A method for constructing a reasoning ability measurement system for goal-oriented action common sense

Abstract

The application discloses a kind of target action common sense-oriented reasoning ability measurement system construction method, specifically as follows: corresponding common sense representation method is designed for target action common sense;Target action common sense library is constructed;Summarize the reasoning ability that should be possessed for target action common sense, and the organization structure of the reasoning ability for target action common sense is initially constructed;Different reasoning abilities are respectively constructed for different measurement indexes, and the reasoning ability measurement system for target action common sense is formed.The scheme is first for the target action common sense existing in target action event, corresponding common sense representation method is constructed, and corresponding common sense library is constructed according to the method, then the reasoning ability item that reasoning engine should have is summarized, and the relationship between each reasoning ability item is analyzed.Finally, for each reasoning ability, corresponding measurement index and calculation method are respectively designed, and the reasoning ability measurement system for target action common sense is formed, to provide the basis for evaluating reasoning engine for target action common sense reasoning.

Description

Technical Field

[0001] This invention belongs to the field of knowledge quality measurement, and in particular relates to a method for constructing a reasoning ability measurement system for goal-oriented common sense actions. Background Technology

[0002] The representation and reasoning of common sense knowledge is currently recognized as a core area of ​​artificial intelligence research. The representation and reasoning of common sense knowledge has become a common core area of ​​intelligence in most military information systems. Conducting research on the measurement and testing of the representation and reasoning of common sense knowledge is one of the main core contents of artificial intelligence evaluation.

[0003] (1) Artificial intelligence evaluation has become a requirement for the current implementation of intelligent military information systems. In developed countries such as the United States, related work is being carried out rapidly, but there are no mature common sense representation, reasoning measurement and testing methods yet.

[0004] (2) Knowledge bases and inference engines are unique components in most intelligent system software. They are very different from general software and databases. As a result, the measurement and testing methods of general software and data quality cannot be directly applied to the evaluation of knowledge bases and inference engines. Targeted research on measurement and testing techniques is required.

[0005] (3) At present, entity-relationship databases represented by knowledge graphs are fact-based knowledge bases. Testing of knowledge graphs often adopts existing data (database) quality evaluation methods, which cannot cover more complex common sense measurements and tests such as more abstract causal relationships, event associations, and physical attribute changes.

[0006] (4) The measurement and testing of major recognition-type knowledge representation methods, including (deep) neural networks, and their reasoning performance have been extensively studied. However, the relevant research often focuses on single-item recognition knowledge areas such as image processing, speech recognition, and entity relationship recognition in natural language. It cannot cover the measurement and testing of more complex common sense such as abstract causal relationships and event associations. The interpretability of the test results is weak, and the requirements for large-scale industrial applications are high.

[0007] (5) At present, common sense model testing with common sense learning as the main goal is usually based on a large amount of data. The main quality metrics are general machine learning evaluation methods such as recall and accuracy. The test data collection does not take into account the multi-level abstraction, anomalies, spatiotemporal changes, causal relationships and other characteristics of common sense. It lacks the measurement and testing of the reasoning effectiveness of the learned common sense in practical applications.

[0008] (6) In the application field, industry practitioners are the main users of intelligent information systems, especially intelligent auxiliary systems, and are the main participants in the formulation and evaluation of intelligent requirements. Their industry common sense is often irreplaceable and difficult to obtain. However, practitioners often lack the enthusiasm and energy to evaluate common sense reasoning in intelligent systems. At present, there is a lack of common sense reasoning measurement and testing methods to improve human participation. Summary of the Invention

[0009] The purpose of this invention is to provide a method for constructing a reasoning ability measurement system based on common sense of target actions. This method takes the task of measuring reasoning ability based on common sense of target actions as its core, providing a basis for evaluating the reasoning engine for common sense reasoning of target actions.

