Ship navigation rule autonomous interpretation method considering scene features

Abstract

The invention provides a ship navigation rule autonomous interpretation method considering scene features, which comprises the following steps: designing a navigation rule text labeling method, and labeling scene features and operation attributes; the method comprises the following steps: designing a navigation rule automatic interpretation method based on semantic analysis, analyzing a navigation rule text by utilizing named entity recognition and a semantic role labeling model, extracting navigation scene information and a corresponding ship operation rule, establishing a mapping relation between a scene and the navigation rule, and generating a logic triple set of the ship navigation rule; a ship navigation rule knowledge graph is constructed, a ship navigation scene and an operation rule are dynamically associated, and textual navigation rules are converted into structured semantic situation interpretation. According to the method, the understanding and explanation of the ship navigation rule are realized, the pure named entity identification level is exceeded, the practice dimension of scenarized application is deepened, the integrity of a logic chain as a scene constraint-operation rule is emphasized, and intelligent support is provided for port traffic management and ship navigation safety.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent port traffic supervision and natural language processing technology, and in particular to a method for autonomously interpreting ship navigation rules taking scene characteristics into consideration. Background Art

[0002] With the introduction of the concept of smart shipping, intelligent shipping has become an industry consensus. By restricting ship behavior through traffic rules, unmanned ships can achieve intelligent navigation, and ship behavior supervision has become more and more common.

[0003] However, how to effectively convert the non-binary text form of port navigation rules (oriented to human reading comprehension) into a machine-executable language, that is, the interpretation of navigation rules, is still in the initial stage of exploration. Different from other rules, the text of navigation rules has more obvious scenario characteristics, so complex language structures are often used in expression, including ambiguous expressions, long sentences, nested sentences and polysemous words, aiming to cover more situations and circumstances, which are very difficult for machines to understand and process. Given that the ship navigation rules system is not only intertwined with the interactive relationship between multiple participants such as ships, port facilities and the environment, but also reflects a high degree of constraint and precision in logical construction. Therefore, the understanding and interpretation of navigation rules needs to go beyond the simple level of named entity recognition, go deep into the practical dimension of scenario-based applications, and emphasize the integrity of the logical chain of "scenario constraints-operation rules". Summary of the invention

[0004] According to the technical problems raised above, a method for autonomous interpretation of ship navigation rules taking into account scenario characteristics is provided. The present invention focuses on establishing a clear mapping relationship between scenarios and navigation rules by designing a text annotation method with scenario and operation characteristics, so that the model can identify and define specific scenarios and conditions and operation rules applicable to navigation rules. An automated interpretation method based on semantic parsing is proposed, which combines machine learning with deep learning technology, and uses semantic role labeling (SRL) and named entity recognition (NER) models to parse scenarios and operation rules in regulatory texts. Finally, by constructing a rule knowledge graph, a rule interpretation system with clear semantic hierarchy is constructed to achieve automated rule matching and contextualized interpretation.

[0005] The technical means adopted by the present invention are as follows:

[0006] A method for autonomous interpretation of ship navigation rules considering scene characteristics, comprising:

[0007] S1. Integrate scene features and operation attributes, design a navigation rules text annotation method, and annotate scene features and operation attributes;

[0008] S2. Based on semantic analysis, design an automated interpretation method for navigation rules, use named entity recognition and semantic role labeling models to parse the navigation rules text, accurately extract navigation scenario information and corresponding ship operation rules, establish a clear mapping relationship between scenarios and navigation rules, and generate a logical triple set of ship navigation rules;

[0009] S3. Construct a knowledge graph of ship navigation rules to achieve dynamic association between ship navigation scenarios and operating rules, and convert textual navigation rules into structured semantic contextual interpretations.

[0010] Furthermore, in step S1, the marked scene features include the port water functional area, navigation speed, weather conditions, visibility, and the marked operation attributes include turning, deceleration, and U-turn.

[0011] Furthermore, step S1 specifically includes:

[0012] S11. Expand the concepts of context and operation to represent the context semantics including environment, state and ship maneuvering, and use the doccano platform to semi-automatically annotate the navigation rules text;

[0013] S12. Two conceptual entities, Situation and Rules, are introduced to perform a new round of annotation on navigation rules to identify the scenario description and rule description in the rules.

[0014] Furthermore, in step S11:

[0015] The elements of the situational semantics include Value, Compare, Introduction, Object, Location, Number, Attribute, Operate, and Interaction.

[0016] Furthermore, in step S2, named entity recognition and semantic role labeling models are used to parse the navigation scenarios, operating rules and their corresponding relationships in the navigation rules text, and accurately extract navigation scenario information including ship navigation paths, waterway regulations, and weather influences.

