Condition adjustment instruction real-time generation method and system

Abstract

The invention relates to the technical field of simulation training, in particular to a situation adjustment instruction real-time generation method and system, and the method comprises the steps: constructing a training situation comprehensive knowledge graph based on training basic knowledge and training adjustment plan field knowledge; in combination with expert knowledge and adjustment plan preprocessing, a condition adjustment rule set suitable for the training is constructed; acquiring a training condition real-time identification result through training condition real-time convergence and training condition intelligent identification; and generating a real-time training condition adjustment instruction by adopting a multi-level rule comprehensive reasoning and rule parameter reasoning technology. The training condition adjustment instruction generated by the technical scheme of the invention has the effects of short time consumption, high accuracy and high reliability.

Description

[0001] The invention belongs to the field of simulation training, and in particular to a method and system for real-time generation of training situation adjustment instructions. Background Art

[0002] With the continuous deepening of information and intelligent construction, the training model is constantly innovating, gradually changing from planned "script-based" drills to random adjustments of the entire process and link. The functions of the training situation adjustment system mainly include initial training situation release, multi-dimensional training situation aggregation, real-time situation adjustment, etc. Its purpose is to induce the trainees to train according to the intention of the director's office. It is the "brain" of the entire training activity. As the core technology of real-time situation adjustment, the generation of training situation adjustment instructions is the focus of attention in the field of simulation training.

[0003] Prior art CN115311923A and prior art CN118197129A disclose methods for generating training situation adjustment instructions. However, the methods for generating situation adjustment instructions involved in the above prior art all require professionals familiar with the equipment to send relevant instructions through software, which requires a high knowledge system of the sender of the adjustment instructions, and the efficiency is also very low, which is difficult to meet the needs of modern training development. Summary of the invention

[0004] In view of the above technical problems in the art, the present invention proposes a method for generating a situation adjustment instruction in real time, such as Figure 1 As shown, it is characterized in that it includes the following steps:

[0005] S1: Based on the basic knowledge of training and the domain knowledge of training adjustment plans, a comprehensive knowledge graph of training situations is constructed;

[0006] S2: Combining expert knowledge and adjustment plan preprocessing, a collection of adjustment rules suitable for this training situation is constructed;

[0007] S3: Obtain real-time identification results of training situations through real-time aggregation of training situations and intelligent identification of training situations;

[0008] S4: Use multi-level rule comprehensive reasoning and rule parameter reasoning technology to generate real-time training situation adjustment instructions.

[0009] The training situation adjustment instructions generated by the technical solution of the present invention have the effects of short time consumption, high accuracy and strong reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 : A flow chart of a method for real-time generation of situation adjustment instructions;

[0011] Figure 2 : A flowchart of a specific embodiment. DETAILED DESCRIPTION

[0012] The following examples further illustrate the content of the present invention, but should not be construed as limiting the present invention. Without departing from the spirit and substance of the present invention, modifications or substitutions made to the methods, steps or conditions of the present invention all fall within the scope of the present invention.

[0013] First, the relevant terms involved in the present invention are explained and illustrated.

[0014] The "Training Situation Adjustment System" is the "brain" that integrates training information, deploys training resources, and controls training activities. It is mainly used for situation integration monitoring of exercise training, real-time situation adjustment, and adjustment of command issuance. It induces participants to conduct confrontation training in accordance with the intentions of the director's office and is the core of the training information system.

[0015] "Training situation adjustment instructions" are executable adjustment instructions issued based on the training situation adjustment system, according to the training adjustment plan and the current training situation, in order to adjust the content of the training environment and intervene in and control the personnel and equipment of the participating parties.

[0016] Based on the above scheme, some preferred schemes, S1, constructing a comprehensive knowledge graph of training conditions, specifically includes the following steps:

[0017] S11: It is mainly aimed at basic knowledge in the field of training. The data source is mainly structured data such as equipment performance, training style, and training command process accumulated in the training industry. After importing the data, the Bert-score text similarity model is used to calculate the similarity between different entities and attributes. When the similarity threshold is higher than a certain level, the data of entities and attributes are fused; then the Chinese relation extraction model based on BERT-wwm is used to establish the association relationship between different entities;

[0018] Based on the above scheme, in an embodiment for basic knowledge in the training field, the similarity threshold is set to 0.9; in order to cope with the fact that the text of basic knowledge in the training field is relatively long, the specific model used by BERT-wwm is RoBERTa-wwm-ext-large.

[0019] S12: Mainly for training and adjusting the domain knowledge of the plan, the data source is mainly the unstructured and semi-structured data such as the command and control plan, situation handling, analysis and evaluation accumulated in the historical training. For this type of data, first use the entity extraction method based on the BERT-BiLSTM-CRF model to automatically identify the named entities, use the attribute extraction model to extract the entity attributes, and use the Chinese relation extraction model based on BERT-wwm to extract the association relationship between entities from the relevant corpus;

[0020] Based on the above scheme, in an implementation example for training adjustment plan domain knowledge, in order to cope with the actual situation that the training adjustment plan text is relatively long, the specific model used by BERT-wwm is RoBERTa-wwm-ext-large.

