CN106886572B - Knowledge graph relation type inference method based on Markov logic network and device thereof - Google Patents
Knowledge graph relation type inference method based on Markov logic network and device thereof Download PDFInfo
- Publication number
- CN106886572B CN106886572B CN201710035828.4A CN201710035828A CN106886572B CN 106886572 B CN106886572 B CN 106886572B CN 201710035828 A CN201710035828 A CN 201710035828A CN 106886572 B CN106886572 B CN 106886572B
- Authority
- CN
- China
- Prior art keywords
- nodes
- inference
- knowledge graph
- logic network
- relation
- 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.)
- Active
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
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2455—Query execution
- G06F16/24564—Applying rules; Deductive queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a knowledge graph relation type conjecture method and a device thereof based on a Markov logic network, wherein the device comprises the following components: the inference rule acquisition module is used for generating inference rules according to the path characteristics among the known nodes of the data set knowledge graph; the credibility weight learning module is used for carrying out credibility weight learning on the inference rule through a Markov logic network and acquiring the inference rule with weight; the probabilistic reasoning module is used for carrying out probabilistic reasoning on the relation types existing among the nodes to be presumed according to the reasoning rule with the weight to acquire the relation type probability among the nodes to be presumed; and the relation type determining module is used for selecting the relation type with a larger probability value as the relation type between the nodes to be presumed according to the relation type probability acquired by the probability reasoning module. The invention realizes the automatic learning of the inference rule in the knowledge graph and the probabilistic inference of the relationship type between the nodes, and effectively ensures the accuracy of the relationship type inference possibly existing between the nodes.
Description
Technical Field
The invention belongs to the technical field of big data analysis, and particularly relates to a knowledge graph relation type conjecture method and a knowledge graph relation type conjecture device based on a Markov logic network.
Background
With the gradual transition of the internet from a document world wide web to a data world wide web, the correlated entity objects in the internet gradually transform into a knowledge map form which can be understood by a computer, and quick question answering, correlation query and entity recommendation based on the knowledge map provide great convenience for the life of people. However, the construction of the knowledge graph is a complex process, the association relationship between the entities is often difficult to obtain comprehensively in the knowledge extraction process, and how to guess the unknown relationship type between the entities based on the existing relationship type between the entities, so that the knowledge graph is improved, and the knowledge graph has very important practical significance and practical value.
The knowledge graph is taken as a new research direction and is a research hotspot in recent years, domestic and foreign researchers propose a large number of models for solving the link speculation problem of the knowledge graph, and mainly comprise a tensor decomposition-based method, a conversion-based method and a path inference-based method, wherein the tensor decomposition-based method expresses the relationship between entities by using a matrix, the conversion-based method embeds the entities and the relationship in the high-dimensional knowledge graph into a low-dimensional continuous vector, and the path inference-based method constructs a relationship classification eigenvector by counting a large number of relationship paths in the knowledge graph. The methods have certain effect on the general knowledge graph, but the problem of low reasoning accuracy is generally existed.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a knowledge graph relation type inference method and a knowledge graph relation type inference device based on a Markov logic network, which are used for realizing automatic learning of inference rules in a knowledge graph and probabilistic inference of relation types among nodes, solving the problems of unscientific and incomplete inference rules, poor expandability and the like caused by manual formulation of the inference rules in the knowledge graph perfecting process, having high reliability and effectively improving the accuracy of inference.
According to the design scheme provided by the invention, the knowledge graph relation type inference method based on the Markov logic network comprises the following contents:
step 1, determining path characteristics among known nodes and nodes to be presumed aiming at a knowledge graph of a known data set;
step 3, carrying out credibility weight learning on the inference rule through a Markov logic network to obtain an inference rule with weight;
and 5, determining the relationship type between the nodes to be inferred according to the probabilistic reasoning result.
As described above, step 2 includes the following steps:
step 21, traversing path features among known nodes of the knowledge graph by adopting a graph traversal method to generate evidence predicates and query predicates;
and 22, constructing an inference rule from the evidence predicate to the query predicate according to the evidence predicate and the query predicate.
Preferably, step 21 comprises the following steps: and traversing the path characteristics among the known nodes of the knowledge graph by setting the path length among the known nodes and adopting a breadth-first traversal method to generate an evidence predicate and an inquiry predicate.
As mentioned above, step 2 further includes the following steps: and step 23, constructing training set data for Markov logic network learning according to the inference rule and by combining the node information.
Preferably, step 23 further comprises: and if the node information meets the inference rule, adding the evidence predicate in the inference rule to the closed predicate corresponding to the query predicate into the training set data.
As described above, step 4 includes the following steps: and (4) carrying out Markov logic network inference according to the inference rule with the weight to obtain the relation type probability existing between the nodes to be inferred.
As described above, step 4 includes the following steps: performing Markov logic network inference according to an inference rule with weight and by combining known information among nodes to be inferred to obtain the relation type probability existing among the nodes to be inferred, wherein the known information at least comprises: and path characteristics between the node to be presumed and other nodes.
As mentioned above, step 4 further includes: and if the evidence predicates in the inference rules exist among the nodes to be inferred, adding the closed predicates corresponding to the evidence predicates into the known information.
As described above, step 5 includes the following steps: and selecting the relationship type with a larger probability value as the relationship type between the nodes to be presumed according to the relationship type probability existing between the nodes to be presumed.
A knowledge graph relation type inference device based on a Markov logic network comprises:
the inference rule acquisition module is used for generating inference rules according to the path characteristics among the known nodes of the data set knowledge graph;
the credibility weight learning module is used for carrying out credibility weight learning on the inference rule generated by the inference rule acquisition module through a Markov logic network and acquiring the inference rule with weight;
the probabilistic reasoning module is used for carrying out probabilistic reasoning on the relation types existing among the nodes to be presumed according to the reasoning rules with the weights acquired by the credibility weight learning module to acquire the relation type probabilities among the nodes to be presumed;
and the relation type determining module is used for selecting the relation type with a larger probability value as the relation type between the nodes to be presumed according to the relation type probability acquired by the probability reasoning module.
The invention has the beneficial effects that:
the method generates the inference rule according to the path characteristics of the known association relationship between the nodes, applies the Markov logic network to carry out weight learning on the credibility of the inference rule, carries out probability inference on the relationship types possibly existing between the nodes through the inference rule with the weight obtained by learning, solves the problems of unscientific and incomplete inference rule, poor expandability and the like of manual formulation of the inference rule in the process of improving the knowledge map, realizes the automatic learning of the inference rule in the knowledge map and the probability inference of the relationship types between the nodes, has high credibility, and effectively ensures the accuracy of the inference of the relationship types possibly existing between the nodes.
Description of the drawings:
FIG. 1 is a schematic view of an apparatus of the present invention;
FIG. 2 is a schematic flow diagram of the process of the present invention;
FIG. 3 is a diagrammatic view of the speculation process of the present invention;
FIG. 4 is a schematic diagram of predicate composition;
FIG. 5 is a diagram illustrating weight learning according to a third embodiment;
FIG. 6 is a graph showing the effect of the present invention under different missing ratios;
fig. 7 is a graph showing the effect of the present invention under different length path characteristics.
The specific implementation mode is as follows:
the concepts involved in the present invention are defined as follows:
and (3) inference rules: setting P in knowledge mapn(A,B)=P1P2...PnIn which P is1=A,PnB denotes a path of length n between node a and node B, Pn=r1r2...rn-1A sequence of relationship types representing the correspondence of the path, wherein riRepresenting a node PiAnd Pi+1The type of relationship between. If P is simultaneously present between A, Bn(A, B) and P1(A, B) path, then Pn(n>1) Can be regarded as P1Characteristic of the path of (1), PnAnd P1Form P betweenn→P1The inference rule of (1).
And (3) rule training set: the closed predicate set is used for training the credibility of the inference rule, and the higher the credibility of the inference rule is, the larger the weight is, and the higher the probability of occurrence in the real world is.
Rule evidence set: the known evidence closed predicate used for reasoning the unknown relationship types among the nodes can be used for reasoning the probability of various relationship types possibly existing among the unknown nodes by using the trained reasoning rule with the weight.
The present invention will be described in further detail below with reference to the accompanying drawings and technical solutions, and embodiments of the present invention will be described in detail by way of preferred examples, but the embodiments of the present invention are not limited thereto.
In an embodiment, referring to fig. 1, an apparatus for estimating a relationship type of a knowledge graph based on a Markov logic network includes:
the inference rule acquisition module is used for generating inference rules according to the path characteristics among the known nodes of the data set knowledge graph;
the credibility weight learning module is used for carrying out credibility weight learning on the inference rule generated by the inference rule acquisition module through a Markov logic network and acquiring the inference rule with weight;
the probabilistic reasoning module is used for carrying out probabilistic reasoning on the relation types existing among the nodes to be presumed according to the reasoning rules with the weights acquired by the credibility weight learning module to acquire the relation type probabilities among the nodes to be presumed;
and the relation type determining module is used for selecting the relation type with a larger probability value as the relation type between the nodes to be presumed according to the relation type probability acquired by the probability reasoning module.
The inference rules are generated according to the path characteristics of the known association relationship between the nodes, the Markov logic network is applied to weight learning of the credibility of the inference rules, the inference rules with the weights obtained through learning are used for probability inference of the relationship types possibly existing between the nodes, the problems that the inference rules are unscientific and incomplete and poor in expandability when being manually formulated in the knowledge graph perfecting process are solved, automatic learning of the inference rules in the knowledge graph and probability inference of the relationship types between the nodes are achieved, the credibility is high, and therefore the inference result is more real and credible.
In a second embodiment, referring to fig. 1 to 2, a knowledge graph relationship type inference method based on a Markov logic network includes the following steps:
step 1, determining path characteristics among known nodes and nodes to be presumed aiming at a knowledge graph of a known data set;
step 3, carrying out credibility weight learning on the inference rule through a Markov logic network to obtain an inference rule with weight;
and 5, determining the relationship type between the nodes to be inferred according to the probabilistic reasoning result.
By introducing attitude information, the deviation amount between the Beidou receiver antenna and the GPS receiver antenna is fully considered, so that the detection result is more real and credible; through a coordinate transfer algorithm, the deviation amount between the Beidou receiver antenna and the GPS receiver antenna can be accurately considered, and the purpose that a plurality of Beidou receivers participate in testing at the same time is achieved.
In a third embodiment, referring to fig. 1 to 7, a knowledge graph relationship type inference method based on a Markov logic network includes the following steps:
a. and determining path characteristics among the known nodes and the nodes to be inferred according to the knowledge graph of the known data set.
b. Generating an inference rule according to the path characteristics among the known nodes, wherein the inference rule comprises the following contents:
b1, traversing path features among the known nodes of the knowledge graph by adopting a graph traversal method to generate evidence predicates and query predicates; the path characteristics among the known nodes of the knowledge graph are traversed by setting the path length among the known nodes and adopting a breadth-first traversal method, so that an evidence predicate and an inquiry predicate are generated, wherein the predicate is composed as shown in FIG. 4.
And b2, constructing an inference rule from the evidence predicate to the query predicate according to the evidence predicate and the query predicate.
And b3, constructing training set data for Markov logic network learning according to the inference rule and by combining the node information. And if the node information meets the inference rule, adding the evidence predicate in the inference rule to the closed predicate corresponding to the query predicate into the training set data.
c. And (4) carrying out credibility weight learning on the inference rule through a Markov logic network to obtain the inference rule with weight.
d. And performing probabilistic reasoning on the relationship types among the nodes to be presumed through the reasoning rules with the weights, and performing Markov logic network reasoning according to the reasoning rules with the weights to obtain the relationship type probability existing among the nodes to be presumed. Or carrying out Markov logic network inference by an inference rule with weight and combining with known information among the nodes to be inferred to obtain the relation type probability existing among the nodes to be inferred, wherein the known information at least comprises: and path characteristics between the node to be presumed and other nodes. And if the evidence predicates in the inference rules exist among the nodes to be inferred, adding the closed predicates corresponding to the evidence predicates into the known information.
e. Determining the relationship type between the nodes to be inferred according to the probabilistic reasoning result, which comprises the following contents: and selecting the relationship type with a larger probability value as the relationship type between the nodes to be presumed according to the relationship type probability existing between the nodes to be presumed.
To further illustrate the effectiveness of the present invention, the following examples are provided to further illustrate the invention:
experiments were conducted on a Kinships family dataset, which describes 112 relationship type triplets between 12 pairs of entities of 12 types of relationships between 24 family person entity objects, as shown in fig. 3, and the following implementation is specifically made:
(1) and setting query predicates and evidence predicates of the association rules, and constructing a rule description file required by the Markov logic network system for parameter learning according to the concepts of the inference rules and predicate logics shown in FIG. 4, wherein the rule description file is recorded as rule. The knowledge graph is essentially a directed graph, and P among nodes of the knowledge graph is determined in a graph traversal mode1Characteristic P of the relationship pathn(n>1) And traversing, generating a corresponding evidence predicate and a corresponding query predicate, constructing an inference rule from the evidence predicate to the query predicate, and writing the inference rule into' rule. In the construction process, in order to control the number of inference rules, a threshold value of n can be set, and a breadth-first traversal method is adopted to traverse the knowledge graph.
