A description and practice method of inferable and precise semantics based on a knowledge graph

Abstract

The present invention discloses a description and practice method for inferable and precise semantics based on a knowledge graph, which relates to the field of intelligent education technology. In the present invention, a cell description language for the mathematics subject is defined to describe the semantics of questions, and a complete description framework for concepts, individuals, and relationships is defined, which not only includes common binary relationships but also extends to the description of multi - element relationships. At the same time, for multi - element relationships, corresponding Atom extension rules are defined to obtain all atomic information from simple relationship descriptions, providing a basis for knowledge reasoning, eliminating the diversity and ambiguity of natural language, and giving definitions with uniqueness and accuracy. In addition, aiming at the problem of non - uniqueness of semantic information, the normalization rules of the cell language are defined to eliminate ambiguity, solving the problem that the existing methods cannot accurately and unambiguously describe the semantics of questions. On this basis, the present invention can realize the intelligent solution of questions through an inference engine, greatly reducing the teaching cost and tutoring cost of teachers.

Description

Technical Field

[0001] The present invention relates to the field of intelligent education technology, and particularly to a description and practice method for inferable and precise semantics based on a knowledge graph. Background Art

[0002] In the field of education, the question bank search technology is one of the most commonly used tools at present, aiming to help users quickly find questions related to learning, exams, or homework grading. However, the limitation of such technology is that when there is no matching question in the question bank, it can only provide references to similar questions and cannot deeply understand the semantics of the questions or complete reasoning. To achieve truly intelligent problem-solving capabilities, the system needs to be able to understand the internal logic of the questions and perform reasoning like a teacher, which has become a key research direction for artificial intelligence in the field of education.

[0003] Currently, the reasoning technologies based on artificial intelligence and knowledge representation mainly include expert systems and knowledge graphs. Expert systems solve complex problems in specific fields by simulating the knowledge and decision-making processes of domain experts and using pre-written rules, and are widely used in fields such as medical diagnosis, financial analysis, and information retrieval. A knowledge graph is a graphical knowledge organizational structure that can represent entities and their relationships, and based on this, reasoning can be carried out, playing an important role in scenarios such as information retrieval, recommendation systems, and data analysis. However, these technologies still have certain limitations when facing complex, cross-domain, or semantically ambiguous reasoning tasks.

[0004] Although there have been many application results for expert systems and knowledge graphs, they still face many problems in the reasoning process: The construction of expert systems is complex, relying on the collection and collation of a large amount of domain knowledge, and it is difficult to adapt to the reasoning needs of new problems or cross-domains. Although the knowledge graph has powerful knowledge representation capabilities, its update and maintenance costs are high, and it is often incomplete, making it difficult to support certain reasoning tasks. In addition, although the SWRL rule language can handle complex logic and conditional reasoning, it has a strong dependence on rules and performs poorly when facing fuzzy or uncertain information, easily leading to inconsistent reasoning results. Therefore, there is an urgent need for a description and practice method for inferable and precise semantics based on a knowledge graph to solve such problems. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a description and practice method for inferable and precise semantics based on a knowledge graph to solve the problems in the existing solutions that the construction of expert systems is complex, relying on the collection and collation of a large amount of domain knowledge, and it is difficult to adapt to the reasoning needs of new problems or cross-domains. Although the knowledge graph has powerful knowledge representation capabilities, its update and maintenance costs are high, and it is often incomplete, making it difficult to support certain reasoning tasks.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] The present invention provides a description and practice method for inferable and precise semantics based on a knowledge graph, which includes,

[0009] Step S1, define a unified Cell language description. To achieve semantic reasoning, the present invention proposes a special description language - the Cell language. The Cell language is used to convert the natural language information in the question into a structured expression form that can be processed by a computer, and define the Cell language to describe complex semantic relationships.

[0010] Step S2, expand the ternary relationship into a multi - element relationship. Traditional knowledge representation methods usually adopt the form of triples (subject, predicate, object), which are difficult to express the complex semantics in math problems. The present invention expands the triples, allows the relationship between multiple individuals, and defines the multi - element relationship structure.