[0021] S13: Based on the training terminology, training encyclopedia, training dictionary and other related ontologies accumulated in the industry, combined with the training plan classification table and other contents, we invite experts in the training field to use the Progtégé software to semi-automatically construct a multi-level comprehensive ontology of training situations, mainly by manually linking key entities of different ontologies and automatically linking non-key entities;

[0022] S14: Based on the multi-level training situation comprehensive ontology, the Bert-score text similarity model and the Chinese relationship extraction model based on BERT-wwm are used to perform knowledge fusion operations such as entity linking, coreference resolution, and entity disambiguation on the entities, attributes, and associations in S11 and S12 to obtain the comprehensive knowledge graph of the training situation.

[0023] Based on the above schemes, the entity linking step of some schemes specifically includes a candidate entity generation step and an entity disambiguation step. In the candidate entity generation step, fuzzy matching is performed based on the text information of the entity mention and the entity in the training situation comprehensive knowledge graph, and at the same time, screening is performed based on the domain label of the situation plan context in which the entity mention is located, and the situation plan entities with the same domain label as the entity mention are retained as candidate entities.

[0024] The entity linking step in the above scheme also includes the following steps: if there is only one candidate entity, a direct connection is established; if there is no candidate entity, the entity is not mentioned and marked as an entity; if there is more than one candidate entity, the candidate entities are sorted through the entity disambiguation step, and the candidate entities are matched with the plan entities in the comprehensive knowledge graph of the training situation according to the sorting results, and each plan entity is outputted and associated with the entity of the comprehensive knowledge graph of the training situation.

[0025] The domain labels of the situation plan context in the above scheme are generated by a fine-tuned BERT model.

[0026] Based on the above scheme, some preferred schemes, S2, constructing a set of adjustment rules suitable for this training situation, specifically includes the following steps:

[0027] S21: The training director department selects a suitable initial situation adjustment plan from the situation adjustment plan library collected in the industry in the early stage through expert voting, keyword matching, text similarity comparison and other means, and through comprehensive scoring. After the training field experts conduct content verification, detail modification, digital correction and other modification operations, the plan becomes the situation adjustment plan used in this training;

[0028] S22: Adjust the plan for the situation in S21. First, use the jieba word segmentation tool to segment the plan text, and then use the entity extraction method based on the BERT-BiLSTM-CRF model to perform named entity recognition.

[0029] S23: In the comprehensive knowledge graph of the training situation described in S1, search for the named entity identified by S22, combine the situation adjustment context information of the entity in the situation adjustment plan in S21, and the graph reasoning result of the comprehensive knowledge graph of the training situation in S14 for the above situation adjustment, to obtain a collection of situation adjustment rules suitable for this training.

[0030] Based on the above scheme, some preferred schemes, S3, obtaining the real-time recognition result of the training situation, specifically includes the following steps:

[0031] S31: Real-time convergence of training situations. The training situation adjustment system automatically collects status information from the participating equipment and systems of each participating unit, reads training environment information from sensors in the virtual training environment or the real training environment, and converges it to the director's training information system in real time through the training network. The automatic fusion function of the training situation of the director's training information system is used to form a dynamic and real-time training situation diagram.

[0032] S32: Intelligent recognition of training situations. Collect common training situation diagrams and image-text pair data of training situation recognition results, and fine-tune the image-text dual-stream Bert multimodal model. Based on the fine-tuned image-text dual-stream Bert model, recognize the real-time training situation diagram described in S31 to obtain the real-time recognition result of the training situation.

[0033] Based on the above scheme, in an embodiment of intelligent identification of training conditions, the ViLBERT architecture is used. The architecture pre-trains visual and text input information separately through the common attention Transformer layer. The text uses a unimodal Transformer encoder to extract context information, and the image uses a pre-trained Faster R-CNN to obtain image representation.

[0034] Based on the above scheme, in some preferred schemes, generating the real-time training situation adjustment instruction in S4 specifically includes the following steps:

[0035] S41: Based on the real-time recognition result of the training situation described in S31 and the collection of situation adjustment rules described in S23, a multi-level rule comprehensive reasoning technology based on confidence is used to obtain an activation rule suitable for the current training situation, which is the real-time rule reasoning result.

[0036] S42: Based on the real-time rule reasoning results described in S41, combined with the context of the situation adjustment plan and the comprehensive knowledge graph of the training situation, the rapid reasoning function of the graph is used to perform parameter reasoning on the situation adjustment rules.

[0037] S43: Start the situation adjustment rule operation, and convert the result of the rule operation into a collection of situation adjustment instructions through the knowledge graph. Through named entity matching, the shortest instruction order path is found, and the collection of situation adjustment instructions is quickly issued to the participants, equipment or environment, so that the exercise situation can be adjusted to meet the overall requirements of the exercise.

[0038] Based on the above scheme, a specific embodiment, the process is as follows Figure 2 shown.

[0039] Some solutions propose a method and system for real-time generation of situation adjustment instructions. The system includes at least one processor; and a memory storing instructions. When the instructions are executed by at least one processor, the steps of any of the above solutions are implemented.

