CN114722211A - Quality evaluation method and device of network optimization knowledge graph and electronic equipment - Google Patents

Quality evaluation method and device of network optimization knowledge graph and electronic equipment Download PDF

Info

Publication number
CN114722211A
CN114722211A CN202210162918.0A CN202210162918A CN114722211A CN 114722211 A CN114722211 A CN 114722211A CN 202210162918 A CN202210162918 A CN 202210162918A CN 114722211 A CN114722211 A CN 114722211A
Authority
CN
China
Prior art keywords
entity
level
confidence
knowledge graph
confidence coefficient
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210162918.0A
Other languages
Chinese (zh)
Inventor
喻鹏
李文璟
丰雷
周凡钦
阎钰洁
方宏林
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN202210162918.0A priority Critical patent/CN114722211A/en
Publication of CN114722211A publication Critical patent/CN114722211A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Animal Behavior & Ethology (AREA)
  • Computational Linguistics (AREA)
  • Databases & Information Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a quality evaluation method and device of a network optimization knowledge graph and electronic equipment, wherein the method comprises the following steps: acquiring a three-element data set of a knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to the entity type in the knowledge graph based on the three-element data set of the knowledge graph; determining the resource amount of a head entity flowing to a tail entity, and obtaining the confidence coefficient of an entity level in a local level; obtaining the confidence coefficient of a relation level in a local level based on the entity in the knowledge graph, the entity relation and the embedded vector representation corresponding to the entity type; and obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level. The quality evaluation method and device for the network optimization knowledge graph and the electronic equipment can accurately and effectively realize the quality evaluation of the knowledge graph.