[0017] Further, step S2 specifically includes:

[0018] S21, use fine-tuned Albert-BiLSTM-CRF to identify conceptual entities of situational semantics;

[0019] S22, using the Electra small model, a Chinese semantic role labeling model pre-trained in HanLP, to analyze the semantic roles in the navigation rules text. To address the recognition defects caused by the lack of shipping proper nouns in its pre-trained corpus, the Location identified in S21 is structurally modified in the text;

[0020] S23. Combining the entity recognition result in step S21 and the semantic role recognition result in step S22, the relationship between entities is identified through rule setting.

[0021] Furthermore, the specific operation of the structural change in step S22 is:

[0022] Change "Location" to "At Location,"

[0023] Furthermore, in step S23, the semantic role recognition result in step S22 only retains the original annotation results of ARG0, ARG1, PRED, and ARGM-ADV to better reflect the relationship between the subject and the object.

[0024] Furthermore, in step S23, the rules are set as follows:

[0025] Object / Location / Attribute is parallel to ARG0 annotation as the head entity;

[0026] Object / Attribute / Value / Location is parallel to ARG1 annotation as the tail entity;

[0027] ARGM-ADV is parallel to the Introduction annotation as an adverb, which is the limiting state of the predicate relationship and forms a relationship with the predicate;

[0028] PRED and Interaction / Compare are marked in parallel as corresponding head entity and tail entity relationships;

[0029] The relationship between Object and the corresponding Attribute is HasAttribute;

[0030] The relationship between Object and the corresponding Operator is HasOperate;

[0031] The corresponding relationship between Situation and Rule is represented by the rule number, and the specific relationship is HasRule;

[0032] The relationship between Situation / Rule and the corresponding inspection object and entity is Contains.

[0033] Furthermore, in step S3, the logical hierarchy of the constructed ship navigation rules knowledge graph is: navigation scenario → navigation inspection elements → navigation attribute requirements and operating specifications.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] 1. The present invention provides an autonomous interpretation method for ship navigation rules that takes into account scenario characteristics, which can realize scenario-oriented autonomous interpretation of navigation rules, and can accurately identify domain entities and relationships in navigation rules. On the basis of ensuring the authenticity and diversity of information, it constructs a clear-layered and logically feasible automated rule knowledge graph, which is convenient for understanding and structural management of navigation rules. It thus provides a powerful auxiliary tool for intelligent port management.

[0036] 2. The present invention provides an autonomous interpretation method for ship navigation rules that takes scenario characteristics into consideration. By accurately identifying applicable scenarios and operating conditions, effective rule management is performed, reducing risk accidents caused by misjudgment and improper execution, while accelerating ship passage and optimizing logistics and operational efficiency.

[0037] 3. The present invention provides an autonomous interpretation method for ship navigation rules that takes scenario characteristics into consideration. The role-based interpretation of the knowledge graph enhances the decision-making support capabilities of port managers, promotes the transformation process of traffic management, achieves information transparency, and promotes everyone's good work, thereby providing strong support for the development of modern port operations.

[0038] Based on the above reasons, the present invention can be widely promoted in the fields of intelligent port traffic supervision, natural language processing, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 It is the overall framework diagram of the method of the present invention.

[0041] Figure 2 A schematic diagram of BIO annotation provided in an embodiment of the present invention.

[0042] Figure 3 A traffic rule named entity recognition model provided by an embodiment of the present invention.

[0043] Figure 4This is a graph of semantic role recognition results provided by an embodiment of the present invention.

[0044] Figure 5 An example of a navigation rules knowledge graph provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0046] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0047] like Figure 1 As shown, the present invention provides a method for autonomous interpretation of ship navigation rules taking into account scene characteristics, comprising:

[0048] S1. Integrate scene features and operation attributes, design a navigation rules text annotation method, and annotate scene features and operation attributes;

[0049] S2. Based on semantic analysis, an automated interpretation method for navigation rules is designed. Named entity recognition (NER) and semantic role labeling (SRL) models are used to parse navigation scenarios, operation rules and their corresponding relationships in the navigation rules text, accurately extract navigation scenario information including ship navigation paths, channel regulations, weather impacts and corresponding ship operation rules, and establish a clear mapping relationship between scenarios and navigation rules to generate a logical triple set of ship navigation rules.

[0050] S3. Construct a knowledge graph of ship navigation rules to achieve dynamic association between ship navigation scenarios (ship navigation status in specific waters) and operating rules (seaworthiness requirements, equipment requirements, operating specifications), and convert textual navigation rules into structured semantic contextual interpretations.

[0051] In specific implementation, as a preferred embodiment of the present invention, step S1 specifically includes:

[0052] S11. Expand the concepts of context and operation to represent the context semantics including environment, state (position, navigation status) and ship maneuvering, and use the doccano platform to semi-automatically annotate the navigation rules text;

[0053] S12. Two conceptual entities, Situation and Rules, are introduced to perform a new round of annotation on navigation rules to identify the scenario description and rule description in the rules.

[0054] In specific implementation, as a preferred embodiment of the present invention, in step S11:

[0055] The elements of the situational semantics include Value, Compare, Introduction, Object, Location, Number, Attribute, Operate, and Interaction. See Table 1 for details:

[0056] Table 1 BIOES scenario semantic entity element expansion table

[0057]

[0058] In this embodiment, in step S1, the rule text needs to be annotated with BIO, where B represents the beginning of an entity, I represents the inside of an entity, and O represents a non-entity. By applying BIO to the ship navigation rules and naming the entities, a variety of B, I, and O categories can be obtained, such as B-Situation, I-Situation, B-Object, and I-Object. Before annotation, the navigation rule text needs to be preprocessed, including manually selecting the document body, deleting invalid information, and standardizing the rule sequence. After the annotation on the doccano platform is completed (specific annotations such as Figure 2 As shown), it needs to be converted into BIO format and then converted into the input of the entity recognition model.

[0059] In specific implementation, as a preferred embodiment of the present invention, step S2 specifically includes:

[0060] S21, use fine-tuned Albert-BiLSTM-CRF to identify conceptual entities of situational semantics;

[0061] In this embodiment, the model first vectorizes the corpus labeled in step S1 and inputs it into the Albert layer for initial encoding to capture basic language features. To enhance the model's ability to capture context-dependent and complex relationships, the model adopts a combination of bidirectional gated recurrent unit and multi-head attention mechanism (BiLSTM) to extract subtle state features and form a robust hidden layer representation. Finally, the model uses conditional random field (CRF) as the decoding layer to ensure that the output entity labels have the highest probability, thereby achieving accurate and context-related entity recognition in the field of port traffic rules. As Figure 3 shown, it is an entity recognition model.

[0062] S22. Use the pre-trained Chinese semantic role labeling model Electra small model in HanLP to perform semantic role analysis on the navigation rules text. To address the recognition defect caused by the lack of shipping proper nouns in its pre-trained corpus, change the structure of the Location identified in S21 in the text;

[0063] In this embodiment, the specific operation of the structural change in step S22 is: change "Location" to the form of "at Location,". The semantic roles in HanLP are specifically shown in Table 2 as follows:

[0064] Table 2 HanLP Semantic Role Label Table

[0065] Label Character Label Character ARG0 Agent ARGM-ADV adverbial ARG1 The person who is involved ARGM-BNF Beneficiaries PRED predicate ARGM-CND condition ARGM-EXT scope ARGM-LOC Location

[0066] S23. Combine the entity recognition results in step S21 and the semantic role recognition results in step S22, and identify the relationships between entities through rule setting.

[0067] Specifically, as a preferred implementation manner of the present invention, in step S23, only the original annotation results of ARG0, ARG1, PRED, and ARGM-ADV in the semantic role recognition results in step S22 are retained to better reflect the relationship between the subject and the object. The semantic recognition results are as Figure 4 shown. By retaining the four semantic roles of ARG0, ARG1, PRED, and ARGM-ADV, the entity and predicate relationships can be corresponded, thereby realizing the representation in the form of a rule text logical triple.

[0068] Specifically, as a preferred implementation manner of the present invention, in step S23, the rule setting is as follows:

[0069] Object / Location / Attribute and ARG0 annotation parallel are used as the head entity;

[0070] Object / Attribute / Value / Location is parallel to ARG1 annotation as the tail entity;

[0071] ARGM-ADV is parallel to the Introduction annotation as an adverb, which is the limiting state of the predicate relationship and forms a relationship with the predicate;

[0072] PRED and Interaction / Compare are marked in parallel as corresponding head entity and tail entity relationships;

[0073] The relationship between Object and the corresponding Attribute is HasAttribute;

[0074] The relationship between Object and the corresponding Operator is HasOperate;

[0075] The corresponding relationship between Situation and Rule is represented by the rule number, and the specific relationship is HasRule;

[0076] The relationship between Situation / Rule and the corresponding inspection object and entity is Contains.

