CN109033129A - Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method - Google Patents
Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method Download PDFInfo
- Publication number
- CN109033129A CN109033129A CN201810563786.6A CN201810563786A CN109033129A CN 109033129 A CN109033129 A CN 109033129A CN 201810563786 A CN201810563786 A CN 201810563786A CN 109033129 A CN109033129 A CN 109033129A
- Authority
- CN
- China
- Prior art keywords
- entity
- indicate
- indicates
- knowledge mapping
- structured message
- 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.)
- Granted
Links
- 238000013507 mapping Methods 0.000 title claims abstract description 34
- 238000000034 method Methods 0.000 title claims abstract description 31
- 230000004927 fusion Effects 0.000 title claims abstract description 25
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 22
- 230000006870 function Effects 0.000 claims abstract description 44
- 239000013598 vector Substances 0.000 claims abstract description 23
- 230000014509 gene expression Effects 0.000 claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 abstract description 13
- 238000005457 optimization Methods 0.000 abstract description 7
- 238000013519 translation Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 240000000233 Melia azedarach Species 0.000 description 1
- 239000008186 active pharmaceutical agent Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Landscapes
- Machine Translation (AREA)
Abstract
The present invention discloses a kind of Multi-source Information Fusion knowledge mapping expression learning method based on adaptive weighting, the fusion of text information and structured message is considered first, using the model based on translation between entity vector sum relation vector, optimize scoring function by adjusting weight between the two, and by carrying out type constraint training to categorized good structured message early period, and without introducing more parameters;Then entity vector sum relation vector is associated using loss function, and optimizes the loss function, when reaching optimization aim, so that it may the vector of the vector sum relationship of each entity in learned knowledge map.The present invention solves the problems, such as that text information and structured message fusion do not account for weight in knowledge base, and the existing hierarchical information of structured message in knowledge base is utilized, more accurately connecting each other between presentation-entity and relationship, and be applied in extensive knowledge mapping.
Description
Technical field
The present invention relates to knowledge mappings and depth learning technology field, and in particular to a kind of multi-source based on adaptive weighting
Information, which merges knowledge mapping, indicates learning method.
Background technique
With the fast development of society, we slowly enter an information-based epoch.The new data of magnanimity and information are every
It is all generated in different forms.Mobile Internet nowadays at the most effective convenient and fast information acquisition platform of today's society,
Demand of the user to real information obtains also increasingly increases, and how to obtain effective information from mass data and has become various fields
The main bugbear faced.Thus knowledge mapping also comes into being.
Knowledge of the people usually in the form of network in knowledge base of organization, each node presentation-entity in network, and every
Side indicates that the relationship between two entities, the form of triple are (entity 1, relationship, entity 2).Fig. 1 is typical in knowledge mapping
Triple exemplary diagram.Wherein the node " Shakespear " " Romeo and Juliet " of ellipse representation is all entity, Lian Bianbiao
" author " shown is relationship.Therefore, most of knowledge can be indicated with triple, correspond to one in knowledge base network
Chain and two entities of link, here it is the generic representation modes of knowledge base.Recent years, deep learning in speech recognition,
Image analysis and natural language processing field obtain extensive concern.Indicate that study is intended to for the semantic information of research object being expressed as
Dense low-dimensional real-valued vectors.In the low-dimensional vector space, two object distances are closer higher with regard to declarative semantics similarity.The party
To achieving impressive progress recently, can in lower dimensional space efficient computational entity and relationship semantic relation, effective solution
Sparse Problem, makes knowledge acquisition, and the performance of fusion and reasoning is significantly improved.
The significant challenge that representation of knowledge study faces is how to realize Multi-source Information Fusion.Existing knowledge mapping
Triple structural information such as TransE etc., be indicated study merely with the triple structural information of knowledge mapping, there are also big
Other information related with knowledge is measured not to be utilized effectively as the other information of knowledge base, such as description of entity and relationship are believed
Breath, classification information etc..
Summary of the invention
The present invention indicates to be unable to fully benefit after merging present in learning method with text information for existing knowledge map
With the problem of relationship, proposing a kind of Multi-source Information Fusion knowledge based on adaptive weighting between structural model and text information
Map indicates learning method.
To solve the above problems, the present invention is achieved by the following technical solutions:
Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method, specifically includes that steps are as follows:
Step 1, the fusion that text information and structured message are balanced using adaptive weight, define text information and
Structured message is mutually related total score function f (h, r, t):
F (h, r, t)=(1- λ) (| | hd+r-td||+||hd+r-MrttS||+||MrhhS+r-td||)+λ(||Mrhh+r+Mrtt
||)
Wherein, λ indicates weight, and h indicates that head entity, t indicate that tail entity, r indicate the relationship of head entity h and tail entity t, hd
Indicate that head entity text based indicates, tdIndicate that tail entity text based indicates, hSIndicate table of the head entity based on structuring
Show, tSIndicate expression of the tail entity based on structuring, MrhIt is according to the projection matrix of head substantial definition, MrhIt is fixed according to tail entity
The projection matrix of justice;
Step 2 is based on total score function f (h, r, t) defined in step 1, establishes the text envelope based on adaptive weighting
The loss function merged with structured message is ceased, and by minimizing loss function, the vector expression of learn entity and relationship reaches
To optimization aim.
In above-mentioned steps 1, the value range of weight λ is λ ∈ (0,1).
In above-mentioned steps 2, loss function is minimized using stochastic gradient descent method.
In above-mentioned steps 2, constructed loss function L are as follows:
Wherein, [f (h, r, t)+γ-f (h', r, t')]+=max (0, f (h, r, t)+γ-f (h', r, t'));γ is to set
Fixed boundary value;(h, r, t) indicates triple, that is, positive example triple of knowledge mapping, and h indicates that head entity, t indicate tail entity, r
Indicate the relationship between head entity and tail entity, f (h, r, t) indicates that the scoring function of positive example triple, S (h, r, t) indicate just
Example triplet sets;(h', r, t') indicates that random replacement turns around negative example triple constructed by entity h and tail entity t, f (h',
R, t') indicate that the scoring function of negative example triple, S ' (h, r, t) indicate negative example triplet sets.
Compared with prior art, the present invention considers the fusion of text information and structured message first, using entity to
Model based on translation between amount and relation vector, optimizes scoring function by adjusting weight between the two, and by pair
Early period, categorized good structured message carried out type constraint training, and without introducing more parameters;Then loss is utilized
Function associates entity vector sum relation vector, and optimizes the loss function, when reaching optimization aim, so that it may learn
Obtain the vector of the vector sum relationship of each entity in knowledge mapping.The present invention solves text information and structuring letter in knowledge base
Breath merges the problem of not accounting for weight, and the existing hierarchical information of structured message in knowledge base is utilized, more accurately table
Show connecting each other between entity and relationship, and is applied in extensive knowledge mapping.
Detailed description of the invention
Fig. 1 is the exemplary diagram of relationship triple in knowledge mapping.
Fig. 2 is the flow diagram that knowledge mapping of the present invention indicates learning method.
Fig. 3 a is the exemplary diagram for indicating the triple that learning method obtains according to existing knowledge mapping and indicating.
Fig. 3 b is the exemplary diagram for indicating the triple that learning method obtains according to knowledge mapping of the present invention and indicating.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific example, and referring to attached
Figure, the present invention is described in more detail.
The fusion of text information and structured message is only accounted in view of the prior art, there is no adequately consider how to adjust
The weight of section between the two is best to achieve the effect that, and does not use the existing level letter of structured message in knowledge base
Breath, and learning parameter number is more, thus connection that can not accurately between presentation-entity and relationship, it can not be well by it
It is applied in large-scale knowledge mapping.
The present invention adequately considers the fusion of the text information and structured message of adaptive weighting, and according to no benefit
Hierarchical information enriches structured message.When text information and structured message merge, proposition one adaptive
Weight to balance the fusion of text information and structured message, and type constraint is carried out to the structured message classified in advance
Training minimizes scoring function by such method, achievees the purpose that optimization aim.By the way that the more of adaptive weighting are added
Metamessage fusion method solves the heterogeneity and disequilibrium of entity and relationship in knowledge base, more accurately presentation-entity and pass
System and its between connect each other, and be applied in extensive knowledge mapping.
Specifically, a kind of Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method, such as Fig. 2 institute
Show, includes the following steps:
Step 1, when text information and structured message merge, propose an adaptive weight to balance text
The fusion of information and structured message, and type constraint training is carried out to the structured message classified in advance.
Fusion based on text information and structured message defines the total score function f that is mutually related:
F (h, r, t)=(1- λ) fD(h,r,t)+λfS(h,r,t)
FD (h, r, t) indicates the scoring function based on text 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) indicates the scoring function based on structured representation:
fS(h, r, t)=| | Mrhh+r+Mrtt||
Wherein, λ indicates weight, and λ ∈ (0,1), h indicate that head entity, t indicate that tail entity, r indicate head entity h and tail entity t
Relationship, hdIndicate that head entity text based indicates, tdIndicate that tail entity text based indicates, hSIndicate that head entity is based on
The expression of structuring, tSIndicate expression of the tail entity based on structuring, MrhIt is according to the projection matrix of head substantial definition, MrhIt is root
According to the projection matrix of tail substantial definition.
Step 2 is proposed a loss function merged based on the text information of adaptive weighting with structured message, and led to
Minimum loss function is crossed, the vector expression of the entity that learns, relationship reaches optimization aim.
Step 21 defines loss function are as follows:
Wherein, [f (h, r, t)+γ-f (h', r, t')]+=max (0, f (h, r, t)+γ-f (h', r, t'))+;γ is to set
Fixed boundary value;(h, r, t) indicates triple, that is, positive example tuple of knowledge mapping, and h indicates that head entity, t indicate tail entity, r table
Show the relationship of an entity h and tail entity t, f (h, r, t) indicates that the scoring function of positive example triple, S (h, r, t) indicate positive example three
Tuple-set;Negative example triple constructed by the head entity h and tail entity t that (h', r, t') expression replaces immediately, f (h', r,
T' the scoring function of negative example triple, S'(h, r, t) are indicated) indicate negative example triplet sets;
Step 22 minimizes loss function using stochastic gradient descent method, and study obtains each entity in knowledge mapping
Vector sum relation vector and its between connect each other.
The process for minimizing loss function is to minimize the process of scoring function, and the process minimized is exactly to reach excellent
Change the process of target.E in triple scoring function is using the energy function in TransE model, then minimizing loss
During function, when the type of relationship r is simple relation Class1-1 or complex relationship Class1-N, N-1, N-N, by not
Disconnected adjustment h, t and r, keep h+r as equal with t as possible.
Thus method learns and obtains the knowledge mapping expression learning method of the Multi-source Information Fusion based on adaptive weighting,
And the model for carrying out type constraint training to the structured message classified in advance is more accurate effectively.
Be added text information can solve the insurmountable problem of existing method: when predict an emerging entity (not
Trained entity) when, original method can be come to indicate to its vector at random, in this way its scoring function and process instruction
The scoring function of white silk entity is compared will be very poor, its loss function can also become larger, so that the effect of prediction also can be very poor.At me
It is unmodified be added text information method, when occur a new entity (untrained entity), structural method
It can come to indicate to its vector at random, but have the text description to this novel entities in knowledge base, we pass through new
The text description of entity can handle into the vector of text representation, by the vector sum text representation of training structure method to
The Calais Liang Xiang obtains new scoring function, to reach optimization aim.Occurring, an emerging entity is (untrained
Entity) when, although text information, which is added, optimizes scoring function by text description, it is added by the two scoring function
It obtains in the method for new scoring function, structured message is still able to provide the information of mistake, and specific gravity is very big, this is clearly
Unreasonable.
The invention proposes the fusions of structured message and text information in adaptive weighting.Pass through the training time of entity
Number indicates come the weights both updated, when the weight that the every trained primary structure information of entity indicates suitably increases a bit, and it is literary
The weight that this information indicates then is reduced a bit.This is because when no addition text information, entity frequency of training is also in
Entity and more with multi-class entity frequency of occurrence is commonly used in long-tail distribution, and opposite frequency of training is also enough, then frequency of training
Enough entities more close will correctly indicate that the expression of the few entity of frequency of training also can be relatively weak.In this case I
Think that the more entity of frequency of occurrence is good enough in the expression of structured message, we can be real in this way occurring
The weight of the structured representation part of body increases, and frequency of occurrence is more, and the weight that structured message indicates can also increase with number
Increase more.Instead, the less entity of frequency of occurrence, it is not good enough in the expression of structured message, then we are just at these
The text information of entity indicates that the weight of part increases, if sporocarp does not occur once, then in structured message part
Weight is also 0, that is, all indicated with text information, text information has been utilized both in this way to indicate the novel entities not occurred,
The random structured message for assigning vector has also been filtered out simultaneously.
Fig. 3 a is the exemplary diagram for indicating the triple that learning method obtains according to existing knowledge mapping and indicating.In Fig. 3 a
In, do not account for the hierarchical information of knowledge mapping triple.Hierarchical information refers to that under different scenes, entity may have not
Same role, for example Shakespear is both writer and musician, Bao Bai also has the property that.It is considered that possessing a variety of
The entity of type should have different expressions under different relationships.We construct certain types of projection matrix from hierarchical structure
Mr, then head entity h and tail entity t are indicated by the specific projection matrix of construction.Entity has how many kinds of relationship in this way
Will how many kind mapping to respectively indicate special representation of this entity under every kind of relationship.In fig 3b, we are entity
Type showed by particular kind of relationship, the entity in training with same type tend to a cluster and have it is similar
It indicates, in fact this is also the main reason for causing error in entity prediction.Selection can be improved in the present invention in particular kind of relationship
The training probability of entity under type information with same type, carrys out optimization aim in this way.
The present invention solve entity in the prior art and relationship disequilibrium and heterogeneity and parameter it is excessive and cause
Calculating it is excessively complicated, can not indicate connecting each other between entity and relationship in knowledge mapping well and cannot be very
It is applied to the problems in extensive knowledge mapping well, there is good practicability.
It should be noted that although the above embodiment of the present invention be it is illustrative, this be not be to the present invention
Limitation, therefore the invention is not limited in above-mentioned specific embodiment.Without departing from the principles of the present invention, all
The other embodiment that those skilled in the art obtain under the inspiration of the present invention is accordingly to be regarded as within protection of the invention.
Claims (4)
1. the Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method, characterized in that specifically include step
It is as follows:
Step 1, the fusion that text information and structured message are balanced using adaptive weight define text information and structure
Change information to be mutually related total score function f (h, r, t):
F (h, r, t)=(1- λ) (| | hd+r-td||+||hd+r-MrttS||+||MrhhS+r-td||)+λ(||Mrhh+r+Mrtt||)
Wherein, λ indicates weight, and h indicates that head entity, t indicate that tail entity, r indicate the relationship of head entity h and tail entity t, hdIt indicates
Head entity text based expression, tdIndicate that tail entity text based indicates, hSIndicate expression of the head entity based on structuring,
tSIndicate expression of the tail entity based on structuring, MrhIt is according to the projection matrix of head substantial definition, MrhIt is according to tail substantial definition
Projection matrix;
Step 2, based on total score function f (h, r, t) defined in step 1, establish text information based on adaptive weighting with
The loss function of structured message fusion, and by minimizing loss function, the vector expression of learn entity and relationship reaches excellent
Change target.
2. the Multi-source Information Fusion knowledge mapping according to claim 1 based on adaptive weighting indicates learning method,
It is characterized in, in step 1, the value range of weight λ is λ ∈ (0,1).
3. the Multi-source Information Fusion knowledge mapping according to claim 1 based on adaptive weighting indicates learning method,
It is characterized in, in step 2, loss function is minimized using stochastic gradient descent method.
4. the Multi-source Information Fusion knowledge mapping according to claim 1 based on adaptive weighting indicates learning method,
It is characterized in, in step 2, constructed loss function L are as follows:
Wherein, [f (h, r, t)+γ-f (h', r, t')]+=max (0, f (h, r, t)+γ-f (h', r, t'));γ is the side of setting
Dividing value;(h, r, t) indicates triple, that is, positive example triple of knowledge mapping, and h indicates that head entity, t indicate that tail entity, r indicate head
Relationship between entity and tail entity, f (h, r, t) indicate that the scoring function of positive example triple, S (h, r, t) indicate positive example ternary
Group set;(h', r, t') indicates that random replacement turns around negative example triple constructed by entity h and tail entity t, f (h', r, t') table
Show that the scoring function of negative example triple, S ' (h, r, t) indicate negative example triplet sets.
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 true CN109033129A (en) | 2018-12-18 |
CN109033129B 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) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109739996A (en) * | 2018-12-29 | 2019-05-10 | 北京航天数据股份有限公司 | A kind of construction method and device of industry knowledge mapping |
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 |
CN110263324A (en) * | 2019-05-16 | 2019-09-20 | 华为技术有限公司 | Text handling method, model training method and device |
CN110275959A (en) * | 2019-05-22 | 2019-09-24 | 广东工业大学 | A kind of Fast Learning method towards large-scale knowledge base |
CN111159485A (en) * | 2019-12-30 | 2020-05-15 | 科大讯飞(苏州)科技有限公司 | Tail entity linking method, device, server and storage medium |
WO2020135048A1 (en) * | 2018-12-29 | 2020-07-02 | 颖投信息科技(上海)有限公司 | Data merging method and apparatus for knowledge graph |
CN111488402A (en) * | 2020-03-26 | 2020-08-04 | 天津大学 | Representation learning method with hierarchical relation structure knowledge graph |
CN111680109A (en) * | 2020-04-22 | 2020-09-18 | 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 |
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 |
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 |
CN113590843A (en) * | 2021-08-06 | 2021-11-02 | 中国海洋大学 | Knowledge representation learning method fusing molecular structure characteristics |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110137919A1 (en) * | 2009-12-09 | 2011-06-09 | Electronics And Telecommunications Research Institute | Apparatus and method for knowledge graph stabilization |
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 |
-
2018
- 2018-06-04 CN CN201810563786.6A patent/CN109033129B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110137919A1 (en) * | 2009-12-09 | 2011-06-09 | Electronics And Telecommunications Research Institute | Apparatus and method for knowledge graph stabilization |
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 |
---|
徐增林等: "知识图谱技术综述", 《电子科技大学学报》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020135048A1 (en) * | 2018-12-29 | 2020-07-02 | 颖投信息科技(上海)有限公司 | Data merging method and apparatus for knowledge graph |
CN109739996A (en) * | 2018-12-29 | 2019-05-10 | 北京航天数据股份有限公司 | A kind of construction method and device of industry 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 |
CN110263324A (en) * | 2019-05-16 | 2019-09-20 | 华为技术有限公司 | Text handling 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 |
CN110275959A (en) * | 2019-05-22 | 2019-09-24 | 广东工业大学 | A kind of Fast Learning method towards large-scale knowledge base |
CN111159485A (en) * | 2019-12-30 | 2020-05-15 | 科大讯飞(苏州)科技有限公司 | Tail entity linking method, device, server and storage medium |
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 |
CN111488402A (en) * | 2020-03-26 | 2020-08-04 | 天津大学 | Representation learning method with hierarchical relation structure knowledge graph |
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 |
CN111680109A (en) * | 2020-04-22 | 2020-09-18 | 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 |
CN113590843A (en) * | 2021-08-06 | 2021-11-02 | 中国海洋大学 | Knowledge representation learning method fusing molecular structure characteristics |
CN113590843B (en) * | 2021-08-06 | 2023-06-23 | 中国海洋大学 | Knowledge representation learning method integrating molecular structural features |
Also Published As
Publication number | Publication date |
---|---|
CN109033129B (en) | 2021-08-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109033129A (en) | Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method | |
CN112131404B (en) | Entity alignment method in four-risk one-gold domain knowledge graph | |
CN110825881B (en) | Method for establishing electric power knowledge graph | |
CN108509519B (en) | General knowledge graph enhanced question-answer interaction system and method based on deep learning | |
CN108492200B (en) | User attribute inference method and device based on convolutional neural network | |
WO2022057669A1 (en) | Method for pre-training knowledge graph on the basis of structured context information | |
CN108197290A (en) | A kind of knowledge mapping expression learning method for merging entity and relationship description | |
CN110232186A (en) | The knowledge mapping for merging entity description, stratification type and text relation information indicates learning method | |
CN107885760B (en) | Knowledge graph representation learning method based on multiple semantics | |
WO2020143163A1 (en) | Named entity recognition method and apparatus based on attention mechanism, and computer device | |
CN108509425A (en) | Chinese new word discovery method based on novelty | |
CN108509654A (en) | The construction method of dynamic knowledge collection of illustrative plates | |
CN114757182A (en) | BERT short text sentiment analysis method for improving training mode | |
CN110457689A (en) | Semantic processes method and relevant apparatus | |
CN114090783A (en) | Heterogeneous knowledge graph fusion method and system | |
CN110647904A (en) | Cross-modal retrieval method and system based on unmarked data migration | |
CN112163607B (en) | Network social media emotion classification method based on multi-dimensional and multi-level joint modeling | |
CN113360582B (en) | Relation classification method and system based on BERT model fusion multi-entity information | |
CN115099409A (en) | Text-image enhanced multi-mode knowledge map embedding method | |
CN114676687A (en) | Aspect level emotion classification method based on enhanced semantic syntactic information | |
CN108052683B (en) | Knowledge graph representation learning method based on cosine measurement rule | |
CN110096713A (en) | A kind of Laotian organization names recognition methods based on SVM-BiLSTM-CRF | |
CN113901208A (en) | Method for analyzing emotion tendentiousness of intermediate-crossing language comments blended with theme characteristics | |
CN115686868A (en) | Cross-node-oriented multi-mode retrieval method based on federated hash learning | |
CN115114409A (en) | Civil aviation unsafe event combined extraction method based on soft parameter sharing |
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 |