Description

Quality evaluation method and device of network optimization knowledge graph and electronic equipment
Technical Field
The invention relates to the technical field of knowledge graphs, in particular to a quality evaluation method and device of a network optimization knowledge graph and electronic equipment.
Background
The knowledge graph lays a solid foundation for knowledge interconnection on the world wide web due to strong semantic processing capability and open interconnection capability. In the field of communication networks, network optimization knowledge maps convert various types of network related information (including network optimization problem judgment rules, network optimization scheme formulation rules, network configuration resource information, network operation quality judgment information, and the like) into structured knowledge information. By utilizing the constructed network optimization knowledge graph, knowledge data related to the problem network can be conveniently acquired, a basis is provided for judging the network problem and formulating a network optimization scheme, the labor cost and the optimization time cost are reduced, and the optimization efficiency is improved. However, a potential error in the network optimization knowledge graph may mislead the judgment of the network problem, thereby affecting the optimization efficiency. The prior art scheme can not accurately and effectively evaluate the quality of the knowledge graph.
Disclosure of Invention
The invention provides a quality evaluation method and device of a network optimization knowledge graph and electronic equipment, which are used for accurately and effectively realizing quality evaluation of the knowledge graph.
The invention provides a quality evaluation method of a network optimization knowledge graph, which comprises the following steps:
acquiring a three-element data set of a knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to an entity type in the knowledge graph based on the three-element data set of the knowledge graph;
determining the resource amount of a head entity flowing to a tail entity based on the entities, and obtaining the confidence coefficient of an entity level in a local level based on the resource amount of the head entity flowing to the tail entity;
obtaining the confidence of the relationship level in the local level based on the entity in the knowledge graph, the entity relationship and the embedded vector representation corresponding to the entity type;
determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
According to the quality evaluation method of the provided network optimization knowledge graph, the determining the resource amount of the head entity to the tail entity based on the entities comprises the following steps:
determining a directed subgraph of each entity by taking each entity in the knowledge graph as a central node of a search range and taking a preset value as a search depth;
and determining the resource amount of the head entity flowing to the tail entity based on the directed subgraph.
According to the provided quality evaluation method of the network optimization knowledge graph, the calculation formula for determining the resource amount of the head entity to the tail entity based on the directed subgraph is as follows:
Figure BDA0003515511230000021
wherein Q (t | h) represents the amount of resources flowing from the head entity h to the tail entity t, ItA set of head entities, O (e), representing entities t corresponding to the directed subgraphi) Representing entity e corresponding to the directed subgraphiThe out-of-range of (c) is,
Figure BDA0003515511230000022
representing entity e corresponding to the directed subgraphiAnd the number of entity relations between the directed subgraph and the entity t, wherein N represents the total number of the entities corresponding to the directed subgraph; e represents the probability of the direct jump of the resource flow controlling each entity to a random entity.
According to the provided quality evaluation method of the network optimization knowledge graph, the calculation formula for obtaining the confidence coefficient of the entity level in the local level based on the resource amount of the head entity flowing to the tail entity is as follows:
Figure BDA0003515511230000031
wherein ET (h, r, t) is the confidence of the entity level in the local level, r is the entity relationship, λ is the hyper-parameter for smoothing, λ is greater than or equal to 1, and θ is the preset threshold corresponding to the entity relationship.
According to the provided quality evaluation method of the network optimization knowledge graph, the confidence of the relationship level in the local level is determined based on the entity, the entity relationship and the embedded vector representation corresponding to the entity type in the knowledge graph, and the calculation formula is as follows:
Figure BDA0003515511230000032
wherein RT (h, r, t) is the confidence of the relationship level in the local level.
According to the provided quality evaluation method of the network optimization knowledge graph, the quality evaluation result of the knowledge graph is obtained based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level, and the method comprises the following steps:
and performing weighted calculation on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level based on a preset entity level confidence coefficient weight value, a relation level confidence coefficient weight value and a global level confidence coefficient weight value to obtain a quality evaluation result of the knowledge graph.
The invention also provides a quality evaluation device of the network optimization knowledge graph, which comprises the following steps:
the data processing module is used for acquiring a three-element data set of the knowledge graph and acquiring an entity, an entity relation and an embedded vector representation corresponding to the entity type in the knowledge graph based on the three-element data set of the knowledge graph;
the first confidence coefficient calculation module is used for determining the resource amount of the head entity flowing to the tail entity based on the entity and obtaining the confidence coefficient of the entity level in the local level based on the resource amount of the head entity flowing to the tail entity;
the second confidence coefficient calculation module is used for obtaining the confidence coefficient of a relation level in a local level based on the entity in the knowledge graph, the entity relation and the embedded vector representation corresponding to the entity type;
a third confidence calculation module for determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and the quality evaluation module is used for obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of the network optimization knowledge-graph quality assessment method.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method for quality assessment of a network optimization knowledgegraph as described in any of the above.
The present invention also provides a computer program product comprising a computer program which, when executed by a processor, performs the steps of the method for quality assessment of a network optimization knowledgegraph as described in any of the above.
According to the quality evaluation method and device for the network optimization knowledge graph and the electronic equipment, the entity relation and the embedded vector representation corresponding to the entity type in the knowledge graph are obtained by obtaining the three-element data set of the knowledge graph and based on the three-element data set of the knowledge graph; in the process of learning the entity and the entity type embedded vector, the entity type and the entity embedded vector have a relationship, and when the entity and the entity type are correctly matched, the loss function value corresponding to the relationship between the entity type and the entity embedded vector is small, so that a large error is avoided.
Therefore, the energy function in the local level of confidence calculation is improved by using the entity type embedded vector, and on the basis of considering the traditional triple errors, the matching errors of the entity and the entity type are considered at the same time, so that more accurate confidence evaluation of the knowledge graph is realized.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for evaluating the quality of a network optimization knowledge graph provided by the present invention;
FIG. 2 is a second schematic flow chart of the method for evaluating the quality of a network optimization knowledge-graph according to the present invention;
FIG. 3 is a detailed flow chart of triple data preprocessing provided by the present invention;
FIG. 4 is a detailed flow chart of triple local confidence calculation provided by the present invention;
FIG. 5 is a schematic diagram of a wireless network optimization knowledge-graph provided by the present invention;
FIG. 6 is a schematic structural diagram of a network optimization knowledge-graph quality assessment apparatus provided by the present invention;
fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The following describes the quality evaluation method, apparatus and electronic device of the network optimization knowledge-graph according to the present invention with reference to fig. 1 to 7.
Fig. 1 is a schematic flow chart of a quality evaluation method of a network optimization knowledge graph provided by the present invention, and as shown in fig. 1, the quality evaluation method of a network optimization knowledge graph provided by the present invention includes:
and step 110, acquiring a three-element data set of the knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to the entity type in the knowledge graph based on the three-element data set of the knowledge graph.
It is to be appreciated that the knowledge-graph of the present invention optimizes the knowledge-graph for the wireless network. The knowledge graph is stored in a database in the form of a Resource Description Framework (RDF) dataset. The resource description framework data set is composed of triples, namely a head entity h, an entity relation r and a tail entity t, and can be represented by embedded vectors, specifically (h, r, t). Each triplet represents a real-world fact, and thus the knowledge-graph enables a formal description of real-world facts and their entity relationships.
Wherein the entities comprise a head entity and a tail entity; entity relationships, that is, relationships between head and tail entities. The triple data set contains entities and entity relationships.
To calculate knowledge-graph confidence, data in the form of triples is first obtained from a knowledge-graph database. Further, a special entity relationship "typeIs" can be given to the binary (entity, entity type), and the binary form thereof is converted into a triple form of triple (entity, typeIs, entity type).
And 120, determining the resource amount of the head entity flowing to the tail entity based on the entities, and obtaining the confidence of the entity level in the local level based on the resource amount of the head entity flowing to the tail entity.
It can be appreciated that, with reference to the PageRank algorithm, the amount of resources between each pair of entities flowing from the head entity to the tail entity is obtained, and data is prepared in advance for subsequent calculation of confidence levels of the entity levels in the local levels.
The confidence of the local level triple comprises two levels of an entity level and an entity relationship level, and the calculation is carried out in two steps.
First, the entity hierarchy reflects the confidence with the strength of association between two entities. It can be considered that when a head entity has more resources to flow to a tail entity in a certain entity pair (h, t), the entity pair has stronger association, and the entity relationship is more likely to exist between the entity pair, so that the entity pair should have higher confidence for the triple containing the entity pair. And finally, converting the resource quantity into an entity level confidence coefficient by using a sigmoid function.
And step 130, obtaining the confidence of the relation level in the local level based on the entity in the knowledge graph, the entity relation and the embedded vector representation corresponding to the entity type.
It is understood that the confidence of the relationship level, i.e. the confidence of the relationship level of the head entity and the tail entity.
The entity relationship level confidence coefficient utilizes a translation invariance principle in embedded representation learning to construct an energy function, meanwhile, the accuracy of a triple and the accuracy of entity and entity type matching are considered, and the triple meeting the energy function is considered to have higher confidence coefficient, namely, the higher confidence coefficient is considered to be possessed when the value of the energy function E (h, r, t) is smaller, so that the energy function value is converted into the relationship level confidence coefficient by utilizing the sigmoid function.
Step 140, determining a global level confidence based on multi-step paths between the entity pairs in the knowledge-graph.
It is understood that the global level confidence is the global level confidence of the knowledge-graph triple data set. In the global hierarchy, a multi-step path between two entity pairs is utilized to reflect confidence. The multi-step path between two entity pairs and the direct relationship between two entity pairs should have similar semantic information, i.e. the embedded vector representation of the multi-step path should be a small distance from the embedded vector representation of the direct relationship.
And 150, obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
It is understood that the quality assessment result of the knowledge-graph, i.e. the integrated confidence of the whole knowledge-graph.
In some embodiments, said determining, based on said entities, an amount of resources that a head entity flows to a tail entity comprises:
determining a directed subgraph of each entity by taking each entity in the knowledge graph as a central node of a search range and taking a preset value as a search depth;
and determining the resource amount of the head entity flowing to the tail entity based on the directed subgraph.
It can be understood that, in order to obtain the resource quantity owned by each entity conveniently, the whole knowledge graph is preprocessed, each entity is taken as a center entity, the search depth is limited to be epsilon, and a directed subgraph taking each entity as the center is obtained by a deep search mode.
Fig. 2 is a flowchart of a method for evaluating the quality of a network optimization knowledge-graph in another embodiment provided by the present invention, fig. 3 is a flowchart of triple data preprocessing, and fig. 4 is a flowchart of triple local confidence calculation.
In some embodiments, the calculation formula for determining the resource amount flowing from the head entity to the tail entity based on the directed subgraph is:
Figure BDA0003515511230000081
wherein Q (t | h) represents the amount of resources flowing from the head entity h to the tail entity t, ItA set of head entities, O (e), representing entities t corresponding to the directed subgraphi) Representing the entity e corresponding to the directed subgraphiOut of rotation of WeitRepresenting entity e corresponding to the directed subgraphiAnd the number of entity relations between the directed subgraph and the entity t, wherein N represents the total number of the entities corresponding to the directed subgraph; e is the probability that the resource flow controlling each entity jumps directly to a random entity.
Optionally, e may be set by itself, and the corresponding value range is 0-1, for example, may be 0.15.
To numerically quantify and compute the knowledge-graph, it is first necessary to project the entities and entity relationships in the knowledge-graph into a vector space in order to preserve the structural information of the knowledge-graph while computing.
In order to obtain the embedded vector representation of the entity, the entity relationship and the entity type at the same time, the improvement is carried out on the basis of the traditional embedding algorithm TransE, namely, an energy function part for embedding and representing the entity type is added on the basis of the original energy function, and the new energy function is obtained as follows:
Figure BDA0003515511230000082
the energy function consists of two parts, wherein G (h, r, t) | | h + r-t | | represents an dissimilar fraction under the assumption of a translation invariance principle, the dissimilar fraction is the same as the fraction in TransE, and lower G (h, r, t) represents that a triplet formed by an entity and an entity relation is more in line with the translation invariance principle. On the basis of the original TransE, the entity type energy function is introduced into the invention as a second part of the overall energy function.
The invention projects entities and entity types into different spaces, and utilizes a mapping matrix associated with entity relationship "typeIs" to communicate two projection spaces, namely, e is approximately equal to M tau, and the concept of the principle of translational invariance is imitated, G (e, tau) | | e-M tau | is defined as a distance function, and lower G (e, tau) represents the assumption that e is approximately equal to M tau more.
For learning embedding, based on the energy function, a maximum margin optimization method is adopted, and a first loss function is constructed as follows:
Figure BDA0003515511230000091
the second loss function is as follows:
Figure BDA0003515511230000092
wherein, γ1And gamma2Are two hyper-parameters; h and t respectively represent head and tail entities, and all the entities form an entity set E; r represents entity relationships, and all entity relationships form an entity relationship set R; τ represents an entity type, all entity types constituting a set P; t represents the set of all positive triples, and T' represents the set of all negative triples; s represents the set of all positive doublets, S' represents the set of all negative doublets:
T’＝{(h′,r,t)|h′∈E}∪{(h,r,t′)|t′∈E}∪{(h,r′,t)|r′∈R},(h,r,t)∈T
S’＝{(e′,τ)|e′∈E}∪{(e,τ′)|τ′∈P},(e,τ)∈S
in the training process, better embedded vector representation is obtained by continuously enlarging the distance between the negative sample and the positive sample.
The invention divides the training process of each round into two stages: step one, obtaining an embedded vector representation of an entity and an entity relation through an optimization loss function; and step two, on the basis of obtaining the entity vector, calculating to obtain an embedded representation of the entity type through the loss function, and keeping the entity embedding unchanged in the process.
In order to conveniently obtain the resource quantity of each entity, the whole knowledge graph is preprocessed, each entity in the entities is used as a central entity, the searching depth is limited to be epsilon, and a directed subgraph taking each entity as the center is obtained in a deep searching mode.
Referring to a PageRank algorithm, iterative solution is carried out by using the formula for calculating the resource amount of the head entity flowing to the tail entity, the resource amount of each entity pair flowing from the head entity to the tail entity is obtained, and data are prepared in advance for subsequently calculating the confidence coefficient of the entity level in the local level.
In some embodiments, the calculation formula for obtaining the confidence of the entity level in the local level based on the resource amount of the head entity flowing to the tail entity is:
Figure BDA0003515511230000101
wherein ET (h, r, t) is the confidence of the entity level in the local level, r is the entity relationship, λ is the hyper-parameter for smoothing, and θ is the preset threshold corresponding to the entity relationship.
It will be appreciated that the entity hierarchy reflects the confidence level with the strength of association between two entities. It is considered that when a head entity has more resources to flow to a tail entity in a certain entity pair (h, t), the entity pair has stronger relevance, and entity relationship is more likely to exist between the entity pair, so that the triples containing the entity pair should have higher confidence. And finally, converting the resource quantity into an entity level confidence ET (h, r, t) by using a sigmoid function.
Smoothing here, for controlling the scaling scale, may be of meaningful value to let some extreme values. The hyper-parameter for smoothing may be set based on historical experience.
For the relation a, in the training process, statistics are performed on values of E (h, r, t) of all triples including the relation a and labels thereof, and the statistics is determined according to a statistical result.
In some embodiments, the confidence of the relationship level in the local level is determined based on the embedded vector representation corresponding to the entity, the entity relationship and the entity type in the knowledge-graph, and the formula is:
Figure BDA0003515511230000111
wherein RT (h, r, t) is the confidence of the relationship level in the local level.
It can be understood that entity relationship level confidence utilizes the translation invariance principle in embedded representation learning to construct an energy function E (h, r, t) | | M τH+r-MτtConsidering the correctness of the triples and the correctness of the entity and the entity type matching, the triples satisfying the energy function should have higher confidence level, that is, the smaller the value of the energy function E (h, r, t), the higher the confidence level, and therefore, the sigmoid function is used to convert the energy function value into the confidence level of the relationship level.
In some embodiments, the obtaining a quality assessment result of a knowledge graph based on the confidence level of the entity level in the local level, the confidence level of the relationship level in the local level, and the global level confidence level includes:
and performing weighted calculation on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level based on a preset entity level confidence coefficient weight value, a relation level confidence coefficient weight value and a global level confidence coefficient weight value to obtain a quality evaluation result of the knowledge graph.
It will be appreciated that the stage of computing the confidence of the global level of the triplet, in which the confidence is reflected by the multi-step path between two entity pairs, is used. The multi-step path between two entity pairs and the direct entity relationship between two entity pairs have similar semantic information, i.e. the embedded vector representation of the multi-step path should be a small distance from the embedded vector representation of the direct entity relationship. The conversion into a mathematical expression can be expressed as: the more the embedded vectors of the multi-step path and the direct entity relationship between the entity pair satisfy the expression | | p-r | | (where p represents the embedded vector of the multi-step path and r represents the embedded vector of the direct entity relationship), i.e., the smaller its value, the greater the confidence the triplet should have.
From the above description, the embedded vector representation of the multi-step path is an important ring of the global level confidence calculation. The invention fully considers the information of the dependency entity relationship, the entity type, the position of the entity on the path and the like between adjacent or non-adjacent entities, utilizes a TransFormer encoder, and combines the embedded vectors of the entity relationship, the entity and the entity type to learn the vector representation of the multi-step path, so that the vector representation of the multi-step path can fully reflect the semantic information contained in the multi-step path.
In addition, the special case that only a direct entity relationship exists between two entity pairs and no multi-step path exists is considered. According to the mathematical expression in the above description, when there is only a direct entity relationship between two entity pairs, there is | | p-r | | ═ 0, (p ═ r), then according to the definition that the smaller the value of | | p-r | | | is, the triplets have greater confidence, on the contrary, there is no entity pair of the multi-step path that has higher confidence and is not in accordance with the original purpose of the global level confidence calculation, therefore, in order to solve the problem brought by this situation, the penalty factor α is added to the entity pair that has only a direct entity relationship in the calculation.
And finally, obtaining a global level confidence GT (h, r, t) by using a sigmoid function:
Figure BDA0003515511230000121
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003515511230000122
and calculating the comprehensive confidence coefficient of the triple by using the local level confidence coefficient and the global level confidence coefficient obtained in the two previous stages, and further calculating the comprehensive confidence coefficient of the whole knowledge graph.
T(h,r,t)＝β1ET(h,r,t)+β2RT(h,r,t)+β3GT(h,r,t)
T (h, r, T) is the overall confidence of each triplet, where β1，β2，β3Is weighted by three levels of confidence and satisfies beta1+β2+β31, the integrated confidence of the entire knowledge-graph can be expressed asThe following:
Figure BDA0003515511230000123
wherein | T | represents the number of triples in the entire knowledge-graph, and T represents the set of triples.
In other embodiments, as shown in fig. 5, which is a partial example of a wireless network optimization knowledge graph, the wireless network optimization knowledge graph can provide a basis for the judgment of network problems and the formulation of network optimization schemes, so that the labor cost and the optimization time cost are reduced. For example: the triplets (capacity problem, satisfaction, value condition G) and (value condition G, probability 0.997, intra-cell interference) jointly describe that when the capacity problem satisfies the value condition G, the probability of 0.997 is caused by intra-cell interference, and provide a basis for network judgment; the triad (intra-cell interference, action set, state set, D _ DDQN algorithm (namely: double-depth competition reinforcement learning algorithm)) describes that the D _ DDQN algorithm can be adopted when the intra-cell interference is solved, and provides a basis for formulating a network optimization scheme.
However, the quality of the wireless network optimization knowledge graph has great influence on the accuracy of judging the network problems and the quality of the basis for formulating the network optimization scheme, so the quality evaluation of the wireless network optimization knowledge graph is carried out by using the quality evaluation scheme provided by the invention, and the specific flow is as follows
Firstly, acquiring the triple data of the knowledge graph and preprocessing the triple data, namely: obtaining embedded vector representations corresponding to each entity, entity relationship and entity type in the knowledge graph by using an improved TransE embedded learning algorithm; taking each entity in the knowledge graph as a central entity, setting the maximum search distance as m, and obtaining a directed subgraph of each central entity through a depth-first search algorithm; and (4) performing iterative computation to obtain the resource quantity of the head entity to the tail entity in each entity pair.
Obtaining the confidence of the entity level in the local level by using the calculated resource amount obtained by the tail entity between each entity pair from the head entity; and obtaining the confidence of the relationship level in the local level by using the learned entity, entity relationship and embedded vector representation of the entity type.
And taking the learned entity, entity relationship, embedded vector representation of entity type and position information of the entity in a certain multi-step path as input of a TransFormer encoder, obtaining a multi-step path vector representation capable of reflecting all semantic information of the multi-step path through learning of the TransFormer encoder, and obtaining the confidence of the global level.
And finally, obtaining a comprehensive confidence value of the whole knowledge graph to realize the quality evaluation of the whole knowledge graph.
In summary, the quality evaluation method for the network optimization knowledge graph provided by the invention obtains the entity, the entity relationship and the embedded vector representation corresponding to the entity type in the knowledge graph by obtaining the three-element data set of the knowledge graph and based on the three-element data set of the knowledge graph; in the process of learning the entity and the entity type embedded vector, the entity type and the entity embedded vector have a relationship, and when the entity and the entity type are correctly matched, the loss function value corresponding to the relationship between the entity type and the entity embedded vector is small, so that a large error is avoided.
Therefore, the entity type embedded vector is used for improving the energy function in the confidence coefficient calculation local level, and on the basis of considering the traditional triple errors, the matching errors of the entity and the entity type are considered at the same time, so that more accurate confidence coefficient evaluation of the knowledge graph is realized.
Further, a TransFormer encoder is used for learning the embedded representation of the multi-step path, the dependency relationship between adjacent entities, the dependency relationship between non-adjacent entities, the entity type and the position of the entity in the multi-step path are fully considered, and the embedded representation of the multi-step path can fully reflect the implied semantic information.
The quality evaluation device of the network optimization knowledge graph provided by the invention is described below, and the quality evaluation device of the network optimization knowledge graph described below and the quality evaluation method of the network optimization knowledge graph described above can be referred to correspondingly.
As shown in fig. 6, the quality evaluation apparatus 600 for network optimization knowledge-graph provided by the present invention includes: a data processing module 610, a first confidence calculation module 620, a second confidence calculation module 630, a third confidence calculation module 640, and a quality assessment module 650.
The data processing module 610 is configured to obtain a three-component data set of a knowledge graph, and obtain an entity, an entity relationship, and an embedded vector representation corresponding to an entity type in the knowledge graph based on the three-component data set of the knowledge graph;
a first confidence calculation module 620, configured to determine, based on the entities, a resource amount of a head entity flowing to a tail entity, and obtain a confidence of an entity level in a local level, based on the resource amount of the head entity flowing to the tail entity;
a second confidence calculation module 630, configured to obtain a confidence of a relationship level in a local level based on the entity in the knowledge graph, the entity relationship, and the embedded vector representation corresponding to the entity type;
a third confidence calculation module 640, configured to determine a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and the quality evaluation module 650 is configured to obtain a quality evaluation result of the knowledge graph based on the confidence of the entity level in the local level, the confidence of the relationship level in the local level, and the confidence of the global level.
In some embodiments, the first confidence computation module 620 comprises: a subgraph determination unit and a resource amount determination unit.
The subgraph determining unit is used for determining a directed subgraph of each entity by taking each entity in the knowledge graph as a central node of a search range and taking a preset value as a search depth;
and the resource amount determining unit is used for determining the resource amount of the head entity flowing to the tail entity based on the directed subgraph.
In some embodiments, the resource amount determination unit is further configured to determine the amount of resources flowing from the head entity to the tail entity based on the following formula:
Figure BDA0003515511230000151
wherein Q (t | h) represents the amount of resources flowing from the head entity h to the tail entity t, ItA set of head entities, O (e), representing entities t corresponding to the directed subgraphi) Representing entity e corresponding to the directed subgraphiOut of rotation of WeitRepresenting entity e corresponding to the directed subgraphiAnd the number of entity relations between the directed subgraph and the entity t, wherein N represents the total number of the entities corresponding to the directed subgraph; e is the probability that the resource flow controlling each entity jumps directly to a random entity.
In some embodiments, the first confidence calculation module 620 includes: a first confidence calculation unit.
The first confidence coefficient calculating unit is used for calculating the confidence coefficient of the entity level in the local level based on the following formula:
Figure BDA0003515511230000152
wherein ET (h, r, t) is the confidence of the entity level in the local level, r is the entity relationship, λ is the hyper-parameter for smoothing, and θ is the preset threshold corresponding to the entity relationship.
In some embodiments, the second confidence calculation module 630 is further configured to calculate the confidence of the relationship level in the local level based on the following formula:
Figure BDA0003515511230000161
wherein RT (h, r, t) is the confidence of the relationship level in the local level.
In some embodiments, the quality evaluation module 650 is further configured to perform weighted calculation on the entity level confidence in the local level, the relationship level confidence in the local level, and the global level confidence based on preset entity level confidence weight values, relationship level confidence weight values, and global level confidence weight values, so as to obtain a quality evaluation result of the knowledge graph.
The invention provides a quality evaluation device for a network optimization knowledge graph, which obtains an entity, an entity relation and an embedded vector representation corresponding to an entity type in the knowledge graph by obtaining a three-element data set of the knowledge graph and based on the three-element data set of the knowledge graph; in the process of learning the entity and the entity type embedded vector, the entity type and the entity embedded vector have a relationship, and when the entity and the entity type are correctly matched, the loss function value corresponding to the relationship between the entity type and the entity embedded vector is small, so that a large error is avoided.
Therefore, the energy function in the local level of confidence calculation is improved by using the entity type embedded vector, and on the basis of considering the traditional triple errors, the matching errors of the entity and the entity type are considered at the same time, so that more accurate confidence evaluation of the knowledge graph is realized.
The electronic device, the computer program product, and the storage medium provided by the present invention are described below, and the electronic device, the computer program product, and the storage medium described below may be referred to in correspondence with the quality evaluation method of the network optimization knowledge graph described above.
Fig. 7 illustrates a physical structure diagram of an electronic device, and as shown in fig. 7, the electronic device may include: a processor (processor)710, a communication Interface (Communications Interface)720, a memory (memory)730, and a communication bus 740, wherein the processor 710, the communication Interface 720, and the memory 730 communicate with each other via the communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform a method for quality assessment of a network optimization knowledgegraph, the method comprising:
acquiring a three-element data set of a knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to an entity type in the knowledge graph based on the three-element data set of the knowledge graph;
determining the resource amount of a head entity flowing to a tail entity based on the entities, and obtaining the confidence of an entity level in a local level based on the resource amount of the head entity flowing to the tail entity;
obtaining the confidence coefficient of the relationship level in the local level based on the entity in the knowledge graph, the entity relationship and the embedded vector representation corresponding to the entity type;
determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
In addition, the logic instructions in the memory 730 can be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being storable on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the method for quality assessment of network optimization knowledge-graph provided by the above methods, the method comprising:
acquiring a three-element data set of a knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to an entity type in the knowledge graph based on the three-element data set of the knowledge graph;
determining the resource amount of a head entity flowing to a tail entity based on the entities, and obtaining the confidence of an entity level in a local level based on the resource amount of the head entity flowing to the tail entity;
obtaining the confidence of the relationship level in the local level based on the entity in the knowledge graph, the entity relationship and the embedded vector representation corresponding to the entity type;
determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for quality assessment of a network-optimized knowledgegraph provided by performing the above methods, the method comprising:
acquiring a three-element data set of a knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to an entity type in the knowledge graph based on the three-element data set of the knowledge graph;
determining the resource amount of a head entity flowing to a tail entity based on the entities, and obtaining the confidence of an entity level in a local level based on the resource amount of the head entity flowing to the tail entity;
obtaining the confidence of the relationship level in the local level based on the entity in the knowledge graph, the entity relationship and the embedded vector representation corresponding to the entity type;
determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (10)

1. A quality evaluation method of a network optimization knowledge graph is characterized by comprising the following steps:
acquiring a three-element data set of a knowledge graph, and acquiring an entity, an entity relation and an embedded vector representation corresponding to an entity type in the knowledge graph based on the three-element data set of the knowledge graph;
determining the resource amount of a head entity flowing to a tail entity based on the entities, and obtaining the confidence of an entity level in a local level based on the resource amount of the head entity flowing to the tail entity;
obtaining the confidence coefficient of the relationship level in the local level based on the entity in the knowledge graph, the entity relationship and the embedded vector representation corresponding to the entity type;
determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
2. The method of claim 1, wherein the determining the amount of resources from the head entity to the tail entity based on the entities comprises:
determining a directed subgraph of each entity by taking each entity in the knowledge graph as a central node of a search range and taking a preset value as a search depth;
and determining the resource amount of the head entity flowing to the tail entity based on the directed subgraph.
3. The method for evaluating the quality of the network optimization knowledge-graph according to claim 2, wherein the calculation formula for determining the resource amount of the head entity to the tail entity based on the directed subgraph is as follows:
Figure FDA0003515511220000011
wherein Q (t | h) represents the amount of resources flowing from the head entity h to the tail entity t, ItRepresenting correspondence of said directed subgraphHead entity set of entity t, O (e)i) Representing entity e corresponding to the directed subgraphiThe out-of-range of (c) is,
Figure FDA0003515511220000012
representing entity e corresponding to the directed subgraphiAnd the number of entity relations between the directed subgraph and the entity t, wherein N represents the total number of the entities corresponding to the directed subgraph; e is the probability that the resource flow controlling each entity jumps directly to a random entity.
4. The method for evaluating the quality of a network-optimized knowledge-graph according to claim 3, wherein the calculation formula for obtaining the confidence of the entity level in the local level based on the resource amount of the head entity flowing to the tail entity is as follows:
Figure FDA0003515511220000021
wherein ET (h, r, t) is the confidence of the entity level in the local level, r is the entity relationship, λ is the hyper-parameter for smoothing, and θ is the preset threshold corresponding to the entity relationship.
5. The method according to claim 4, wherein the confidence of the relationship level in the local level is determined based on the embedded vector representation corresponding to the entity, the entity relationship and the entity type in the knowledge graph, and the formula is as follows:
Figure FDA0003515511220000022
wherein RT (h, r, t) is the confidence of the relationship level in the local level.
6. The method for evaluating the quality of the network-optimized knowledge-graph according to any one of claims 1 to 5, wherein the obtaining of the quality evaluation result of the knowledge-graph based on the confidence of the entity level in the local level, the confidence of the relationship level in the local level, and the confidence of the global level comprises:
and performing weighted calculation on the entity level confidence coefficient in the local level, the relationship level confidence coefficient in the local level and the global level confidence coefficient based on preset entity level confidence coefficient weight values, relationship level confidence coefficient weight values and global level confidence coefficient weight values to obtain a quality evaluation result of the knowledge graph.
7. A quality assessment device for a network optimization knowledge graph is characterized by comprising:
the data processing module is used for acquiring a three-element data set of the knowledge graph and acquiring an entity, an entity relation and an embedded vector representation corresponding to the entity type in the knowledge graph based on the three-element data set of the knowledge graph;
the first confidence coefficient calculation module is used for determining the resource amount of the head entity flowing to the tail entity based on the entity and obtaining the confidence coefficient of the entity level in the local level based on the resource amount of the head entity flowing to the tail entity;
the second confidence coefficient calculation module is used for obtaining the confidence coefficient of a relation level in a local level based on the entity in the knowledge graph, the entity relation and the embedded vector representation corresponding to the entity type;
a third confidence calculation module for determining a global level confidence based on multi-step paths between pairs of entities in the knowledge-graph;
and the quality evaluation module is used for obtaining a quality evaluation result of the knowledge graph based on the confidence coefficient of the entity level in the local level, the confidence coefficient of the relation level in the local level and the confidence coefficient of the global level.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps of the method for quality assessment of a network optimization knowledge-graph as claimed in any one of claims 1 to 6.
9. A non-transitory computer readable storage medium, having stored thereon a computer program, wherein the computer program, when being executed by a processor, implements the steps of the method for quality assessment of a network optimization knowledgegraph according to any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, carries out the steps of the method for quality assessment of a network optimization knowledgegraph according to any one of claims 1 to 6.
CN202210162918.0A 2022-02-22 2022-02-22 Quality evaluation method and device of network optimization knowledge graph and electronic equipment Pending CN114722211A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210162918.0A CN114722211A (en) 2022-02-22 2022-02-22 Quality evaluation method and device of network optimization knowledge graph and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210162918.0A CN114722211A (en) 2022-02-22 2022-02-22 Quality evaluation method and device of network optimization knowledge graph and electronic equipment

Publications (1)

Publication Number Publication Date
CN114722211A true CN114722211A (en) 2022-07-08

Family

ID=82235183

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210162918.0A Pending CN114722211A (en) 2022-02-22 2022-02-22 Quality evaluation method and device of network optimization knowledge graph and electronic equipment

Country Status (1)

Country Link
CN (1) CN114722211A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117725231A (en) * 2024-02-08 2024-03-19 中国电子科技集团公司第十五研究所 Content generation method and system based on semantic evidence prompt and confidence

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117725231A (en) * 2024-02-08 2024-03-19 中国电子科技集团公司第十五研究所 Content generation method and system based on semantic evidence prompt and confidence
CN117725231B (en) * 2024-02-08 2024-04-23 中国电子科技集团公司第十五研究所 Content generation method and system based on semantic evidence prompt and confidence

Similar Documents

Publication Publication Date Title
Grubinger et al. evtree: Evolutionary learning of globally optimal classification and regression trees in R
CN110222771B (en) Class identification method of zero-sample picture
US20050278139A1 (en) Automatic match tuning
CN111127364B (en) Image data enhancement strategy selection method and face recognition image data enhancement method
Lee et al. A hierarchical Bayesian framework for constructing sparsity-inducing priors
CN113190688A (en) Complex network link prediction method and system based on logical reasoning and graph convolution
US11797507B2 (en) Relation-enhancement knowledge graph embedding method and system
CN114254093A (en) Multi-space knowledge enhanced knowledge graph question-answering method and system
CN112084341A (en) Knowledge graph completion method based on triple importance
Shahmohammadi et al. Identification of System Software Components Using Clustering Approach.
CN114722211A (en) Quality evaluation method and device of network optimization knowledge graph and electronic equipment
Huang et al. Partitioned hybrid learning of Bayesian network structures
Kuzina et al. Diagnosing vulnerability of variational auto-encoders to adversarial attacks
CN113239699B (en) Depth knowledge tracking method and system integrating multiple features
CN111090765A (en) Social image retrieval method and system based on missing multi-modal hash
CN111339258B (en) University computer basic exercise recommendation method based on knowledge graph
CN111723010A (en) Software BUG classification method based on sparse cost matrix
CN111553401A (en) QoS prediction method based on graph model and applied to cloud service recommendation
CN111126443A (en) Network representation learning method based on random walk
Chen et al. Fuzzy clustering ensemble considering cluster dependability
CN115329139A (en) Knowledge graph embedding method based on group representation theory
CN114610967A (en) Data augmentation method applied to user portrait field
Fei et al. Soft Reasoning on Uncertain Knowledge Graphs
Hauser et al. An improved assessing requirements quality with ML methods
US20230106295A1 (en) System and method for deriving a performance metric of an artificial intelligence (ai) model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination