CN109033129B - Multi-source information fusion knowledge graph representation learning method based on self-adaptive weight - Google Patents
Multi-source information fusion knowledge graph representation learning method based on self-adaptive weight Download PDFInfo
- Publication number
- CN109033129B CN109033129B CN201810563786.6A CN201810563786A CN109033129B CN 109033129 B CN109033129 B CN 109033129B CN 201810563786 A CN201810563786 A CN 201810563786A CN 109033129 B CN109033129 B CN 109033129B
- Authority
- CN
- China
- Prior art keywords
- entity
- information
- weight
- relation
- representation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000004927 fusion Effects 0.000 title claims abstract description 29
- 230000006870 function Effects 0.000 claims abstract description 42
- 239000013598 vector Substances 0.000 claims abstract description 31
- 238000005457 optimization Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 abstract description 13
- 238000013519 translation Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 239000008186 active pharmaceutical agent Substances 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
Images
Landscapes
- Machine Translation (AREA)
Abstract
The invention discloses a multisource information fusion knowledge graph representation learning method based on self-adaptive weight, which comprises the steps of firstly considering the fusion of text information and structured information, adopting a translation-based model between an entity vector and a relation vector, optimizing a score function by adjusting the weight between the entity vector and the relation vector, and performing type constraint training on the structured information which is classified at the early stage without introducing more parameters; and then, associating the entity vectors and the relation vectors by using a loss function, optimizing the loss function, and learning the vectors of the vectors and the relation of each entity in the knowledge graph when the optimization target is reached. The invention solves the problem that the weight is not considered in the fusion of the text information and the structured information in the knowledge base, utilizes the existing hierarchical information of the structured information in the knowledge base, more accurately represents the mutual connection between the entity and the relation, and applies the mutual connection to the large-scale knowledge map.
Description
Technical Field
The invention relates to the technical field of knowledge maps and deep learning, in particular to a multi-source information fusion knowledge map representation learning method based on adaptive weight.
Background
With the rapid development of society, we slowly enter an information-oriented era. Huge amounts of new data and information are produced in different forms each day. The mobile internet is the most effective and convenient information acquisition platform in the current society, the demand of users for acquiring real information is increasing day by day, and how to acquire effective information from mass data becomes a major problem in many fields. The knowledge-graph is generated accordingly.
One typically organizes knowledge in a knowledge base in the form of a network, where each node represents an entity and each edge represents a relationship between two entities, the triplets being in the form of (entity 1, relationship, entity 2). FIG. 1 is an exemplary diagram of a typical triplet within a knowledge-graph. Wherein the nodes "Shashibia", "Romeo and Julie" represented by the ellipses are all entities, and the "authors" represented by the connecting edges are relations. Therefore, most knowledge can be represented by triples, corresponding to a chain and two linked entities in the knowledge base network, which is a common representation of the knowledge base. In recent years, deep learning has gained widespread attention in the fields of speech recognition, image analysis, and natural language processing. Representation learning aims at representing semantic information of a study object as a dense low-dimensional real-valued vector. In the low-dimensional vector space, the closer the two objects are, the higher the semantic similarity is. The direction has recently made an important progress, semantic relation of entities and relations can be calculated efficiently in a low-dimensional space, the problem of data sparsity is effectively solved, and the performance of knowledge acquisition, fusion and reasoning is remarkably improved.
One of the major challenges facing knowledge representation learning is how to achieve multi-source information fusion. The triple structure information of the existing knowledge graph, such as TransE, is only used for representing and learning, and a large amount of other information related to knowledge is not effectively used, such as other information of a knowledge base, such as description information of entities and relations, category information and the like.
Disclosure of Invention
The invention provides a multisource information fusion knowledge map representation learning method based on self-adaptive weight, aiming at the problem that the relation between a structural model and text information cannot be fully utilized after the structural model and the text information are fused in the existing knowledge map representation learning method.
In order to solve the problems, the invention is realized by the following technical scheme:
the multi-source information fusion knowledge graph representation learning method based on the self-adaptive weight specifically comprises the following steps:
step 1, balancing fusion of text information and structured information by using self-adaptive weight, and defining a total score function f (h, r, t) of the text information and the structured information which are mutually associated:
f(h,r,t)＝(1-λ)(||hd+r-td||+||hd+r-MrttS||+||MrhhS+r-td||)+λ(||Mrhh+r+Mrtt||)
wherein λ represents weight, h represents head entity, t represents tail entity, rRepresenting the relationship of a head entity h and a tail entity t, hdRepresenting a text-based representation of a head entity, tdRepresenting the text-based representation of the tail entity, hSRepresenting head entities based on a structured representation, tSRepresenting tail entities based on a structured representation, MrhIs based on a projection matrix defined by the head entity, MrhIs a projection matrix defined according to the tail entity;
and 2, establishing a loss function based on the fusion of the text information and the structured information of the self-adaptive weight based on the total score function f (h, r, t) defined in the step 1, and learning the vector representation of the entity and the relation by minimizing the loss function to achieve the optimization goal.
In the step 1, the value range of the weight λ is λ ∈ (0, 1).
In step 2, a random gradient descent method is used to minimize the loss function.
In the step 2, the constructed loss function L is:
wherein, [ f (h, r, t) + gamma-f (h ', r, t')]+Max (0, f (h, r, t) + γ -f (h ', r, t')); gamma is a set boundary value; (h, r, t) represents a triplet of the knowledge graph, namely a positive example triplet, h represents a head entity, t represents a tail entity, r represents a relation between the head entity and the tail entity, f (h, r, t) represents a score function of the positive example triplet, and S (h, r, t) represents a positive example triplet set; (h ', r, t ') represents a negative example triple constructed by randomly replacing the head entity h and the tail entity t, f (h ', r, t ') represents a score function of the negative example triple, and S ' (h, r, t) represents a negative example triple set.
Compared with the prior art, the method has the advantages that the fusion of text information and structured information is considered, a translation-based model between an entity vector and a relation vector is adopted, the score function is optimized by adjusting the weight between the entity vector and the relation vector, type constraint training is carried out on the structured information which is classified at the early stage, and more parameters are not required to be introduced; and then, associating the entity vectors and the relation vectors by using a loss function, optimizing the loss function, and learning the vectors of the vectors and the relation of each entity in the knowledge graph when the optimization target is reached. The invention solves the problem that the weight is not considered in the fusion of the text information and the structured information in the knowledge base, utilizes the existing hierarchical information of the structured information in the knowledge base, more accurately represents the mutual connection between the entity and the relation, and applies the mutual connection to the large-scale knowledge map.
Drawings
FIG. 1 is an exemplary diagram of relationship triples in a knowledge-graph.
FIG. 2 is a flow chart of a learning method according to the present invention.
FIG. 3a is an exemplary diagram of a triple representation obtained from a prior knowledge graph representation learning method.
FIG. 3b is an exemplary diagram of a triplet representation resulting from a knowledge graph representation learning method according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings in conjunction with specific examples.
In view of the fact that the prior art only considers the fusion of text information and structured information, how to adjust the weight between the text information and the structured information is not fully considered to achieve the best effect, the existing hierarchical information of the structured information in a knowledge base is not utilized, the number of learning parameters is large, the relation between an entity and a relation cannot be accurately represented, and the method cannot be well applied to a large-scale knowledge map.
The invention fully considers the fusion of the text information and the structured information of the self-adaptive weight and enriches the structured information according to the unused hierarchical information. When text information and structured information are fused, a self-adaptive weight is provided to balance the fusion of the text information and the structured information, and type constraint training is carried out on the structured information classified in advance. By adding the multivariate information fusion method of the self-adaptive weight, the heterogeneity and the unbalance of the entities and the relations in the knowledge base are solved, the entities and the relations and the mutual connection among the entities and the relations are more accurately represented, and the multivariate information fusion method is applied to a large-scale knowledge graph.
Specifically, a multi-source information fusion knowledge graph representation learning method based on adaptive weight, as shown in fig. 2, includes the following steps:
step 1, when text information and structured information are fused, a self-adaptive weight is provided to balance the fusion of the text information and the structured information, and type constraint training is carried out on the structured information classified in advance.
Based on the fusion of the text information and the structured information, a mutually associated overall score function f is defined:
f(h,r,t)＝(1-λ)fD(h,r,t)+λfS(h,r,t)
fD (h, r, t) represents a scoring function based on the textual representation:
fD(h,r,t)＝fDD(h,r,t)+fDS(h,r,t)+fSD(h,r,t)
＝||hd+r-td||+||hd+r-MrttS||+||MrhhS+r-td||
fS(h, r, t) represents a score function based on the structured representation:
fS(h,r,t)＝||Mrhh+r+Mrtt||
wherein, λ represents weight, λ belongs to (0, 1), h represents head entity, t represents tail entity, r represents relation between head entity h and tail entity t, h represents weight, t represents weight, r represents weight, h represents weight, t represents weight, r represents weight, h represents weight, r represents weight, h represents weight, r represents weight, h represents head entity, h represents weight, h represents head entity, r, h represents weight, represents head entity, h represents weight, r, represents weight, h represents weight, h represents weight, h represents head entity, h represents weight, represents head entity, h, represents weight, r, h represents weight, represents head entity, represents weight, r, h represents weight, h represents head entity, r, h represents weight, r, h represents weight, h, represents weight, representsdRepresenting a text-based representation of a head entity, tdRepresenting the text-based representation of the tail entity, hSRepresenting head entities based on a structured representation, tSRepresenting tail entities based on a structured representation, MrhIs based on a projection matrix defined by the head entity, MrhIs a projection matrix defined from the tail entity.
And 2, providing a loss function based on the fusion of the text information and the structured information of the self-adaptive weight, and learning the vector representation of the entity and the relation by minimizing the loss function to achieve the optimization target.
Step 21, defining a loss function as:
wherein, [ f (h, r, t) + gamma-f (h ', r, t')]+＝max(0,f(h,r,t)+γ-f(h',r,t'))+(ii) a Gamma is a set boundary value; (h, r, t) represents a triplet of the knowledge graph, namely a positive example triplet, h represents a head entity, t represents a tail entity, r represents the relation between the head entity h and the tail entity t, f (h, r, t) represents a score function of the positive example triplet, and S (h, r, t) represents a positive example triplet set; (h ', r, t ') represents a negative example triple constructed by the head entity h and the tail entity t which are replaced immediately, f (h ', r, t ') represents a score function of the negative example triple, and S ' (h, r, t) represents a negative example triple set;
and step 22, minimizing a loss function by adopting a random gradient descent method, and learning to obtain each entity vector and relationship vector in the knowledge graph and the mutual connection between the entity vectors and the relationship vectors.
The process of minimizing the loss function is the process of minimizing the score function, and the process of minimizing is the process of achieving the optimization goal. And E in the triple scoring function adopts an energy function in a TransE model, so that in the process of minimizing the loss function, when the type of the relation r is a simple relation type 1-1 or a complex relation type 1-N, N-1, N-N, h, t and r are continuously adjusted to ensure that h + r is equal to t as far as possible.
The method learns and obtains the self-adaptive weight-based knowledge graph representation learning method of multi-source information fusion, and the model for performing type constraint training on the structural information classified in advance is more accurate and effective.
The problem that the existing method cannot solve can be solved by adding text information: when predicting a new entity (untrained entity), the original method randomly gives a vector representation to it, so that its score function is poor compared with that of the trained entity, its loss function is large, and the prediction effect is poor. In the method for adding text information which is not improved by us, when a new entity (an untrained entity) appears, the structural method randomly gives a vector representation to the new entity, but a text description of the new entity exists in a knowledge base, the text description of the new entity can be processed into a vector represented by the text, and a new score function is obtained by adding the vector represented by the text and the vector represented by the training structural method, so that the optimization goal is achieved. When a new entity (untrained entity) appears, although the score function is optimized by text description by adding text information, in the method of obtaining the new score function by adding the two score functions, the structured information can provide wrong information and the weight is high, which is obviously unreasonable.
The invention proposes the fusion of structured information and text information at adaptive weights. And updating the weight representations of the entity and the text information through the training times of the entity, wherein when the weight represented by the structured information is properly increased a little while the weight represented by the text information is decreased a little every time the entity is trained. This is because the training times of the entities are distributed in a long tail when no text information is added, and the occurrence times of the common entities and the entities with multiple classes are large, and the relative training times are also sufficient, so that the entities with large training times are closer to the correct representation, and the representation of the entities with small training times is relatively weak. In this way, we consider that the entity with the larger number of occurrences is good enough to represent the structured information, we can increase the weight of the part of the structured representation where such entity appears, and the weight of the structured representation increases with the number of occurrences. On the contrary, for entities with a small number of occurrences, the representation of the structured information is not good enough, so that the weights of the text information representation parts of the entities are increased, and if the entities do not appear at one time, the weights of the structured information parts are also 0, namely all the entities are represented by the text information, so that the text information is utilized to represent new entities which do not appear, and meanwhile, the structured information randomly endowed with vectors is filtered.
FIG. 3a is an exemplary diagram of a triple representation obtained from a prior art knowledge graph representation learning method. In fig. 3a, the hierarchical information of the knowledge-graph triplets is not considered. Hierarchical information means that entities may have different roles in different scenarios, such as shakespeare being both a writer and a musician, and also bob having such attributes. We consider that having multiple types of entities with different relationships should have different representations. We construct a projection matrix M of a particular type from a hierarchyrThe head entity h and the tail entity t are then represented by a particular projection matrix constructed. Thus, how many relationships an entity has will have how many mappings to represent the particular representation of the entity under each relationship, respectively. In fig. 3b, we show the kind of entity by specific relations, and entities with the same type tend to be a cluster and have similar representation in training, which is also the main cause of error in entity prediction in fact. In the invention, the training probability of selecting entities with the same type under the specific relation type information can be improved, and the target is optimized in such a way.
The invention solves the problems that the imbalance and heterogeneity of the entities and the relations and the calculation caused by excessive parameters in the prior art are too complex, the mutual relation between the entities and the relations in the knowledge graph cannot be well represented by a method, and the method cannot be well applied to large-scale knowledge graphs, and has good practicability.
It should be noted that, although the above-mentioned embodiments of the present invention are illustrative, the present invention is not limited thereto, and thus the present invention is not limited to the above-mentioned embodiments. Other embodiments, which can be made by those skilled in the art in light of the teachings of the present invention, are considered to be within the scope of the present invention without departing from its principles.
Claims (4)
1. The multi-source information fusion knowledge graph representation learning method based on the self-adaptive weight is characterized by specifically comprising the following steps of:
step 1, balancing fusion of text information and structured information by using self-adaptive weight, and defining a total score function f (h, r, t) of the text information and the structured information which are mutually associated:
f(h,r,t)＝(1-λ)(||hd+r-td||+||hd+r-MrttS||+||MrhhS+r-td||)+λ(||Mrhh+r+Mrtt||)
wherein, λ represents weight, h represents head entity, t represents tail entity, r represents relation between head entity h and tail entity t, h represents weight, t represents tail entitydRepresenting a text-based representation of a head entity, tdRepresenting the text-based representation of the tail entity, hSRepresenting head entities based on a structured representation, tSRepresenting tail entities based on a structured representation, MrhIs based on a projection matrix defined by the head entity, MrhIs a projection matrix defined according to the tail entity;
and 2, establishing a loss function based on the fusion of the text information and the structured information of the self-adaptive weight based on the total score function f (h, r, t) defined in the step 1, and learning the vector representation of the entity and the relation by minimizing the loss function to achieve the optimization goal.
2. The adaptive weight-based multi-source information fusion knowledge graph representation learning method as claimed in claim 1, wherein in step 1, the value range of the weight λ is λ ∈ (0, 1).
3. The adaptive weight-based multi-source information fusion knowledge graph representation learning method according to claim 1, wherein in step 2, a random gradient descent method is adopted to minimize a loss function.
4. The adaptive weight-based multi-source information fusion knowledge graph representation learning method according to claim 1, wherein in the step 2, the constructed loss function L is:
wherein, [ f (h, r, t) + gamma-f (h ', r, t')]+Max (0, f (h, r, t) + γ -f (h ', r, t')); gamma is a set boundary value; (h, r, t) represents a triplet of the knowledge graph, namely a positive example triplet, h represents a head entity, t represents a tail entity, r represents a relation between the head entity and the tail entity, f (h, r, t) represents a score function of the positive example triplet, and S (h, r, t) represents a positive example triplet set; (h ', r, t ') represents a negative example triple constructed by randomly replacing the head entity h and the tail entity t, f (h ', r, t ') represents a score function of the negative example triple, and S ' (h, r, t) represents a negative example triple set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810563786.6A CN109033129B (en) | 2018-06-04 | 2018-06-04 | Multi-source information fusion knowledge graph representation learning method based on self-adaptive weight |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810563786.6A CN109033129B (en) | 2018-06-04 | 2018-06-04 | Multi-source information fusion knowledge graph representation learning method based on self-adaptive weight |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109033129A CN109033129A (en) | 2018-12-18 |
| CN109033129B true CN109033129B (en) | 2021-08-03 |
Family
ID=64612062
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810563786.6A Active CN109033129B (en) | 2018-06-04 | 2018-06-04 | Multi-source information fusion knowledge graph representation learning method based on self-adaptive weight |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109033129B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109739939A (en) * | 2018-12-29 | 2019-05-10 | 颖投信息科技(上海)有限公司 | The data fusion method and device of knowledge mapping |
| CN109739996B (en) * | 2018-12-29 | 2020-12-25 | 北京航天数据股份有限公司 | Construction method and device of industrial knowledge map |
| CN110263324B (en) * | 2019-05-16 | 2021-02-12 | 华为技术有限公司 | Text processing method, model training method and device |
| CN110232186A (en) * | 2019-05-20 | 2019-09-13 | 浙江大学 | The knowledge mapping for merging entity description, stratification type and text relation information indicates learning method |
| CN110275959B (en) * | 2019-05-22 | 2023-04-28 | 广东工业大学 | Rapid learning method for large-scale knowledge base |
| CN111159485B (en) * | 2019-12-30 | 2020-11-13 | 科大讯飞(苏州)科技有限公司 | Tail entity linking method, device, server and storage medium |
| CN111488402B (en) * | 2020-03-26 | 2023-11-28 | 天津大学 | Representation learning method of knowledge graph with hierarchical relationship structure |
| CN111680109B (en) * | 2020-04-22 | 2024-03-29 | 北京三快在线科技有限公司 | Knowledge graph representation learning model training method and device and electronic equipment |
| CN111881290A (en) * | 2020-06-17 | 2020-11-03 | 国家电网有限公司 | Distribution network multi-source grid entity fusion method based on weighted semantic similarity |
| CN112596031A (en) * | 2020-12-22 | 2021-04-02 | 电子科技大学 | Target radar threat degree assessment method based on knowledge graph |
| CN113312487A (en) * | 2021-01-16 | 2021-08-27 | 江苏网进科技股份有限公司 | Knowledge representation learning method facing legal text based on TransE model |
| CN113032582A (en) * | 2021-04-20 | 2021-06-25 | 杭州叙简科技股份有限公司 | Knowledge graph based entity unified model establishment and entity unified method |
| CN113590843B (en) * | 2021-08-06 | 2023-06-23 | 中国海洋大学 | Knowledge representation learning method integrating molecular structural features |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106250412A (en) * | 2016-07-22 | 2016-12-21 | 浙江大学 | The knowledge mapping construction method merged based on many source entities |
| CN106886543A (en) * | 2015-12-16 | 2017-06-23 | 清华大学 | The knowledge mapping of binding entity description represents learning method and system |
| CN107273490A (en) * | 2017-06-14 | 2017-10-20 | 北京工业大学 | A kind of combination mistake topic recommendation method of knowledge based collection of illustrative plates |
| CN107885760A (en) * | 2016-12-21 | 2018-04-06 | 桂林电子科技大学 | It is a kind of to represent learning method based on a variety of semantic knowledge mappings |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101306667B1 (en) * | 2009-12-09 | 2013-09-10 | 한국전자통신연구원 | Apparatus and method for knowledge graph stabilization |
-
2018
- 2018-06-04 CN CN201810563786.6A patent/CN109033129B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106886543A (en) * | 2015-12-16 | 2017-06-23 | 清华大学 | The knowledge mapping of binding entity description represents learning method and system |
| CN106250412A (en) * | 2016-07-22 | 2016-12-21 | 浙江大学 | The knowledge mapping construction method merged based on many source entities |
| CN107885760A (en) * | 2016-12-21 | 2018-04-06 | 桂林电子科技大学 | It is a kind of to represent learning method based on a variety of semantic knowledge mappings |
| CN107273490A (en) * | 2017-06-14 | 2017-10-20 | 北京工业大学 | A kind of combination mistake topic recommendation method of knowledge based collection of illustrative plates |
Non-Patent Citations (1)
| Title |
|---|
| 知识图谱技术综述;徐增林等;《电子科技大学学报》;20160731;第45卷(第4期);第589-606页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109033129A (en) | 2018-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109033129B (en) | Multi-source information fusion knowledge graph representation learning method based on self-adaptive weight | |
| CN112131404B (en) | Entity alignment method in four-risk one-gold domain knowledge graph | |
| Zhao et al. | Network schema preserving heterogeneous information network embedding | |
| CN108197290B (en) | Knowledge graph representation learning method integrating entity description and relationship description | |
| CN108763376B (en) | Knowledge representation learning method for integrating relationship path, type and entity description information | |
| CN110880019B (en) | Method for adaptively training target domain classification model through unsupervised domain | |
| CN112988917B (en) | Entity alignment method based on multiple entity contexts | |
| CN109815336B (en) | Text aggregation method and system | |
| US11216739B2 (en) | System and method for automated analysis of ground truth using confidence model to prioritize correction options | |
| CN111353545A (en) | Plant disease and insect pest identification method based on sparse network migration | |
| CN110264372B (en) | Topic community discovery method based on node representation | |
| CN114676687A (en) | Aspect level emotion classification method based on enhanced semantic syntactic information | |
| CN114969278A (en) | Knowledge enhancement graph neural network-based text question-answering model | |
| CN108052683B (en) | Knowledge graph representation learning method based on cosine measurement rule | |
| CN112507077A (en) | Event time sequence relation identification method based on relational graph attention neural network | |
| CN110597986A (en) | Text clustering system and method based on fine tuning characteristics | |
| CN114491036A (en) | Semi-supervised text classification method and system based on self-supervision and supervised joint training | |
| CN114491039A (en) | Meta-learning few-sample text classification method based on gradient improvement | |
| CN115686868A (en) | Cross-node-oriented multi-mode retrieval method based on federated hash learning | |
| CN116341655A (en) | Entity alignment method based on multi-mode collaborative representation learning | |
| Zhao et al. | Network schema preserved heterogeneous information network embedding | |
| CN114723011A (en) | High-order time sequence hypergraph convolution network operation method oriented to dynamic graph link prediction | |
| CN114842257A (en) | Robust image classification method based on multi-model anti-distillation | |
| CN117151429B (en) | Government service flow arranging method and device based on knowledge graph | |
| CN113486143A (en) | User portrait generation method based on multi-level text representation and model fusion |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181218 Assignee: Guilin Biqi Information Technology Co.,Ltd. Assignor: GUILIN University OF ELECTRONIC TECHNOLOGY Contract record no.: X2023980045831 Denomination of invention: A Learning Method for Knowledge Graph Representation of Multi source Information Fusion Based on Adaptive Weighting Granted publication date: 20210803 License type: Common License Record date: 20231107 |