[0077] In specific implementation, as a preferred embodiment of the present invention, in step S3, the logical hierarchy of the ship navigation rules knowledge graph constructed is: navigation scenario → navigation inspection elements → navigation attribute requirements and operating specifications.

[0078] In this embodiment, in the graph construction process of step S3, in addition to the identified relationships in the constructed knowledge graph, the customized entity class relationships are fixed, and the triple set obtained in step 2 is integrated into Neo4j. In order to distinguish entities with the same name in different scenarios and enhance the readability of the rules, entities with the same name in different scenarios are set as different entities in the graph construction process and distinguished by unique identifiers. The knowledge graph finally constructed is as follows: Figure 5 shown.

[0079] In summary, the present invention proposes an autonomous interpretation method of navigation rules under the constraint framework of "scenario constraints-operation rules". This method combines the named entity recognition method (NER) and the semantic role labeling (SRL) information extraction model to extract knowledge from ship navigation rules and realize the representation of knowledge graphs in the neo4j platform.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for autonomous interpretation of ship navigation rules considering scene characteristics, characterized in that: include: S1. Integrate scene features and operation attributes, design a navigation rules text annotation method, and annotate scene features and operation attributes; S2. Based on semantic analysis, design an automated interpretation method for navigation rules, use named entity recognition and semantic role labeling models to parse the navigation rules text, accurately extract navigation scenario information and corresponding ship operation rules, establish a clear mapping relationship between scenarios and navigation rules, and generate a logical triple set of ship navigation rules; S3. Construct a knowledge graph of ship navigation rules to achieve dynamic association between ship navigation scenarios and operating rules, and convert textual navigation rules into structured semantic contextual interpretations.

2. According to claim 1, a method for autonomous interpretation of ship navigation rules considering scene characteristics is characterized in that: In step S1, the marked scene features include the port water functional area, navigation speed, weather conditions, visibility, and the marked operation attributes include turning, deceleration, and U-turn.

3. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 1 is characterized in that: Step S1 specifically includes: S11. Expand the concepts of context and operation to represent the context semantics including environment, state and ship maneuvering, and use the doccano platform to semi-automatically annotate the navigation rules text; S12. Two conceptual entities, Situation and Rules, are introduced to perform a new round of annotation on navigation rules to identify the scenario description and rule description in the rules.

4. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 3 is characterized in that: In step S11: The elements of the situational semantics include Value, Compare, Introduction, Object, Location, Number, Attribute, Operate, and Interaction.

5. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 1 is characterized in that: In step S2, named entity recognition and semantic role labeling models are used to parse the navigation scenarios, operating rules and their corresponding relationships in the navigation rules text, and accurately extract navigation scenario information including ship navigation paths, waterway regulations, and weather impacts.

6. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 1 is characterized in that: Step S2 specifically includes: S21, use fine-tuned Albert-BiLSTM-CRF to identify conceptual entities of situational semantics; S22, using the Electra small model, a Chinese semantic role labeling model pre-trained in HanLP, to analyze the semantic roles in the navigation rules text. To address the recognition defects caused by the lack of shipping proper nouns in its pre-trained corpus, the Location identified in S21 is structurally modified in the text; S23. Combining the entity recognition result in step S21 and the semantic role recognition result in step S22, the relationship between entities is identified through rule setting.

7. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 6 is characterized in that: The specific operation of the structural change in step S22 is: Change "Location" to "At Location," 8. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 6 is characterized in that: In step S23, the semantic role recognition results in step S22 only retain the original annotation results of ARG0, ARG1, PRED, and ARGM-ADV to better reflect the relationship between the subject and the object.

9. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 6 is characterized in that: In step S23, the rules are set as follows: Object / Location / Attribute is parallel to ARG0 annotation as the head entity; Object / Attribute / Value / Location is parallel to ARG1 annotation as the tail entity; ARGM-ADV is parallel to the Introduction annotation as an adverb, which is the limiting state of the predicate relationship and forms a relationship with the predicate; PRED and Interaction / Compare are marked in parallel as corresponding head entity and tail entity relationships; The relationship between Object and the corresponding Attribute is HasAttribute; The relationship between Object and the corresponding Operator is HasOperate; The corresponding relationship between Situation and Rule is represented by the rule number, and the specific relationship is HasRule; The relationship between Situation / Rule and the corresponding inspection object and entity is Contains.

10. The method for autonomous interpretation of ship navigation rules considering scene characteristics according to claim 1, characterized in that: In step S3, the logical hierarchy of the ship navigation rules knowledge graph constructed is: navigation scenario → navigation inspection elements → navigation attribute requirements and operating specifications.

Images