[0040] Embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on one or more tangible, non-transitory program carriers for execution by, or to control the operation of, a data processing device. Alternatively or in addition, the program instructions may be encoded on an artificially generated propagating signal, such as a machine-generated electrical, optical, or electromagnetic signal, which is generated as encoded information for transmission to an appropriate receiver device for execution by a data processing device. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of the foregoing.

[0041] A computer program (which may also be referred to or described as a program, software, software application, module, software module, script or code) may be written in any form of programming language, including compiled or interpreted languages ​​or declarative or procedural languages, and the computer program may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a file in a file system. A program may be stored in a portion of a file that stores other programs or data, for example, one or more scripts stored in the following: in a markup language document; in a single file dedicated to a related program; or in multiple collaborative files, for example, files storing one or more modules, subroutines or code portions. A computer program may be deployed to execute on one or more computers, the computers being located in one place or distributed to multiple locations and interconnected by a communication network.

[0042] Although this specification contains many specific implementation details, these should not be interpreted as limitations on the scope of any invention or the scope of what can be claimed, but rather as an illustration of features that can concretize a particular embodiment of a particular invention. The specific features described in this specification in the context of independent embodiments can also be implemented in combination with a single embodiment. Conversely, the various features described in the context of a single embodiment can also be implemented independently in multiple embodiments, or in any suitable sub-combination. In addition, although the features can be described above as a combination effect and even initially required, one or more features from the required combination can be removed from the combination in some cases, and the required combination can be turned to a sub-combination or a variation of the sub-combination.

Claims

1. A method and system for real-time generation of situation adjustment instructions, characterized in that: The following steps are involved: S1: Based on the basic knowledge of training and the domain knowledge of training adjustment plans, a comprehensive knowledge graph of training situations is constructed; S2: Combine expert knowledge and adjustment plan preprocessing to build a collection of adjustment rules suitable for this training situation. S3: Obtain real-time identification results of training situations through real-time aggregation of training situations and intelligent identification of training situations; S4: Use multi-level rule comprehensive reasoning and rule parameter reasoning technology to generate real-time training situation adjustment instructions.

2. The method according to claim 1, characterized in that S1 constructs a comprehensive knowledge graph of training conditions, specifically including the following steps: S11: Mainly for structured data of basic knowledge in training fields. After importing data, first perform data fusion based on entities and attributes, and then use the relationship extraction model to establish the association relationship between different entities; S12: Mainly for training and adjusting the unstructured and semi-structured data of the domain knowledge of the plan. For this type of data, first use the entity extraction method to automatically identify the named entities, use the attribute extraction model to extract the entity attributes, and use the relationship extraction model to extract the relationship between entities from the relevant corpus; S13: Invite experts in the training field to semi-automatically construct a multi-level comprehensive ontology of training situations based on relevant ontologies accumulated in the industry; S14: Based on the multi-level training situation comprehensive ontology, use the text similarity model to perform entity linking and knowledge fusion on the entities, attributes, and associations in S11 and S12 to obtain the comprehensive knowledge graph of the training situation.

3. The method according to claim 1, characterized in that S2 constructs a set of adjustment rules suitable for this training situation, specifically including the following steps: S21: The training director department selects a suitable situation adjustment plan from the situation adjustment plan library, and after verification and modification by experts, it becomes the situation adjustment plan adopted for this training; S22: Adjust the contingency plan in S21 by first performing word segmentation and then performing named entity recognition. S23: In the comprehensive knowledge graph of the training situation described in S1, search for the named entities identified in S22, combine the context of the entities and the reasoning results of the knowledge graph, and obtain a collection of adjustment rules suitable for this training situation.

4. The method according to claim 1, characterized in that: S3 obtains the real-time recognition result of the training situation, specifically including the following steps: S31: Real-time aggregation of training conditions. The training system automatically collects status information from the participating equipment and systems of each participating unit, and aggregates it to the real-time situation map of the director's comprehensive training system through the training network. S32: Intelligent recognition of training conditions: The real-time situation diagram described in S31 is recognized by using multi-modal intelligent recognition technology to obtain a real-time recognition result of the training conditions.

5. The method according to claim 1, characterized in that S4 generates a real-time training situation adjustment instruction, specifically comprising the following steps: S41: Based on the real-time recognition result of the training situation described in S31 and the collection of situation adjustment rules described in S23, a multi-level rule comprehensive reasoning technology is used to obtain a real-time rule reasoning result. S42: Based on the real-time rule reasoning results described in S41, combined with the context of the situation adjustment plan and the comprehensive knowledge graph of the training situation, parameter reasoning is performed on the situation adjustment rules. S43: Start the situation adjustment rule operation, and convert the result of the rule operation into situation adjustment instructions through the knowledge graph. By issuing these adjustment instructions, the exercise situation can be adjusted to meet the overall requirements of the exercise.

6. A system for generating situation adjustment instructions in real time, characterized in that: The system comprises at least one processor; and a memory storing instructions, which, when executed by at least one processor, implements the steps in a method for real-time generation of a situation adjustment instruction according to any one of claims 1-5.

Images