, and specifically can execute the relevant steps in the above data labeling method embodiment.

[0186] Specifically, the program 610 may include program codes, which include computer operation instructions.

[0187] The processor 602 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0188] The memory 606 is used to store the program 610. The memory 606 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0189] The program 610 may be specifically configured to enable the processor 602 to perform the following operations:

[0190] Obtain target data to be labeled;

[0191] Retrieving data annotation rules matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data;

[0192] The target data is labeled based on the data labeling rule to obtain a labeling result of the target data.

[0193] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0194] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data labeling method, characterized in that: include: Obtain target data to be labeled; Retrieving data annotation rules matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data; Annotate the target data based on the data annotation rule to obtain an annotation result of the target data; The rule engine includes: a rule manager and a rule executor. Before the rule engine is used to call the data annotation rule matching the target data, the method further includes: The rule manager in the rule engine loads the entered rule execution requirement information through the front-end interface. The rule execution requirement information is used to characterize the logical conditions when executing the annotation operation, as well as the rule operation mode and rule operation attributes; Generating, in the rule manager, a data annotation rule that matches the rule execution requirement information; The rule manager submits the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; and / or, The rule manager in the rule engine receives the initial data annotation rule through the front-end interface, and processes the initial data annotation rule to obtain the data annotation rule; The rule manager submits the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; After the rule manager submits the data annotation rule to the rule executor in the rule engine, the method further includes: After receiving the data annotation rules submitted by the rule manager, the rule executor determines the data information that can be annotated by each data annotation rule, wherein the data information includes: a data identifier; The data annotation rules and corresponding data identifiers are stored in a preset storage location in association with each other; The step of retrieving a data annotation rule matching the target data through a rule engine includes: When the rule executor receives the labeling task, it parses the labeling task to obtain the target data to be labeled; Based on the data identifier of the target data to be labeled, the matching data labeling rules in the preset storage location are retrieved, and the labeling task and the corresponding data labeling rules are sent to the task cluster to perform data labeling through the task cluster.

2. The method according to claim 1, characterized in that: Generating a data annotation rule matching the rule execution requirement information in the rule manager includes: The rule manager performs syntax analysis on the rule execution requirement information to obtain rule logic information; The rule manager performs contextual logic identification on the rule logic information to generate data annotation rules.

3. The method according to claim 1, characterized in that The method further comprises: After the target data to be labeled is acquired, a corresponding labeling task is created for the target data, and the labeling task is sent to a task manager in the task cluster, so that the task manager determines an execution node of the task to be labeled from the task cluster.

4. The method according to claim 1, characterized in that After sending the labeling task and the corresponding data labeling rule to the task cluster, the method further includes: Determine an execution node in the task cluster that executes the labeling task of the target data; The labeling task and the corresponding data labeling rule are sent to the execution node through the rule executor in the rule engine, so that the target data corresponding to the labeling task is labeled by the execution node based on the data labeling rule.

5. The method according to claim 4, characterized in that The execution node for executing the target data labeling task in the determination task cluster includes: Acquire the rule execution configuration information of the rule execution requirement information from the rule manager, and / or acquire the data attribute of the target data from the distributed data source; The rule executor determines an execution node that executes the labeling task corresponding to the rule execution configuration information and / or the data attribute.

6. The method according to claim 5, characterized in that The determining, by the rule executor, an execution node that executes the labeling task corresponding to the rule execution configuration information comprises: Obtaining node resource information of execution nodes in the task cluster, wherein the node resource information includes at least one of node resource occupancy, node resource execution language, and node resource execution environment; Based on the node resource occupancy, node resource execution language, and node resource execution environment in the node resource information and matching with the rule execution configuration information, the execution node for executing the labeling task is determined. The rule execution configuration information is used to characterize the system environment, data space size, and operation language required to execute the labeling task.

7. The method according to claim 5, characterized in that The determining, by the rule executor, an execution node for executing the labeling task corresponding to the data attribute comprises: Acquire source data of a to-be-executed labeling task and node resource information of at least one execution node from a distributed data source; Parsing data attributes of the source data, wherein the data attributes include field content and business data source; Execution nodes corresponding to the node resource information matching the data attributes are allocated to different data in the source data, so as to determine execution nodes for executing annotations of different target data.

8. The method according to claim 4, characterized in that The execution node for executing the target data labeling task in the determination task cluster includes: If it is determined based on the rule execution configuration information that there are multiple execution nodes for executing the labeling task, the execution node for executing the labeling task is determined based on matching the data attributes of the target data with the node resource information of the execution node.

9. A data labeling device, characterized in that: include: An acquisition module is used to acquire target data to be labeled; A calling module, used to call a data annotation rule matching the target data through a rule engine, wherein the rule engine stores data annotation rules matching different data; A labeling module, used to label the target data based on the data labeling rule to obtain a labeling result of the target data; The rule engine includes: a rule manager and a rule executor, and the device also includes: A loading module is used for the rule manager in the rule engine to load the entered rule execution requirement information through the front-end interface. The rule execution requirement information is used to characterize the logical conditions when performing the annotation operation, as well as the rule operation mode and rule operation attributes; A generation module, used to generate a data annotation rule matching the rule execution requirement information in the rule manager; a submission module, used for the rule manager to submit the data annotation rule to the rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; and / or, A receiving module, used for a rule manager in a rule engine to receive an initial data annotation rule through a front-end interface, and to process the initial data annotation rule to obtain the data annotation rule; the rule manager submits the data annotation rule to a rule executor in the rule engine, so that the rule executor performs data annotation based on the data annotation rule; The device also includes: A determination module, configured to determine, after the rule executor receives the data annotation rule submitted by the rule manager, data information that can be annotated by each data annotation rule, wherein the data information includes: a data identifier; A storage module, used to associate and store the data labeling rules and corresponding data identifiers in a preset storage location; The calling module includes: A parsing unit, configured to parse the labeling task when the rule executor receives the labeling task, and obtain the target data to be labeled; A retrieving unit is used to retrieve the matching data labeling rules in the preset storage location based on the data identifier of the target data to be labeled, and send the labeling task and the corresponding data labeling rules to the task cluster to perform data labeling through the task cluster.

10. A storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to perform operations corresponding to the data labeling method as described in any one of claims 1-8.

11. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform an operation corresponding to the data labeling method according to any one of claims 1-8.

Images