[0010] To achieve the above objectives, the technical solution of the present invention is as follows: a method for constructing a reasoning ability measurement system based on common sense about target actions. Figure 1 This demonstrates the overall process of the present invention;

[0011] For step S1, based on the common sense about target actions that appears in the target action event, a corresponding representation method for the common sense about target actions is formed, as follows:

[0012] Step S11: Design the corresponding formal definition of common sense regarding target actions. This definition is based on ontology and action theory, using OWL as the ontological framework. The formal definition is as follows:

[0013]

[0014] Here, Class represents a class, used to represent a type of target action common sense; ObjectProperty represents an object property, used to represent the property relationship between Classes, its domain and range are both Class, such as the subclass relationship Subclass_Of; DatatypeProperty represents a data property, used to represent the property of the Class itself, its domain is Class, and its range is DataType data type, such as the status property status; FluentProperty represents a flow property, used to indicate whether a Class, a DatatypeProperty, or an ObjectProperty is a dynamic flow property or a static flow property, where dynamic flow property is dynamic and static flow property is static; DataType represents a data type, used to indicate the type of the value range of a DatatypeProperty.

[0015] Step S12: Based on the formal definition of common sense regarding target actions above, a common sense ontology for target actions is constructed. Specifically, common sense regarding target actions is mainly divided into three parts: first, the conceptual category ontology; second, the non-monotonic common sense ontology; and third, the spatiotemporal change common sense ontology. The conceptual category ontology is similar to the schema layer construction in knowledge graph ontology construction, but there are also differences. It not only includes the schema layer but also common sense related to classes and attributes in common sense regarding target actions. The non-monotonic common sense ontology includes common sense that needs to be described using non-monotonic reasoning. The spatiotemporal change common sense ontology includes common sense regarding target actions related to the spatiotemporal dynamic domain.

[0016] For step S2, a target action common sense base is constructed, which mainly includes:

[0017] Step S21: Construct a target action common sense base. Based on the target action common sense ontology designed in Step S1, organize relevant specific target action common sense and construct it into a target action common sense base. The target action common sense base mainly includes conceptual category common sense, non-monotonic common sense, and spatiotemporal change common sense. Conceptual category common sense mainly includes three categories of concepts and their attributes: target activity common sense (e.g., takeoff), key target common sense (e.g., aircraft), and location common sense (e.g., ports). Non-monotonic common sense mainly includes common sense that cannot be described by monotonic reasoning and needs to be described using non-monotonic reasoning, such as default and anomalies. Spatiotemporal change common sense mainly includes general spatiotemporal common sense (e.g., a key target can only exist in one location at a time) and common sense specific to target actions (e.g., an aircraft's altitude is 0 when stationary).

[0018] For step S3, using the target action common sense base constructed in step S2, the reasoning ability based on target action common sense is summarized, and the organizational structure of the reasoning ability based on target action common sense is initially constructed. The main steps include:

[0019] Step S31: Summarize the reasoning ability of common sense for target actions. By using the common sense base for target actions constructed in step S2, a total of 11 reasoning abilities can be summarized, including conceptual category reasoning ability, attribute category reasoning ability, relation category reasoning ability, relation property reasoning ability, flow attribute reasoning ability, knowledge consistency judgment reasoning ability, default reasoning ability, abnormal reasoning ability, and the interpretability, performance, and correctness of reasoning. Among them, eight reasoning abilities, namely conceptual category reasoning ability, attribute category reasoning ability, relation category reasoning ability, relation property reasoning ability, flow attribute reasoning ability, knowledge consistency judgment reasoning ability, default reasoning ability, and abnormal reasoning ability, belong to monotonic and non-monotonic reasoning abilities. The three reasoning abilities of interpretability, performance, and correctness belong to the quality of reasoning information. Both monotonic and non-monotonic reasoning abilities and the quality of reasoning information belong to the reasoning ability of common sense for target actions. Thus, the organizational structure of reasoning ability for common sense for target actions is constructed.

[0020] For step S4, using the reasoning ability based on common sense for goal-oriented actions summarized in step S3, measurement indicators are constructed for different reasoning abilities, forming a measurement system for reasoning ability based on common sense for goal-oriented actions. The main steps include:

[0021] Step S41: Provide specific definitions for different reasoning abilities;

[0022] Step S42: Design reasoning ability measurement indicators for the reasoning ability defined in step S41.

[0023] Step S43: Design a calculation method for each reasoning ability metric.

[0024] Compared with the prior art, the advantages of the present invention are as follows: The technology utilizes the common sense related to target action events, summarizes and organizes reasoning ability based on common sense regarding target action, and establishes a reasoning ability measurement system based on common sense regarding target action. This enables the reasoning engine that reasones about target action events to meet the reasoning requirements of target action, which not only helps the reasoning engine to conduct self-evaluation of its reasoning ability, but also improves the reasoning ability of the reasoning engine through the evaluation results, thereby further improving the quality of target action reasoning and enhancing the effect of target action reasoning on situational awareness and decision support. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall process of the present invention;

[0026] Figure 2 This invention summarizes common sense about target actions and designs a corresponding flowchart for representing common sense.

[0027] Figure 3 This is a flowchart illustrating the construction of the target action common sense base of the present invention;

[0028] Figure 4 To summarize the reasoning ability of target-oriented action common sense in this invention, a preliminary organizational structure flowchart of the reasoning ability of target-oriented action common sense is constructed.

[0029] Figure 5 The present invention constructs measurement indicators for different reasoning abilities, forming a flowchart of a reasoning ability measurement system oriented towards common sense of target action. Detailed Implementation

[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0031] Example 1: See Figures 1 to 5 A method for constructing a reasoning ability measurement system based on common sense about target actions. Figure 1 The overall process of this invention is demonstrated.

[0032] To achieve the above objectives, the present invention adopts the following technical solution: a method for constructing a reasoning ability measurement system based on common sense of target action, the method comprising the following steps:

[0033] S1: Design a common sense representation method for the target action common sense;

[0034] S2: Construct a common-sense database for target actions;

[0035] S3: Summarize the reasoning ability based on common sense for goal-oriented actions, and initially construct the organizational structure of the reasoning ability based on common sense for goal-oriented actions;

[0036] S4: Construct measurement indicators for different reasoning abilities to form a reasoning ability measurement system oriented towards common sense for target actions.

[0037] In step S1, a corresponding target action common sense representation method is formed for the target action common sense that appears in the target action event.

[0038] Step S2: Based on the target action common sense ontology designed in Step S1, organize the relevant specific target action common sense and build it into a target action common sense library.

[0039] Step S3: Using the target action common sense base constructed in step S2, summarize the reasoning ability oriented towards target action common sense, and initially construct the organizational structure of the reasoning ability oriented towards target action common sense.

[0040] Step S4: Using the reasoning ability of common sense for goal-oriented actions summarized in Step S3, construct measurement indicators for different reasoning abilities to form a measurement system for reasoning ability of common sense for goal-oriented actions.

[0041] In step S1, in this embodiment of the invention, the process is as follows: Figure 2 As shown, it includes:

[0042] Step S11: Design the corresponding formal definition of common sense regarding target actions. This definition is based on ontology and action theory, using OWL as the ontological framework. The formal definition is as follows:

[0043]

[0044] Here, Class represents a class, used to represent a type of target action common sense; ObjectProperty represents an object property, used to represent the property relationship between Classes, its domain and range are both Class, such as the subclass relationship Subclass_Of; DatatypeProperty represents a data property, used to represent the property of the Class itself, its domain is Class, and its range is DataType data type, such as the status property status; FluentProperty represents a flow property, used to indicate whether a Class, a DatatypeProperty, or an ObjectProperty is a dynamic flow property or a static flow property, where dynamic flow property is dynamic and static flow property is static; DataType represents a data type, used to indicate the type of the value range of a DatatypeProperty.

[0045] Step S12: Based on the formal definition of common sense regarding target actions above, a common sense ontology for target actions is constructed. Specifically, common sense regarding target actions is mainly divided into three parts: first, the conceptual category ontology; second, the non-monotonic common sense ontology; and third, the spatiotemporal change common sense ontology. The conceptual category ontology is similar to the schema layer construction in knowledge graph ontology construction, but there are also differences. It not only includes the schema layer but also common sense related to classes and attributes in common sense regarding target actions. The non-monotonic common sense ontology includes common sense that needs to be described using non-monotonic reasoning. The spatiotemporal change common sense ontology includes common sense regarding target actions related to the spatiotemporal dynamic domain.

[0046] Step S2, in this embodiment of the invention, the process is as follows: Figure 3 As shown, it includes:

[0047] Step S21: Construct a target action common sense base. Based on the target action common sense ontology designed in Step S1, organize relevant specific target action common sense and construct it into a target action common sense base. The target action common sense base mainly includes conceptual category common sense, non-monotonic common sense, and spatiotemporal change common sense. Conceptual category common sense mainly includes three categories of concepts and their attributes: target activity common sense (e.g., takeoff), key target common sense (e.g., aircraft), and location common sense (e.g., ports). Non-monotonic common sense mainly includes common sense that cannot be described by monotonic reasoning and needs to be described using non-monotonic reasoning, such as default and anomalies. Spatiotemporal change common sense mainly includes general spatiotemporal common sense (e.g., a key target can only exist in one location at a time) and common sense specific to target actions (e.g., an aircraft's altitude is 0 when stationary).

[0048] Step S3, in this embodiment of the invention, the process is as follows: Figure 4 As shown, it includes:

[0049] Step S31: Summarize the reasoning ability of common sense for target actions. By using the common sense base for target actions constructed in step S2, a total of 11 reasoning abilities can be summarized, including conceptual category reasoning ability, attribute category reasoning ability, relation category reasoning ability, relation property reasoning ability, flow attribute reasoning ability, knowledge consistency judgment reasoning ability, default reasoning ability, abnormal reasoning ability, and the interpretability, performance, and correctness of reasoning. Among them, eight reasoning abilities, namely conceptual category reasoning ability, attribute category reasoning ability, relation category reasoning ability, relation property reasoning ability, flow attribute reasoning ability, knowledge consistency judgment reasoning ability, default reasoning ability, and abnormal reasoning ability, belong to monotonic and non-monotonic reasoning abilities. The three reasoning abilities of interpretability, performance, and correctness belong to the quality of reasoning information. Both monotonic and non-monotonic reasoning abilities and the quality of reasoning information belong to the reasoning ability of common sense for target actions. Thus, the organizational structure of reasoning ability for common sense for target actions is constructed.

[0050] Step S4, in this embodiment of the invention, the process is as follows: Figure 5 As shown, it includes:

[0051] Step S41: Provide specific definitions for different reasoning abilities;

[0052] (1) Conceptual category reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the category relationship within a class and an instance, as well as the category relationship between the two, so as to demonstrate the conceptual category reasoning ability of the reasoning engine.

[0053] (2) Attribute category reasoning ability, that is, the reasoning engine needs to describe and reason about the category relationship within the data attribute and the category relationship between the data attribute and the class, thereby demonstrating the attribute category reasoning ability of the reasoning engine.

[0054] (3) Relational category reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the category relationship within the object attribute and the category relationship between the object attribute and the class, thereby demonstrating the relational category reasoning ability of the reasoning engine.

[0055] (4) Relational reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the properties of the object itself and the relationships that the properties need to satisfy, thereby demonstrating the relational reasoning ability of the reasoning engine.

[0056] (5) Flow attribute reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the properties of the flow attribute itself and the relationships that the properties need to satisfy, thereby demonstrating the flow attribute reasoning ability of the reasoning engine.

[0057] (6) Knowledge consistency reasoning ability, that is, the reasoning engine needs to describe and reason based on the common sense given by the specific common sense base and the rules of the specific domain, detect and locate the inconsistencies, thereby demonstrating the knowledge consistency judgment reasoning ability of the reasoning engine.

[0058] (7) Default reasoning ability, that is, the reasoning engine needs to be able to correctly process the default rules and reason the default results based on the provided default knowledge, thereby demonstrating the default reasoning ability of the reasoning engine.

[0059] (8) Abnormal reasoning ability, that is, the reasoning engine needs to use abnormal knowledge, and under the condition of having the default reasoning ability, reason about the abnormal knowledge and give the correct result, thereby demonstrating the abnormal reasoning ability of the reasoning engine.

[0060] (9) The interpretability of reasoning requires the reasoning engine to explain the reasoning result or reasoning process, thereby making the reasoning process and result more reliable. The interpretability of reasoning is an important indicator of the reliability of the reasoning engine, the reliability of the reasoning result, and the readability of the reasoning process.

[0061] (10) Performance is the reflection of the time and space complexity of the inference algorithm used by the inference engine. The inference engine needs to provide performance feedback for each user request so that users can better understand the performance effect of the inference engine and thus provide a basis for users to choose a more suitable inference engine for their research.

[0062] (11) Correctness is an important indicator of the reliability of an inference engine. Whether the inference result is correct is directly related to the rationality of the inference algorithm within the inference engine, the correctness of the inference engine's rule understanding, and the correctness of the inference engine's base-mapping. Correctness performance most intuitively demonstrates the inference effect of the inference engine.

[0063] Step S42: Design reasoning ability measurement indicators for the reasoning ability defined in step S41.

[0064] (1) The metrics for conceptual category reasoning ability include the ability to describe classes (DCA), the ability to describe subclass relationships (DSA), the ability to describe equal class relationships (DEA), the ability to describe mutual class relationships (DMA), the ability to describe instances (DIA), the ability to describe instances belonging to classes (DICA), and the overall metric CCAS (conceptual category ability score).

[0065] (2) The metrics for attribute category reasoning ability include the ability to describe attributes (DDTA), the ability to describe the domain of attributes (DDTDA), the ability to describe the range of attributes (DDTRA), and the overall metric (DCAS).

[0066] (3) The metrics for relation category reasoning ability include the ability to describe relations (DRA), the ability to describe the domain of relations (DRDA), the ability to describe the range of relations (DRRA), and the overall metric OCAS (object property category ability score).

[0067] (4) The metrics for relational reasoning ability include the ability to describe relation type (DRTA), the ability to describe relation transitivity (DRTSA), the ability to describe relation symmetry (DRSMA), the ability to describe relation inverse (DRIA), the ability to describe relation function (DRFA), the ability to describe relation inverse function (DRIFA), the ability to describe relation self-inverse function (DRSIA), and the overall metric OCAS (object property category ability score).

[0068] (5) The metrics for fluent reasoning ability include the ability to describe fluent attributes (DFA).

[0069] (6) The metrics for knowledge consistency reasoning ability include the ability to determine knowledge consistency (DKCA).

[0070] (7) The metrics for default reasoning ability include the ability to describe defaults (DDA).

[0071] (8) The metrics for anomaly reasoning ability include the ability to describe anomalies (DAA).

[0072] (9) The interpretability of reasoning includes the ability to explain the reasoning process (DEA).

[0073] (10) Performance metrics include the ability to describe performance (time, memory) DPA (describing performance ability).

[0074] (11) The correctness metrics include the reasoning correctness and the requirement fulfilled by the reasoning result (RCA). Step S43: For each reasoning ability metric, design a metric calculation method;

[0075] (1) Conceptual category reasoning ability:

[0076] ① Possesses the ability to describe classes (DCA).

[0077]

[0078] ② It has the ability to describe subclass relationships (DSA).

[0079]

[0080] ③ It has the ability to describe equal class relationships (DEA).

[0081]

[0082] ④ Possesses the ability to describe mutual class relationships (DMA).

[0083]

[0084] ⑤ It has the ability to describe instances (DIA).

[0085]

[0086] ⑥ Possesses the ability to describe the class to which an instance belongs (DICA).

[0087]

[0088] ⑦ Overall indicator CCAS (Conceptual Category Ability Score)

[0089]

[0090] (2) Attribute category reasoning ability:

[0091] ① Possesses the ability to describe attributes (DDTA).

[0092]

[0093] ② It has the ability to describe the domain of an attribute (DDTDA).

[0094]

[0095] ③ It has the ability to describe the range of attribute values ​​(DDTRA).

[0096]

[0097] ④ Overall indicator DCAS (datatypeproperty category ability score)

[0098]

[0099] (3) Reasoning ability in relational categories:

[0100] ① Possesses the ability to describe relationships (DRA).

[0101]

[0102] ② Possesses the ability to describe the domain of a relation (DRDA).

[0103]

[0104] ③ It has the ability to describe the range of relation values ​​(DRRA).

[0105]

[0106] ④ Overall indicator OCAS (object property category ability score)

[0107]

[0108] (4) Reasoning ability based on relational properties:

[0109] ① Possesses the ability to describe relation types (DRTA).

[0110]

[0111] ② It has the ability to describe relation transitivity (DRTSA).

[0112]

[0113] ③ It has the ability to describe both symmetric and asymmetric relationships (DRSMA).

[0114]

[0115] ④ It has the ability to describe the inverse property of a relation (DRIA).

[0116]

[0117] ⑤ Possesses the ability to describe the functional nature of relations (DRFA)

[0118]

[0119] ⑥ It has the ability to describe the inverse functionality of relations (DRIFA).

[0120]

[0121] ⑦ Possesses the ability to describe the reflexivity and non-reflexivity of relations (DRSIA).

[0122]

[0123] ⑧ Overall index OPAS (object property ability score)

[0124]

[0125] (5) Flow-attribute reasoning ability:

[0126] ① It has the ability to describe fluent properties (DFA).

[0127]

[0128] (6) Knowledge consistency reasoning ability:

[0129] ① Possesses the ability to determine knowledge consistency (DKCA)

[0130]

[0131] (7) Default reasoning ability:

[0132] ① It has the ability to describe default values ​​(DDA).

[0133]

[0134] (8) Abnormal reasoning ability:

[0135] ① It possesses the ability to describe anomalies (DAA).

[0136]

[0137] (9) Interpretability of reasoning:

[0138] ① Possesses the ability to explain the reasoning process (DEA).

[0139]

[0140] (10) Performance:

[0141] ① It has the ability to describe performance (time, memory) using the Descriptive Performance Ability (DPA).

[0142]

[0143] (11) Correctness:

[0144] ① The reasoning result is correct and meets the requirements (RCA).

[0145]

[0146] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0147] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the general knowledge content, objectives, target actions, and other specific technologies of the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for constructing a reasoning ability measurement system based on common sense for goal-oriented actions, characterized in that, The method includes the following steps: S1: Design a common sense representation method for the target action common sense; S2: Construct a common-sense database for target actions; S3: Summarize the reasoning ability based on common sense for goal-oriented actions, and initially construct the organizational structure for the reasoning ability based on common sense for goal-oriented actions; S4: Construct measurement indicators for different reasoning abilities to form a reasoning ability measurement system oriented towards common sense for target actions; In step S1, a corresponding target action common sense representation method is formed for the target action common sense that appears in the target action event, as follows: Step S11: Design the corresponding formal definition of common sense regarding target actions. This definition is based on ontology and action theory, using OWL as the ontological framework. The formal definition is as follows: In this context, `Class` represents a class, used to represent a type of target action common sense; `ObjectProperty` represents an object property, used to represent the attribute relationship between classes, its domain and range are both the `Class` class, including the subclass relationship `Subclass_Of`; `DatatypeProperty` represents a data property, used to represent the property of the `Class` itself, its domain is the `Class` class, and its range is the `DataType` data type, including the status property; `FluentProperty` represents a flow property, used to indicate whether a `Class` class, a `DatatypeProperty`, or an `ObjectProperty` is a dynamic flow property or a static flow property, where dynamic flow properties are `dynamic` and static flow properties are `static`; `DataType` represents a data type, used to indicate the type of the value range of a `DatatypeProperty`. Step S12: Based on the formal definition of the common sense of target action above, the ontology of common sense of target action is constructed. Specifically, the common sense of target action is divided into three parts: the first is the conceptual category ontology of common sense of target action, the second is the non-monotonic common sense ontology of common sense of target action, and the third is the spatiotemporal change common sense ontology of common sense of target action. Step S2, construct a target action common sense base, including: Step S21: Construct a target action common sense base. Based on the target action common sense ontology designed in Step S1, organize relevant specific target action common sense and construct it into a target action common sense base. The target action common sense base includes conceptual category common sense, non-monotonic common sense, and spatiotemporal change common sense. Conceptual category common sense includes three categories of concepts and their attributes: target activity common sense, key target common sense, and location common sense. Non-monotonic common sense includes common sense that cannot be described by monotonic reasoning and needs to be described by non-monotonic reasoning. Spatiotemporal change common sense includes spatiotemporal common sense and common sense unique to target actions. In step S3, using the target action common sense base constructed in step S2, the reasoning ability based on target action common sense is summarized, and the organizational structure of the reasoning ability based on target action common sense is initially constructed. The steps include: Step S31: Summarize the reasoning capabilities for target actions. Using the target action common sense base constructed in step S2, summarize 11 reasoning capabilities: concept category reasoning capability, attribute category reasoning capability, relation category reasoning capability, relation property reasoning capability, flow attribute reasoning capability, knowledge consistency judgment reasoning capability, default reasoning capability, abnormal reasoning capability, and reasoning interpretability, performance, and correctness. Among them, concept category reasoning capability, attribute category reasoning capability, relation category reasoning capability, relation property reasoning capability, flow attribute reasoning capability, knowledge consistency judgment reasoning capability, default reasoning capability, and abnormal reasoning capability belong to monotonic and non-monotonic reasoning capabilities. Reasoning interpretability, performance, and correctness belong to reasoning information quality. Monotonic and non-monotonic reasoning capabilities and reasoning information quality both belong to target action common sense reasoning capabilities. Thus, construct the organizational structure of reasoning capabilities for target action common sense. In step S4, the reasoning ability based on common sense regarding target actions summarized in step S3 is used to construct measurement indicators for different reasoning abilities, forming a reasoning ability measurement system oriented towards common sense regarding target actions. The steps include: Step S41: Provide specific definitions for different reasoning abilities; Step S42: Design reasoning ability measurement indicators for the reasoning ability defined in step S41. Step S43: Design a calculation method for each reasoning ability metric.

2. The method for constructing a reasoning ability measurement system based on common sense for target-oriented actions according to claim 1, characterized in that, Step S41: Specific definitions are given for different reasoning abilities, as follows: (1) Conceptual category reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the category relationships within classes and instances, as well as the category relationships between them, thereby demonstrating the conceptual category reasoning ability of the reasoning engine. (2) Attribute category reasoning ability, that is, the reasoning engine needs to describe and reason about the category relationships within data attributes and the category relationships between data attributes and classes, thereby demonstrating the attribute category reasoning ability of the reasoning engine. (3) Relational category reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the category relationships within object attributes and the category relationships between object attributes and classes, thereby demonstrating the relational category reasoning ability of the reasoning engine. (4) Relational reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the properties of the object itself and the relationships that the properties need to satisfy, thereby demonstrating the relational reasoning ability of the reasoning engine. (5) Flow attribute reasoning ability, that is, the reasoning engine needs to be able to describe and reason about the properties of the flow attribute itself and the relationships that the properties need to satisfy, thereby demonstrating the flow attribute reasoning ability of the reasoning engine. (6) Knowledge consistency reasoning ability, that is, the reasoning engine needs to describe and reason based on the common sense given by the specific common sense base and the rules of the specific domain, detect and locate the inconsistencies, thereby demonstrating the knowledge consistency judgment reasoning ability of the reasoning engine. (7) Default reasoning ability, that is, the reasoning engine needs to be able to correctly process the default rules and reason the default results based on the provided default knowledge, thereby demonstrating the default reasoning ability of the reasoning engine. (8) Abnormal reasoning ability, that is, the reasoning engine needs to utilize abnormal knowledge, and under the premise of having the default reasoning ability, reason about the abnormal knowledge and give the correct result, thereby demonstrating the abnormal reasoning ability of the reasoning engine. (9) The interpretability of reasoning requires the reasoning engine to explain the reasoning result or reasoning process, so as to make the reasoning process and result more reliable. The interpretability of reasoning is an important indicator of the reliability of the reasoning engine, the reliability of the reasoning result, and the readability of the reasoning process. (10) Performance is a reflection of the time and space complexity of the inference algorithm used by the inference engine. The inference engine needs to provide performance feedback for each user request so that users can better understand the performance effect of the inference engine and thus provide a basis for users to choose a more suitable inference engine for their research. (11) Correctness is an important indicator of whether the reasoning engine is reliable. Whether the reasoning result is correct is directly related to whether the reasoning algorithm in the reasoning engine is reasonable, whether the reasoning engine understands the rules correctly, and whether the reasoning engine bases the data correctly. Correctness performance most intuitively shows the reasoning effect of the reasoning engine.