(2) And (3) constructing a rule training set: according to defined association rules, and P of entity instance1And PnAnd constructing a training set file used in Markov logic network parameter learning in occurrence conditions in the training set, and recording the training set file as' lean. If the instance meets a certain inference rule in the step (2), the evidence predicate and the query predicate of the rule are usedAnd the closed predicates corresponding to the words are written into' left.
(3) Constructing an evidence file: and (4) according to the occurrence condition of the path characteristics between the nodes to be presumed, constructing an evidence file used when the Markov logic network carries out reasoning, and recording the evidence file as proof. And if some evidence predicate in the inference rule exists among the nodes, writing a closed predicate corresponding to the evidence predicate into' proof.
(4) Inference rule weight learning: the method comprises the steps of compiling in a linux environment to obtain an executable file by adopting an Alchemy system which is a Markov logic network application platform developed by Domingos et al, and realizing parameter learning of the Markov logic network by using a learwts command to obtain a rule file ruleout.
(5) And (3) predicting the relationship type among the nodes: and (3) according to the rule file ruleout.mln and the evidence file proof.db with the weight obtained by learning in the last step, carrying out reasoning of the Markov logic network by using an inference command in an Alchemy system, and outputting a reasoning result file result.
Through the above contents, the occurrence times of part of inference rules and the corresponding weights obtained by applying Markov logic network learning are shown in FIG. 5, and the true relational inference rules have more occurrence times in the graph, so that higher weights can be obtained, that is, the reliability is higher; by removing the semantic edges of the relations among some characters in Kinships and using the known relation type examples as a training set to guess the relation types of the missing characters, the accuracy and recall rate of the experiment are shown in fig. 6 and 7, it can be seen that the technical scheme recorded by the invention can guess the relation types of the missing nodes in the knowledge graph, and the more the training examples are, the more comprehensive the used path characteristics are, the higher the guessing accuracy rate is, the problems of unscientific and incomplete inference rules, poor expandability and the like manually formulated in the knowledge graph perfecting process are solved, the automatic learning of the inference rules in the knowledge graph and the probabilistic inference of the relation types among the nodes are realized, the reliability is high, and the accuracy rate of the inference of the relation types possibly existing among the nodes is effectively ensured.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (8)
1. A knowledge graph relation type inference method based on a Markov logic network is characterized by comprising the following contents:
step 1, determining path characteristics among known nodes and nodes to be presumed aiming at a knowledge graph of a known data set;
step 2, generating an inference rule according to the path characteristics among the known nodes, and setting P in the knowledge graphn(A,B)=P1P2…PnIn which P is1=A,Pn=B,Pn(A, B) represents a path of length n between the person entity object node A and the person entity object node B, if P isn=r1r2…rn-1Then r is1r2…rn-1Indicates the path PnA corresponding relationship type sequence, wherein riRepresenting a node PiAnd Pi+1Type of relationship between P if P is simultaneously present between A, Bn(A, B) and P1(A, B) path, then Pn(n>1) Viewed as P1Characteristic of the path of (1), PnAnd P1Form P betweenn→P1The inference rule of (2);
step 3, carrying out credibility weight learning on the inference rule through a Markov logic network to obtain an inference rule with weight;
step 4, carrying out probabilistic reasoning on the relation types among the nodes to be inferred through a reasoning rule with weight, and utilizing a known relation type example as a training set to infer the missing person relation types;
step 5, determining the relationship type between the nodes to be inferred according to the probabilistic reasoning result;
step 2 comprises the following steps:
step 21, traversing path features among known nodes of the knowledge graph by adopting a graph traversal method to generate evidence predicates and query predicates;
step 22, establishing an inference rule from the evidence predicate to the query predicate according to the evidence predicate and the query predicate; step 21 comprises the following: the method comprises the steps of setting the length of a path between known nodes and traversing the path characteristics between the known nodes of a knowledge graph by adopting a breadth-first traversal method to generate an evidence predicate and an inquiry predicate, wherein the predicate is represented as follows:
。
2. The knowledge graph relationship type inference method based on the Markov logic network as claimed in claim 1, wherein the step 2 further comprises the following contents: and step 23, constructing training set data for Markov logic network learning according to the inference rule and by combining the node information.