[0011] Step S3, expand the inference information. Based on the multi - element relationship in Step S2, expand all relevant individual and attribute information. Suppose the relationship "line segment EF intersects line segment AC at point N" is expanded to expand the relevant individual information of angles, points, and lines.

[0012] Step S4, define the data attributes and binary relationships of individuals. For each individual, further define its attributes and relationships with other individuals.

[0013] Step S5, define a unified naming rule. To ensure the accuracy of reasoning, the present invention introduces a normalization operation to eliminate ambiguity. By defining a unified naming rule, the expression of all geometric elements is unique and unambiguous:

[0014] Step S6, handle major angles and minor angles. The angles in math problems usually have two representation forms: major angles and minor angles. To ensure consistency in the reasoning process, the present invention normalizes the angles and uses the same form for description.

[0015] Step S7, correct the line - segment intersection relationship. For the description of line - segment intersection, correct the order of the starting and ending sides of the line segments in the description. For example, when describing "line segment AB intersects line segment BC at point B", it is necessary to ensure that AB is the starting side and BC is the ending side. If it cannot be directly determined by the graph, use a preset rule to correct such relationships to ensure the uniqueness of the expression in the reasoning process.

[0016] Step S8, the inference engine makes inferences. Steps S1 to S7 are semantic description steps, that is, standardizing the cell language and relationships. After completing the semantic description, the inference engine makes inferences (practices) based on the above-mentioned standardized cell language and relationships. The inference engine analyzes the above relationships and uses SWRL rules to infer new conclusions. For example, for geometric problems, specific results such as the size of an angle, the length of a line segment, and the position of an intersection point can be calculated, and finally, the automatic solution of math problems is achieved.

[0017] Further, in step S1, the constituent elements of the defined Cell language include:

[0018] concept, representing abstract concepts in mathematics, such as "triangle", "angle", etc.

[0019] ind, representing specific individuals, such as "point A", "line segment AB", etc.

[0020] data, representing data or attributes related to individuals, such as "length", "size", etc.

[0021] conceptind, representing the relationship between a concept and an individual, such as "triangle ABC contains point A".

[0022] indind, representing the relationship between two individuals, such as "line segment AB is perpendicular to line segment BC".

[0023] inddata, representing the data attribute of an individual, such as "the length of line segment AB is 5".

[0024] Further, in step S2, the defined form of the multi - relationship structure is:

[0025] indind#relationship description[ind1, ind2, …, ind n , where n represents the number of individuals.

[0026] Further, in step S2, the defined multi - relationship structure includes the following multi - relationship interfaces for expressing common relationships in geometry:

[0027] Two line segments intersect at a certain point:

[0028] indind#SeglineCrossSeglineAtPoint[ind1, ind2, ind3],

[0029] Two line segments are perpendicular at a certain point:

[0030] indind#SeglineVerticalSeglineAtPoint[ind1, ind2, ind3],

[0031] The altitude on a certain side of a triangle:

[0032] indind#TriangleHasEdgeHighline[ind1, ind2, ind3],

[0033] The median on a certain side of a triangle:

[0034] indind#TriangleHasEdgeMiddleline[ind1, ind2, ind3],

[0035] The angle bisector of a certain angle of a triangle:

[0036] indind#TriangleHasAngleMiddleline[ind1, ind2, ind3].

[0037] Furthermore, in step S3, the individual information of the extended angles, points, and lines is:

[0038] The individual information of an angle: conceptind#conceptHasInd[Angle, Angle@FNC],

[0039] The information of a point: conceptind#conceptHasInd[Point, N],

[0040] The information of a line: conceptind#conceptHasInd[Segline, EN].

[0041] Furthermore, in step S4, the way to define the data attributes and binary relations of individuals is:

[0042] The inddata relationship, which describes the data attributes of individuals, including the length of a line segment and the size of an angle:

[0043] inddata#hasDefaultLength(EN, EN@Length),

[0044] inddata#hasSize(Angle@FNC, FNC@Size).

[0045] Furthermore, in step S4, the way to define the data attributes and binary relations of individuals also includes:

[0046] The indind relationship, which is used to describe the relationship between two individuals, including the relationship between a line segment and a point, and an angle:

[0047] indind#seglineHasPoint([AC,N])

[0048] indind#angleEqualAngle([Angle@CNE,Angle@ANF])

[0049] Furthermore, in step S5, the naming rules include:

[0050] Line segment naming rule: Line segments are uniformly named in alphabetical order. For example, line segments AB and BA are uniformly named as AB to avoid ambiguity caused by different writing orders.

[0051] Furthermore, in step S5, the naming rules also include:

[0052] Angle naming rule: The naming of an angle is based on the clockwise or counterclockwise order from the starting side to the ending side. For example, Angle@ABC represents the angle formed by rotating counterclockwise from ray AB to ray BC; regardless of whether the problem involves a major angle or a minor angle, it is uniformly represented in the counterclockwise form.

[0053] Furthermore, in step S6, the methods for normalizing angles include:

[0054] A major angle is defined as an angle less than 180 degrees, and a minor angle is defined as an angle greater than 180 degrees but less than 360 degrees.

[0055] When angles within the same defined range have different names (such as Angle@ABC and Angle@CBA), they are uniformly described in the same form.

[0056] The beneficial effects of the present invention are:

[0057] The present invention eliminates the diversity and ambiguity of natural language, gives definitions with uniqueness and accuracy, and on this basis, performs precise semantic understanding and reasoning to achieve the understanding and automatic solution of problems.

[0058] The present invention defines a cell description language for the mathematics discipline to describe the semantic meaning of problems, defines a complete description framework for concepts, individuals, and relationships, which not only includes common binary relationships but also extends to the description of multi - element relationships. At the same time, for multi - element relationships, corresponding Atom extension rules are defined to obtain all atomic information from simple relationship descriptions, providing a basis for knowledge reasoning.

[0059] The present invention defines the normalization rules of the cell language to eliminate ambiguity for the problem of non - uniqueness of semantic information.

[0060] In summary, the present invention has established semantic description rules for the questions in the mathematics discipline. Compared with the existing semantic extraction methods, it is more comprehensive, with a more rigorous description logic and more complex extracted semantics. At the same time, the present invention further standardizes the definition rules of basic individuals and relationships, solving the problem that the existing methods cannot accurately and unambiguously describe the semantics of questions. On this basis, the present invention can realize the intelligent solution of questions through an inference engine, greatly reducing the teaching cost and tutoring cost of teachers. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0062] Figure 1 It is a schematic flow chart of the method for describing and practicing the inferable and accurate semantics based on the knowledge graph of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0064] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0065] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0066] SWRL is a rule language used to define and execute rules. SWRL has rich logical expression capabilities and can represent various rules and logical conditions, including numerical comparisons, string operations, etc. At the same time, SWRL rules can be executed by an inference engine for reasoning, thereby inferring new facts, relationships, or conclusions. The main purpose of SWRL is to support logical reasoning in order to infer new information or conclusions from knowledge. The performance and effectiveness of SWRL reasoning highly depend on the integrity of the rules. If there are missing or incorrect conditions, incorrect reasoning conclusions will occur. On the other hand, SWRL rules are usually constructed based on precise logical conditions and are not very suitable for dealing with uncertain or ambiguous information. For rules with ambiguity, inconsistent reasoning results will occur. Therefore, the following improvement solutions are proposed:

[0067] Example 1, referring to Figure 1 , this example is for the mathematics discipline and aims to design a method that can accurately describe the complex logical semantics in math problems. Although SWRL supports complex logical reasoning, all knowledge needs to be represented in the form of unambiguous Atom atoms.

[0068] This invention mainly solves two problems: converting the text information in the problem into a unified cell language that can be processed by a computer (the semantic description proposed in this invention is defined as cell); converting the cell language into an unambiguous and accurate Atom atom expression form.

[0069] This invention designs a series of cell definitions to describe complex semantic relationships. This invention expands the representation form of triples, enabling cells to describe multi - element relationships. Use concept to represent concepts, ind to represent individuals, data to represent data, conceptind to represent the relationship between concepts and individuals, indind to represent the relationship between individuals, and inddata to represent the data attributes of individuals. In the conventional representation method, the relationship between individuals and the data attributes of individuals are usually represented as binary relationships. For the relevant knowledge of mathematics, this invention defines multi - element relationships on the basis of the original binary relationships in SWRL. Then the general form of the description of the relationship between n individuals can be expressed as:

[0070] indind#relationship description[ind1, ind2,..., indn]

[0071] Since it is difficult to describe all common problem semantics with binary relationships, although multi - element relationships can be split into binary relationships, from the perspectives of simplifying the engineering process, maintainability, etc., the definition of multi - element relationships is crucial. This invention defines a series of corresponding cell interfaces according to the multi - element relationships between mathematical geometric individuals, for example:

[0072] indind#SeglineCrossSeglineAtPoint: [ind1, ind2, ind3]

[0073] indind#SeglineVerticalSeglineAtPoint: [ind1, ind2, ind3]

[0074] indind#TriangleHasEdgeHighline: [ind1, ind2, ind3]

[0075] indind#TriangleHasEdgeMiddleline: [ind1, ind2, ind3]

[0076] indind#TriangleHasAngleMiddleline: [ind1, ind2, ind3]

[0077] respectively represent the semantics of "two line segments intersect at a point", "two line segments are perpendicular to a point", "a triangle has an altitude on a certain side", "a triangle has a median on a certain side", and "a triangle has an angle bisector of a certain angle".

[0078] For ternary relations, we have implemented the extension of relations for specific definitions, that is, expanding all indind or inddata included in the multi - element relation to facilitate providing the most complete inference information input. Taking SeglineCrossSeglineAtPoint[EF, AC, N] as an example, according to the intersection of two line segments EF and AC at point N, the present invention can expand all the information of lines, points, and angles. For example, expanding the specific individual information of all concepts such as Angle, Point, and Segline, such as:

[0079] conceptind#conceptHasInd[Angle, Angle@FNC],

[0080] conceptind#conceptHasInd[Point, N],

[0081] conceptind#conceptHasInd[Segline, EN]

[0082] All the data attributes of individuals, such as:

[0083] inddata#hasDefaultLength(EN, EN@Length),

[0084] inddata#hasSize(Angle@FNC,FNC@Size),

[0085] All individual binary relations, such as:

[0086] indind#seglineHasPoint([AC,N]),

[0087] indind#angleEqualAngle([Angle@CNE,Angle@ANF]),

[0088] Generally speaking, in the design of the cell language, compared with SWRL, the present invention defines new relations in some concepts, individuals, attributes, etc. according to the characteristics of mathematics, defines multi - element relations according to complex semantics, and expands information at the Atom atomic level for multi - element relations.

[0089] In order to perform deeper reasoning to achieve the goal of intelligent problem - solving, after describing natural language as the Cell language, it is necessary to further eliminate the ambiguity and conflicts in the cell description before the correct conclusion can be inferred through the inference engine. For some questions, it is necessary to combine the text information of the question stem and the reference information of the graph to determine the unique semantic information. For example, triangle ABC and triangle BAC represent the same triangle, and angle ABC and angle CBA represent the same angle. However, for a computer, it cannot perform thinking judgments like a human, and existing inference engines cannot perceive image information. Therefore, making the cell description unique is a prerequisite for accurate computer reasoning.

[0090] The present invention designs a naming rule for semantics to perform the normalization operation of semantic description. Specifically, the present invention makes constraints in the most basic individual definitions, such as the naming of line segments and angles. For line segments, the line segment names are uniformly processed in alphabetical order. For example, AB and BA are uniformly named AB.

[0091] An angle is different from a line segment and has its particularity. The particularity of an angle mainly lies in that an angle itself can be a major angle or a minor angle. That is, Angle@ABC may represent the major angle formed by line segment ray AB and ray BC, or it may refer to the minor angle formed by the two rays. In the actual process, whether Angle@ABC described in a problem represents a major angle or a minor angle is usually judged by people in combination with the actual situation. In addition, Angle@ABC and Angle@CBA usually represent the same angle in the actual situation, which also means that the description of the same angle is not unique. A computer can only judge a major angle or a minor angle or different names of an angle based on clear rules. Therefore, the clear and unambiguous description of an angle is more complex. The present invention makes the following definition: In Angle@ABC, with AB as the initial side and BC as the terminal side, at this time Angle@ABC represents the angle formed by the counterclockwise rotation of ray AB to BC. And Angle@CBA represents the angle formed by the counterclockwise rotation of ray BC to AB. Under such a definition, if Angle@ABC is a minor angle, then Angle@CBA is its corresponding major angle. After Angle@ABC and Angle@CBA are normalized, their names are uniformly corrected to the minor angle represented by ray AB and BC, that is, uniformly corrected to Angle@ABC, to conform to the general habit of the problem. In this way, the problem of the standardized description of an angle is completely solved, ensuring the uniqueness of the angle naming.

[0092] For the intersection relationship of line segments, such as indind#SeglineCrossSeglineAtPoint:[AB,BC,B], it is necessary to ensure that AB is the starting side and BC is the terminal side according to the spatial position relationship between AB and BC in the figure. Otherwise, the parameter order of AB and BC needs to be corrected. If there is no figure for reference, then according to the default order rule, all line, angle, and intersection information will be corrected.

[0093] In the present invention, the Cell language is used as a way of knowledge representation, similar to the "entity-relationship-entity" structure in a knowledge graph, that is, a triple. The main purpose of the Cell language is to convert the semantics in natural language into an unambiguous and structured expression.

[0094] The design of multi-relations is also an extension of the knowledge graph. Traditional knowledge graphs usually use triples to describe the binary relationship between entities (such as entity A is associated with entity B). The present invention extends the knowledge representation ability through the form of multi-relations to express the complex logical relationships in math problems. The extension of multi-relations is regarded as a further development for specific application scenarios based on the knowledge graph.

[0095] At the same time, for IndInd and IndData in the text:

[0096] IndInd represents the relationship between two entities (individuals), such as "line segment AB is perpendicular to line segment BC", which is similar to the description of the relationship between entities in a knowledge graph, that is, the manifestation form of the "predicate" in a knowledge graph.

[0097] IndData is used to describe the relationship between an individual and its attributes. For example, "the length of line segment AB is 5". The association between an individual and data is the same as the typical structure in a knowledge graph, that is, the entity information in the knowledge graph is enriched through data attributes (such as length, size, etc.).

[0098] In a knowledge graph, reasoning is usually based on existing triple relationships. For example, new entity relationships can be discovered through reasoning. The Atom atomic representation form of the present invention disassembles complex relationships into more basic units to support deeper reasoning, which can be understood as the mechanism by which a knowledge graph reasoning engine conducts logical inferences through triples.

[0099] The reasoning engine of the present invention conducts reasoning on complex problems through multi - relationship and attribute information represented by the Cell language, and its process is similar to the working principle of a knowledge graph reasoning engine.

[0100] The present invention, using the Cell language, IndInd / IndData relationships, and the reasoning process of the reasoning engine, is similar to the core idea in a knowledge graph, that is, through the graphical representation of entities and relationships, it supports semantic reasoning and knowledge expansion.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A description and practice method of inferable and precise semantics based on a knowledge graph, characterized in that: including Step S1, define a unified Cell language description, and define the complex semantic relationships in the Cell language description Step S2, expand the ternary relationship into a multi - ary relationship, expand the triples, allow relationships between multiple individuals, and define the multi - ary relationship structure Step S3, expand the inference information. Based on the multi - ary relationships in Step S2, expand all relevant individual and attribute information. Suppose the relationship "line segment EF and AC intersect at point N" is expanded to expand the individual information of related angles, points, and lines Step S4, define the data attributes and binary relationships of individuals. For each individual, further define its attributes and relationships with other individuals Step S5, define a unified naming rule Step S6, handle major angles and minor angles, perform normalization processing on angles, and use the same form for description Step S7, correct the line segment intersection relationship. For the description of line segment intersection, correct the order of the starting and ending sides of the line segments in the description Step S8, use an inference engine for reasoning. Steps S1 to S7 are semantic description steps, that is, normalize the cell language and relationships, convert the text information in the question into a unified cell language that can be processed by a computer. After completing the semantic description, the inference engine performs reasoning based on the above - mentioned normalized cell language and relationships. The inference engine parses the above relationships, uses SWRL rules to infer new conclusions, and finally realizes the automatic solution of math problems In Step S1, the constituent elements of the defined Cell language include concept, representing an abstract concept in mathematics ind, representing a specific individual data, representing data or attributes related to an individual conceptind, representing the relationship between a concept and an individual indind, representing the relationship between two individuals inddata, representing the data attributes of an individual In Step S2, the defined multi - ary relationship structure form is indind#Relationship description [ind1, ind2, …, ind n , where n represents the number of individuals; In Step S2, the defined multi - ary relationship structure includes the following multi - ary relationship interfaces for expressing common relationships in geometry Two line segments intersect at a certain point indind#SeglineCrossSeglineAtPoint[ind1, ind2, ind3] Two line segments are perpendicular at a certain point indind#SeglineVerticalSeglineAtPoint[ind1, ind2, ind3] The altitude on a certain side of a triangle indind#TriangleHasEdgeHighline[ind1, ind2, ind3] The median on a certain side of a triangle indind#TriangleHasEdgeMiddleline[ind1, ind2, ind3] The angle bisector of a certain angle of a triangle indind#TriangleHasAngleMiddleline[ind1, ind2, ind3] 2. The description and practice method of inferable and precise semantics based on a knowledge graph according to claim 1, characterized in that, In Step S3, the expanded individual information of angles, points, and lines is Individual information of angles conceptind#conceptHasInd[Angle, Angle@FNC], Information of point: conceptind#conceptHasInd[Point, N], Information of line segment: conceptind#conceptHasInd[Segline, EN].

3. A description and practice method of inferable precise semantics based on a knowledge graph according to claim 2, characterized in that, In step S4, the data attributes and binary relation modes of individuals are defined as: The inddata relation describes the data attributes of individuals, including the length of the line segment and the size of the angle: inddata#hasDefaultLength(EN, EN@Length), inddata#hasSize(Angle@FNC, FNC@Size).

4. A description and practice method of inferable and precise semantics based on a knowledge graph according to claim 3, characterized in that In step S4, the data attributes and binary relation modes of individuals also include: The indind relation is used to describe the relationship between two individuals, including the relationship between the line segment and the point, and the angle: indind#seglineHasPoint([AC, N]), indind#angleEqualAngle([Angle@CNE, Angle@ANF]).

5. A method for describing and practicing inferable and precise semantics based on a knowledge graph according to claim 4, characterized in that, In step S5, the naming rules include: Naming rule for line segment: The line segments are uniformly named in alphabetical order.

6. A method for describing and practicing inferable precise semantics based on a knowledge graph according to claim 5, characterized in that In step S5, the naming rules also include: Naming rule for angle: The angle is named based on the clockwise or counterclockwise order from the starting side to the ending side.

7. A description and practice method for inferable and precise semantics based on a knowledge graph according to claim 5, characterized in that In step S6, the methods for normalizing the angle include: The reflex angle is defined as an angle less than 180 degrees, and the concave angle is defined as an angle greater than 180 degrees but less than 360 degrees. When there are different names for angles within the same defined range, they are uniformly described in the same form.

Images