3. The method for constructing a reasoning ability measurement system based on common sense for target-oriented actions according to claim 2, characterized in that, Step S42: Design reasoning ability measurement indicators for the reasoning ability defined in step S41, as follows: (1) The metrics for conceptual category reasoning ability include the ability to describe classes (DCA), the ability to describe subclass relationships (DSA), the ability to describe equal class relationships (DEA), the ability to describe mutual class relationships (DMA), the ability to describe instances (DIA), the ability to describe instances belonging to classes (DICA), and the overall metric CCAS (conceptual category ability score). (2) The metrics for attribute category reasoning ability include the ability to describe attributes (DDTA), the ability to describe the domain of attributes (DDTDA), the ability to describe the range of attributes (DDTRA), and the overall metric (DCAS). (3) The metrics for relational category reasoning ability include the ability to describe relations (DRA), the ability to describe the domain of relations (DRDA), the ability to describe the range of relations (DRRA), and the overall metric OCAS (object property category ability score). (4) The metrics for relational reasoning ability include the ability to describe relation type (DRTA), the ability to describe relation transitivity (DRTSA), the ability to describe relation symmetry (DRSMA), the ability to describe relation inverse (DRIA), the ability to describe relation functionability (DRFA), the ability to describe relation inverse functionability (DRIFA), the ability to describe relation self-inverse functionability (DRSIA), and the overall metric OCAS (object property category ability score). (5) Metrics for fluent attribute reasoning include the ability to describe fluent attributes (DFA). (6) The metrics for knowledge consistency reasoning ability include the ability to determine knowledge consistency (DKCA). (7) Metrics for describing default reasoning ability include the ability to describe defaults (DDA). (8) Metrics for anomaly reasoning ability include the ability to describe anomalies (DAA). (9) Measures of interpretability of reasoning include the ability to explain the reasoning process (DEA). (10) Performance metrics include the ability to describe performance-related time and memory usage (DPA). (11) The metrics for correctness include the reasoning correctness and the requirement met (RCA).

4. The method for constructing a reasoning ability measurement system based on common sense for target-oriented actions according to claim 3, characterized in that, Step S43: For each reasoning ability metric, design a calculation method for the metric, as follows: (1) Conceptual category reasoning ability: It possesses the ability to describe classes (DCA). It possesses the ability to describe subclass relationships (DSA). DEA (Describing EqualClass Ability) has the ability to describe equivalence class relationships. It has the ability to describe mutual class relationships (DMA). DIA (Describing Instance Ability) has the ability to describe instances. It possesses the ability to describe the class to which an instance belongs (DICA). Overall indicator CCAS (conceptual category ability score) (2) Attribute category reasoning ability: It has the ability to describe attributes (DDTA). It has the ability to describe the domain of a data type property (DDTDA). DDTRA (Describing Data Type Property Range Ability) has the ability to describe the range of property values. Overall indicator DCAS (datatypeproperty category ability score) (3) Reasoning ability in relational categories: The ability to describe relationships (DRA) It possesses the ability to describe the domain of a relation (DRDA). It possesses the ability to describe the range of relational values ​​(DRRA). Overall indicator OCAS (Object Property Category Ability Score) (4) Reasoning ability based on relational properties: It possesses the ability to describe relation types (DRTA). It possesses the ability to describe relation transitivity (DRTSA). DRSMA (Describing Relation Symmetry Ability) has the ability to describe both symmetric and asymmetric relationships. It possesses the ability to describe the inverse property of a relation (DRIA). DRFA (Describing Relation Function Ability) possesses the ability to describe the functional nature of relations. DRIFA (describing relation inverse functionability) has the ability to describe the inverse functionability of relations. It possesses the ability to describe the reflexivity and non-reflexivity of relations (DRSIA). Overall metric OPAS (object property ability score) (5) Flow attribute reasoning ability: It possesses the ability to describe fluent properties (DFA). (6) Knowledge consistency reasoning ability: The ability to determine knowledge consistency (DKCA) (7) Default reasoning ability: It has the ability to describe default values ​​(DDA). (8) Abnormal reasoning ability: It possesses the ability to describe anomalies (DAA). (9) Interpretability of reasoning: The ability to explain reasoning processes (DEA) (10) Performance: It possesses the ability to describe time and memory performance (DPA). (11) Correctness: The reasoning result is correct and meets the requirements (RCA - Reasoning Correct Ability). 。

Images