3. The knowledge graph relationship type inference method based on Markov logic networks according to claim 2, wherein the step 23 further comprises: and if the node information meets the inference rule, adding the evidence predicate in the inference rule to the closed predicate corresponding to the query predicate into the training set data.
4. The knowledge graph relationship type inference method based on the Markov logic network as claimed in claim 1, wherein the step 4 comprises the following contents: and (4) carrying out Markov logic network inference according to the inference rule with the weight to obtain the relation type probability existing between the nodes to be inferred.
5. The knowledge graph relationship type inference method based on the Markov logic network as claimed in claim 1, wherein the step 4 comprises the following contents: performing Markov logic network inference according to an inference rule with weight and by combining known information among nodes to be inferred to obtain the relation type probability existing among the nodes to be inferred, wherein the known information at least comprises: and path characteristics between the node to be presumed and other nodes.
6. The knowledge graph relationship type inference method based on the Markov logic network as claimed in claim 5, wherein step 4 further comprises: and if the evidence predicates in the inference rules exist among the nodes to be inferred, adding the closed predicates corresponding to the evidence predicates into the known information.
7. The knowledge graph relationship type inference method based on the Markov logic network as claimed in claim 5, wherein the step 5 comprises the following contents: and selecting the relationship type with a larger probability value as the relationship type between the nodes to be presumed according to the relationship type probability existing between the nodes to be presumed.
8. A knowledge graph relationship type inference device based on a Markov logic network, which is realized by the knowledge graph relationship type inference method based on the Markov logic network of claim 1, and comprises:
the inference rule acquisition module is used for generating inference rules according to the path characteristics among the known nodes of the data set knowledge graph;
the credibility weight learning module is used for carrying out credibility weight learning on the inference rule generated by the inference rule acquisition module through a Markov logic network and acquiring the inference rule with weight;
the probabilistic reasoning module is used for carrying out probabilistic reasoning on the relation types existing among the nodes to be presumed according to the reasoning rules with the weights acquired by the credibility weight learning module to acquire the relation type probabilities among the nodes to be presumed;
and the relation type determining module is used for selecting the relation type with a larger probability value as the relation type between the nodes to be presumed according to the relation type probability acquired by the probability reasoning module.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710035828.4A CN106886572B (en) | 2017-01-18 | 2017-01-18 | Knowledge graph relation type inference method based on Markov logic network and device thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710035828.4A CN106886572B (en) | 2017-01-18 | 2017-01-18 | Knowledge graph relation type inference method based on Markov logic network and device thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106886572A CN106886572A (en) | 2017-06-23 |
CN106886572B true CN106886572B (en) | 2020-06-19 |
Family
ID=59176405
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710035828.4A Active CN106886572B (en) | 2017-01-18 | 2017-01-18 | Knowledge graph relation type inference method based on Markov logic network and device thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106886572B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107423820B (en) * | 2016-05-24 | 2020-09-29 | 清华大学 | Knowledge graph representation learning method combined with entity hierarchy categories |
CN110634088A (en) * | 2018-06-25 | 2019-12-31 | 阿里巴巴集团控股有限公司 | Case refereeing method, device and system |
CN110633458A (en) * | 2018-06-25 | 2019-12-31 | 阿里巴巴集团控股有限公司 | Method and device for generating referee document |
CN109308525B (en) * | 2018-07-26 | 2024-04-19 | 南京医科大学第二附属医院 | Intra-hospital infection field-planting investigation intelligent reasoning method based on Markov logic network |
CN110019843B (en) * | 2018-09-30 | 2020-11-06 | 北京国双科技有限公司 | Knowledge graph processing method and device |
CN111178531B (en) * | 2018-11-09 | 2023-09-22 | 百度在线网络技术(北京)有限公司 | Method, device and storage medium for acquiring relationship reasoning and relationship reasoning model |
CN109783652B (en) * | 2019-01-30 | 2023-01-17 | 东南大学 | Construction method of inference network for robot to infer room type according to indoor articles by using common sense |
CN110879843B (en) * | 2019-08-06 | 2020-08-04 | 上海孚典智能科技有限公司 | Method for constructing self-adaptive knowledge graph technology based on machine learning |
CN112163077B (en) * | 2020-09-28 | 2024-06-04 | 华南理工大学 | Knowledge graph construction method for field question and answer |
CN113626612A (en) * | 2021-08-13 | 2021-11-09 | 第四范式(北京)技术有限公司 | Prediction method and system based on knowledge graph reasoning |
CN114254132A (en) * | 2022-03-01 | 2022-03-29 | 北京理工大学 | Retrieval method of material data |
CN116415564B (en) * | 2023-06-12 | 2023-08-25 | 中国电子科技集团公司信息科学研究院 | Functional point amplification method and system based on knowledge graph |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103824115A (en) * | 2014-02-28 | 2014-05-28 | 中国科学院计算技术研究所 | Open-network-knowledge-base-oriented between-entity relationship deduction method and system |
CN105574098A (en) * | 2015-12-11 | 2016-05-11 | 百度在线网络技术(北京)有限公司 | Knowledge graph generation method and device and entity comparing method and device |
CN105760425A (en) * | 2016-01-17 | 2016-07-13 | 曲阜师范大学 | Ontology data storage method |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9195941B2 (en) * | 2013-04-23 | 2015-11-24 | International Business Machines Corporation | Predictive and descriptive analysis on relations graphs with heterogeneous entities |
CN104361396A (en) * | 2014-12-01 | 2015-02-18 | 中国矿业大学 | Association rule transfer learning method based on Markov logic network |
-
2017
- 2017-01-18 CN CN201710035828.4A patent/CN106886572B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103824115A (en) * | 2014-02-28 | 2014-05-28 | 中国科学院计算技术研究所 | Open-network-knowledge-base-oriented between-entity relationship deduction method and system |
CN105574098A (en) * | 2015-12-11 | 2016-05-11 | 百度在线网络技术(北京)有限公司 | Knowledge graph generation method and device and entity comparing method and device |
CN105760425A (en) * | 2016-01-17 | 2016-07-13 | 曲阜师范大学 | Ontology data storage method |
Non-Patent Citations (1)
Title |
---|
基于Web和Email的多元社会网络抽取与分析关键技术研究;尹美娟;《中国博士学位论文全文数据库信息科技辑》;20130615(第06期);I139-21 第1.2.1、3.3、3.4、6.2.2节、图7、8 * |
Also Published As
Publication number | Publication date |
---|---|
CN106886572A (en) | 2017-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106886572B (en) | Knowledge graph relation type inference method based on Markov logic network and device thereof | |
US11631009B2 (en) | Multi-hop knowledge graph reasoning with reward shaping | |
US20220075670A1 (en) | Systems and methods for replacing sensitive data | |
Kalisch et al. | Causal structure learning and inference: a selective review | |
CN112615888B (en) | Threat assessment method and device for network attack behavior | |
WO2022179384A1 (en) | Social group division method and division system, and related apparatuses | |
CN111598461A (en) | Enterprise knowledge graph-based real-time risk generalization and conduction analysis method | |
CN108229731B (en) | System and method for predicting user behavior under hot topics under multi-message interaction | |
WO2020233245A1 (en) | Method for bias tensor factorization with context feature auto-encoding based on regression tree | |
Bojchevski et al. | Is pagerank all you need for scalable graph neural networks | |
US20220300557A1 (en) | Quantifying and improving the performance of computation-based classifiers | |
CN112116069A (en) | Attention-LSTM-based reinforcement learning Agent knowledge inference method | |
Lawrence et al. | Explaining neural matrix factorization with gradient rollback | |
CN116450850A (en) | Space-time knowledge graph completion method based on sequence encoder | |
US11544281B2 (en) | Query-oriented approximate query processing based on machine learning techniques | |
CN117216096B (en) | Approximate query processing optimization method | |
Xin et al. | Subjective knowledge base construction powered by crowdsourcing and knowledge base | |
CN109492924B (en) | Influence evaluation method based on second order of self and behavior value of microblog user | |
US20210065047A1 (en) | Multi-tiered system for scalable entity representation learning | |
Borboudakis et al. | Scoring and searching over Bayesian networks with causal and associative priors | |
US20230018525A1 (en) | Artificial Intelligence (AI) Framework to Identify Object-Relational Mapping Issues in Real-Time | |
Wyllie et al. | Fairness feedback loops: training on synthetic data amplifies bias | |
CN114860952A (en) | Graph topology learning method and system based on data statistics and knowledge guidance | |
CN116962996B (en) | Information propagation prediction method, device and equipment based on particle swarm optimization | |
US11829735B2 (en) | Artificial intelligence (AI) framework to identify object-relational mapping issues in real-time |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |