CN109299284A - A Knowledge Graph Representation Learning Method Based on Structural Information and Text Description - Google Patents
A Knowledge Graph Representation Learning Method Based on Structural Information and Text Description Download PDFInfo
- Publication number
- CN109299284A CN109299284A CN201811011812.0A CN201811011812A CN109299284A CN 109299284 A CN109299284 A CN 109299284A CN 201811011812 A CN201811011812 A CN 201811011812A CN 109299284 A CN109299284 A CN 109299284A
- Authority
- CN
- China
- Prior art keywords
- entity
- vector
- text
- description
- expression vector
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 239000013598 vector Substances 0.000 claims abstract description 75
- 239000002585 base Substances 0.000 claims abstract description 19
- 239000012458 free base Substances 0.000 claims abstract description 4
- 239000013604 expression vector Substances 0.000 claims description 36
- 238000013507 mapping Methods 0.000 claims description 33
- 238000005295 random walk Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 230000004927 fusion Effects 0.000 abstract description 2
- 238000012549 training Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000009412 basement excavation Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Landscapes
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本发明一种基于结构信息与文本描述的知识图谱表示学习方法的目的是将三元组中的实体和关系映射到低维连续的实值空间中。本发明旨在提高知识表示中实体的向量表示；从现有知识库Freebase中得到实体的对应文本描述信息，采用word2vec对每个描述进行词向量表示，然后用词加和均值向量作为该描述的向量表示，还采用doc2vec的句向量生成方式对描述进行向量表示，然后将词向量作为CNN文本编码器的输入，得到每个实体的基于描述文本的表示向量；然后在联合表示中利用权值来评估知识库中基于符号的表示向量、基于网络结构的表示向量和基于描述文本的表示向量对实体的最终表示向量的影响，完成结构信息和文本信息的融合，提高知识图谱表示的准确性。
The purpose of a knowledge graph representation learning method based on structural information and text description of the present invention is to map entities and relationships in triples into a low-dimensional continuous real-valued space. The invention aims to improve the vector representation of entities in knowledge representation; obtains the corresponding text description information of entities from the existing knowledge base Freebase, uses word2vec to represent each description with word vector, and then uses the word sum and mean vector as the description of the description. Vector representation, the description is also represented by the doc2vec sentence vector generation method, and then the word vector is used as the input of the CNN text encoder to obtain the representation vector based on the description text for each entity; Evaluate the influence of the symbol-based representation vector, the network structure-based representation vector and the description text-based representation vector in the knowledge base on the final representation vector of the entity, complete the fusion of structural information and text information, and improve the accuracy of knowledge graph representation.
Description
Technical field
Present invention relates particularly to a kind of knowledge mappings based on structural information and text description to indicate learning method.
Background technique
Knowledge mapping is important component of the NLP technology in the tasks such as intelligent answer, Web search and semantic analysis.
Knowledge mapping is often huge, complete not enough comprising hundreds of entity and billions of knowledge, but usually.So with
Knowledge mapping completion solves the Sparse Problem in knowledge mapping.Knowledge mapping often indicates with latticed form, interior joint
Entity is represented, while the relationship between two entities is represented, the shape of each knowledge triple (head entity, relationship, tail entity)
Formula indicates.Based on symbol representation method as triple, designer is necessary for different application design in knowledge mapping completion
Various nomographys, with being continuously increased since scalability is poor for knowledge mapping scale, calculating complexity becomes more and more infeasible.Separately
Outside, the problems such as KG indicated based on figure faces Sparse in the application, and the KG for scheming to indicate is very not square in machine learning
Just, and in this current big data era, machine learning is that big data automation and intelligence talk about essential technical tool.
In face of these problems, knowledge mapping indicates that study (the also referred to as flush type learning of knowledge mapping) is proposed out.Knowledge mapping
Expression learning method be intended to for the entity of knowledge mapping and relationship to be expressed as dense low-dimensional real-valued vectors, and then in low-dimensional vector
In efficiently computational entity, relationship and its between complicated association, in the building of knowledge mapping, reasoning, fusion, excavation and application
In play a significant role.
Problems Existing at present indicates that study mainly has based on the distributed TransX model indicated and based on nerve net
The model of network.Have a good performance in representation of knowledge study based on the model of translation, however it is most of these
Translation model only only accounts for the expression of the true triple symbol in knowledge mapping when carrying out vector projection, has ignored knowledge graph
Some implicit semantic informations in spectrum, this meeting prevent the vector learnt from accurately expressing the language for including in knowledge mapping
Adopted relationship.There is the text description information of a large amount of entity in existing knowledge library, the corresponding text description of ternary group object contains
Many additional semantic informations can be indicated to provide more accurate semantic expressiveness for entity in conjunction with text information, be helped simultaneously
Semantic dependency between discovery different entities.Certainly, in the existing representation method using entity description, only simply
Triple structure vector and text description vectors simple concatenation based on symbol can not more accurately be judged two kinds of letters by ground
The information in breath source finally indicates whether rationally entity in multi-C vector space.Relationship and entity in knowledge mapping are being done
It indicates all to be projected in multi-C vector space when study, individual each vector is to be difficult to explain its tool at present
Body physical significance, only just there is meaning in relative position.
Such as the triple based on text description in knowledge base Fressbase, wherein the corresponding text description of each entity
Certain information semantically is provided to the expression of entity in the triple, but is learnt in the expression of many knowledge mappings
In method, when handling these triples, the triple study based on symbol only considers the structural information table of triple itself
Show；Text based indicates that learning method simply connects structural information vector sum text information vector and splices；There is no height
Reasonable representation of the entity in vector space is improved using the semantic information in text in effect ground；Moreover, entity is in map
Opposed configuration information is not added in the expression vector of entity yet, has lost the information of entity to a certain extent.
Summary of the invention
The technical problem to be solved in the present invention is that only considering for the expression learning method of above-mentioned current knowledge mapping
True triple symbol in knowledge mapping indicates, has ignored the deficiency of some implicit semantic informations in knowledge mapping,
There is provided a kind of knowledge mapping described based on structural information and text indicates that learning method solves the above problems.
A kind of knowledge mapping expression learning method based on structural information and text description, comprising:
Step 1: triplet information is obtained from default knowledge base, each triplet information includes head entity, relationship and tail
Entity carries out processing using triplet information of the TransE learning method to acquisition and respectively obtains head reality in each triplet information
The expression vector of body, relationship and tail entity, the expression vector composition of head entity, relationship and tail entity in each triplet information breath
One expression vector based on symbol；
Step 2: each expression vector based on symbol that step 1 is obtained is stored in database, and establishes corresponding rope
Draw；
Step 3: successively each entity in the default knowledge base being inquired, inquired entity is obtained and makees respectively
Corresponding entity when for head entity and tail entity；
Step 4: the corresponding entity sets of each inquired entity are obtained according to step 3, for the entity of each inquiry:
The id that all entities for including in correspondent entity set are inquired in the default knowledge base, is obtained by the way of random walk
The id of all entities in correspondent entity set connects and composes an entity id sequence；
Step 5: each entity id sequence obtained according to specific steps 4 is respectively adopted skip-gram model learning and obtains
One expression vector based on network structure；
Step 6: the description text of each entity in the default knowledge base is pre-processed respectively；
Step 7: the description text of each entity pretreated in step 6 is divided using the CBOW method in word2vec
Text is not described and does term vector generation, respectively obtains expression term vector；
Step 8: each expression term vector obtained for step 7: term vector will be indicated as the defeated of CNN encoder
Enter, two convolutional layers, two pond layers are set, study obtains the expression vector based on description text of each entity；
Step 9: respectively by the expression vector, the expression vector sum base based on network structure based on symbol of the same entity
Spliced in the expression vector of description text, obtains the splicing vector of each entity；
Step 10: using TransE learning method, the splicing vector of each entity is learnt to obtain each entity
It is final to indicate vector.
Further, using the subclass FB15k of Freebase as the default knowledge base.
Further, pretreatment specifically includes removal stop words in step 6, and the entity name being made of multiple characters is connected
It connects as a word.
Further, in step 7 in term vector generating process, need to be arranged dimension size, the min- of each term vector
Count and sliding window value.
The present invention proposes to be added to network structure in the expression study of knowledge mapping, it is intended to solve triple symbol table
Show, text description indicates, the vectors of three information sources of Relative Network structure is effectively fused together, with improve each entity to
The expression quality of amount provides the vector source comprising more multi-semantic meaning and structural information to upper layer application, meanwhile, this patent exists for the first time
The network structure information of relative position of the entity in knowledge mapping is introduced into the representation of knowledge.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing:
Fig. 1 is that a kind of knowledge mapping based on structural information and text description of the invention indicates learning method structure chart.
Specific embodiment
For a clearer understanding of the technical characteristics, objects and effects of the present invention, now control attached drawing is described in detail
A specific embodiment of the invention.
A kind of knowledge mapping expression learning method based on structural information and text description, comprising:
Step 1: triplet information is obtained from default knowledge base Freebase, triplet information includes head entity, relationship
With tail entity, the expression vector of head entity, relationship and tail entity in triplet information is respectively obtained using TransE learning method,
Forming the expression vector (h, r, t) based on symbol, wherein h indicates that the expression vector of head entity, r indicate the expression vector of relationship,
T indicates the expression vector of tail entity, and the dimension of each vector is set as 100 dimensions；Loss function is as follows:
Step 2: the expression vector based on symbol that step 1 is obtained is stored in database, and establishes corresponding index；
Step 3: successively each entity in default knowledge base being inquired, obtains institute's query entity respectively as head reality
Corresponding entity and relationship when body and tail entity；
Step 4: the correspondent entity set of institute's query entity being obtained according to step 3, and inquires correspondent entity in the database
The id for all entities for including in set obtains the id of all entities in correspondent entity set by the way of random walk, even
It connects and constitutes entity id sequence；
Step 5: the entity id sequence obtained according to specific steps 4 is obtained using skip-gram model learning based on network
The expression vector of structure.
Step 6: entity description text is selected from default knowledge base, and the entity description text of selection is pre-processed:
Stop words is removed, the entity name being made of multiple characters is connected as a word；
Step 7: entity description text pretreated in step 6 retouches entity using the CBOW method in word2vec
It states text and does term vector generation, obtain indicating term vector；When model training, the dimension size that each term vector is arranged is
100, suitable min-count and sliding window value are set, obtain indicating term vector.
Step 8: two convolutional layers are arranged as the input of CNN encoder in the expression term vector that step 7 is obtained, and two
Pond layer, study obtain the expression vector based on description text.In view of having ignored the word order problem in text in CBOW model,
Therefore this patent is encoded using CNN as description text of the text decoder to each entity on the basis of CBOW model
Study；
Step 9: splicing is indicated vector, is indicated table of the vector sum based on description text based on network structure based on symbol
Show vector, obtain the splicing vector of entity:
E=[es:eg:ed]
Wherein, esRepresentative structure vector, egIndicate the expression vector based on network structure, edIt represents entity and is based on description text
This expression vector, e represent the splicing vector of entity；
Step 10: using TransE learning method, expression vector of the combination learning entity based on symbol, based on graph structure
Indicate that expression vector of the vector sum based on description text, score function are as follows:
F=‖ h+r-t ‖
Wherein, h, t, which respectively represent an entity and the vector of tail entity, indicates that it is corresponding based on symbol that value is equal to the entity
Indicate vector, based on network structure indicate vector sum based on description text expression vector splicing vector；This is scored
Function, which is brought into, participates in model training in the loss function of TransE, obtain the final expression vector of entity.
Wherein, [x]+Indicate to take that number that numerical value is big in 0 and x, two values, ε is hyper parameter, and h is indicated in the triple
Head entity vector, t indicates the tail entity vector in the triple, and r indicates that the relationship in the triple indicates vector,
H ', r ', t ', which then respectively indicate head entity, relationship, tail entity in the negative example triple, indicates vector.S ' expression negative the example
Triplet sets can be obtained by following formula
S '=(h ', r, t) | h ' ∈ E } ∪ (h, r, t ') | t ' ∈ E } ∪ (h, r ', t) | r ' ∈ R }
Wherein, E, R respectively represent the set of entity in knowledge base, relationship, and entity and the negative of relationship are adopted in a triple
Sample loading mode be randomly choose entity in knowledge base in non-present triple or relationship replace entity in current triple or
Head entity or the one of negative example triple of composition of tail entity are only replaced in person's relationship, sampling negative for entity every time.
The final expression vector of entity and relationship in the knowledge mapping obtained by step 10, and apply it to knowledge mapping
In completion task, the effect of model is verified.
Fig. 1, which show combined entity, indicates that vector, the expression vector sum based on graph structure are based on description text based on symbol
This expression vector general frame, from the corresponding text of entity describe term vector input text decoder obtain text describe to
Amount, the entity structure vector that the expression learning model TransE then and based on symbol is obtained and the structure based on relative position
Information vector weighted sum, which obtains vector of the entity in vector space, to be indicated, model using h+r ≈ t as learning objective, wherein
Text vector vector indicates to be obtained by text decoder Encoder training, x1,x2…,xNIn the text description of presentation-entity
The term vector of each word indicates that triple vector is obtained by the representation method TransE training based on tuple, relation to
Amount is equally obtained by TransE training, and Network vector is by the deepwalk internet startup disk that uses in the implementation steps two
Mode obtains, and combination learning is obtained by the target learning function training mentioned in specific steps 12.
The embodiment of the present invention is described with above attached drawing, but the invention is not limited to above-mentioned specific
Embodiment, the above mentioned embodiment is only schematical, rather than restrictive, those skilled in the art
Under the inspiration of the present invention, without breaking away from the scope protected by the purposes and claims of the present invention, it can also make very much
Form, all of these belong to the protection of the present invention.
Claims (4)
1. a kind of knowledge mapping based on structural information and text description indicates learning method characterized by comprising
Step 1: obtain triplet information from default knowledge base, each triplet information includes head entity, relationship and tail entity,
Processing is carried out using triplet information of the TransE learning method to acquisition and respectively obtains head entity in each triplet information, pass
It is the expression vector with tail entity, the expression vector of head entity, relationship and tail entity forms one in each triplet information breath
Expression vector based on symbol；
Step 2: each expression vector based on symbol that step 1 is obtained is stored in database, and establishes corresponding index；
Step 3: successively each entity in the default knowledge base being inquired, obtains inquired entity respectively as head
Corresponding entity when entity and tail entity；
Step 4: the corresponding entity sets of each inquired entity being obtained according to step 3, for the entity of each inquiry: in institute
The id for inquiring all entities for including in correspondent entity set in default knowledge base is stated, is corresponded to by the way of random walk
The id of all entities in entity sets connects and composes an entity id sequence；
Step 5: each entity id sequence obtained according to specific steps 4 is respectively adopted skip-gram model learning and obtains one
Expression vector based on network structure；
Step 6: the description text of each entity in the default knowledge base is pre-processed respectively；
Step 7: to the description text of each entity pretreated in step 6 using the CBOW method in word2vec respectively into
Row description text does term vector generation, respectively obtains expression term vector；
Step 8: each expression term vector obtained for step 7: will indicate term vector as the input of CNN encoder,
Two convolutional layers, two pond layers are set, and study obtains the expression vector based on description text of each entity；
Step 9: being based on retouching by the expression vector based on symbol of the same entity, the expression vector sum based on network structure respectively
The expression vector for stating text is spliced, and the splicing vector of each entity is obtained；
Step 10: using TransE learning method, the splicing vector of each entity is learnt to obtain the final of each entity
Indicate vector.
2. a kind of knowledge mapping based on structural information and text description according to claim 1 indicates learning method,
It is characterized in that, using the subclass FB15k of Freebase as the default knowledge base.
3. a kind of knowledge mapping based on structural information and text description according to claim 1 indicates learning method,
It is characterized in that, pretreatment specifically includes removal stop words in step 6, and the entity name being made of multiple characters is connected as one
A word.
4. a kind of knowledge mapping based on structural information and text description according to claim 1 indicates learning method,
It is characterized in that, in step 7 in term vector generating process, needs to be arranged dimension size, min-count and the sliding of each term vector
Window value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811011812.0A CN109299284B (en) | 2018-08-31 | 2018-08-31 | A Knowledge Graph Representation Learning Method Based on Structural Information and Text Description |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811011812.0A CN109299284B (en) | 2018-08-31 | 2018-08-31 | A Knowledge Graph Representation Learning Method Based on Structural Information and Text Description |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109299284A true CN109299284A (en) | 2019-02-01 |
| CN109299284B CN109299284B (en) | 2021-07-20 |
Family
ID=65165826
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811011812.0A Active CN109299284B (en) | 2018-08-31 | 2018-08-31 | A Knowledge Graph Representation Learning Method Based on Structural Information and Text Description |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109299284B (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109871542A (en) * | 2019-03-08 | 2019-06-11 | 广东工业大学 | A text knowledge extraction method, device, equipment and storage medium |
| CN109918162A (en) * | 2019-02-28 | 2019-06-21 | 集智学园（北京）科技有限公司 | A learnable interactive display method of massive information and high-dimensional graphics |
| CN109992673A (en) * | 2019-04-10 | 2019-07-09 | 广东工业大学 | A kind of knowledge graph generation method, apparatus, equipment and readable storage medium |
| CN110119355A (en) * | 2019-04-25 | 2019-08-13 | 天津大学 | A kind of knowledge based map vectorization reasoning common software defect modeling method |
| CN110851620A (en) * | 2019-10-29 | 2020-02-28 | 天津大学 | Knowledge representation method based on combination of text embedding and structure embedding |
| CN110955764A (en) * | 2019-11-19 | 2020-04-03 | 百度在线网络技术（北京）有限公司 | Scene knowledge graph generation method, man-machine conversation method and related equipment |
| CN111046187A (en) * | 2019-11-13 | 2020-04-21 | 山东财经大学 | One-sample knowledge graph relation learning method and system based on adversarial attention mechanism |
| CN111198950A (en) * | 2019-12-24 | 2020-05-26 | 浙江工业大学 | Knowledge graph representation learning method based on semantic vector |
| CN111581392A (en) * | 2020-04-28 | 2020-08-25 | 电子科技大学 | An automatic composition score calculation method based on sentence fluency |
| CN111680163A (en) * | 2020-04-21 | 2020-09-18 | 国网内蒙古东部电力有限公司 | A knowledge graph visualization method for electric power scientific and technological achievements |
| CN112288091A (en) * | 2020-10-30 | 2021-01-29 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Knowledge inference method based on multi-mode knowledge graph |
| CN112541589A (en) * | 2020-12-21 | 2021-03-23 | 福州大学 | Text knowledge embedding method based on AHE alignment hyperplane |
| CN112559734A (en) * | 2019-09-26 | 2021-03-26 | 中国科学技术信息研究所 | Presentation generation method and device, electronic equipment and computer readable storage medium |
| CN112784066A (en) * | 2021-03-15 | 2021-05-11 | 中国平安人寿保险股份有限公司 | Information feedback method, device, terminal and storage medium based on knowledge graph |
| CN113032415A (en) * | 2021-03-03 | 2021-06-25 | 西北工业大学 | Personalized product description generation method based on user preference and knowledge graph |
| CN113488165A (en) * | 2021-07-26 | 2021-10-08 | 平安科技（深圳）有限公司 | Text matching method, device and equipment based on knowledge graph and storage medium |
| CN113536742A (en) * | 2020-04-20 | 2021-10-22 | 阿里巴巴集团控股有限公司 | Method and device for generating description text based on knowledge graph and electronic equipment |
| CN113792544A (en) * | 2021-07-06 | 2021-12-14 | 中国地质大学（武汉） | Text emotion classification method and device considering geographic spatial distribution |
| CN114329234A (en) * | 2022-03-04 | 2022-04-12 | 深圳佑驾创新科技有限公司 | Collaborative filtering recommendation method and system based on knowledge graph |
| CN114817424A (en) * | 2022-05-27 | 2022-07-29 | 中译语通信息科技(上海)有限公司 | Graph characterization method and system based on context information |
| CN115099504A (en) * | 2022-06-29 | 2022-09-23 | 中南民族大学 | Identification method of cultural relics security risk factors based on knowledge graph completion model |
| CN116306924A (en) * | 2023-02-08 | 2023-06-23 | 西北工业大学 | Associated information mining method based on space-time hypergraph |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824802A (en) * | 2016-03-31 | 2016-08-03 | 清华大学 | Method and device for acquiring knowledge graph vectoring expression |
| US9779085B2 (en) * | 2015-05-29 | 2017-10-03 | Oracle International Corporation | Multilingual embeddings for natural language processing |
| CN107818164A (en) * | 2017-11-02 | 2018-03-20 | 东北师范大学 | A kind of intelligent answer method and its system |
| EP3367256A1 (en) * | 2017-02-28 | 2018-08-29 | Fujitsu Limited | Analysis method and analysis device |
-
2018
- 2018-08-31 CN CN201811011812.0A patent/CN109299284B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9779085B2 (en) * | 2015-05-29 | 2017-10-03 | Oracle International Corporation | Multilingual embeddings for natural language processing |
| CN105824802A (en) * | 2016-03-31 | 2016-08-03 | 清华大学 | Method and device for acquiring knowledge graph vectoring expression |
| EP3367256A1 (en) * | 2017-02-28 | 2018-08-29 | Fujitsu Limited | Analysis method and analysis device |
| CN107818164A (en) * | 2017-11-02 | 2018-03-20 | 东北师范大学 | A kind of intelligent answer method and its system |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918162A (en) * | 2019-02-28 | 2019-06-21 | 集智学园（北京）科技有限公司 | A learnable interactive display method of massive information and high-dimensional graphics |
| CN109918162B (en) * | 2019-02-28 | 2021-11-02 | 集智学园（北京）科技有限公司 | A learnable interactive display method of massive information and high-dimensional graphics |
| CN109871542B (en) * | 2019-03-08 | 2024-03-08 | 广东工业大学 | Text knowledge extraction method, device, equipment and storage medium |
| CN109871542A (en) * | 2019-03-08 | 2019-06-11 | 广东工业大学 | A text knowledge extraction method, device, equipment and storage medium |
| CN109992673A (en) * | 2019-04-10 | 2019-07-09 | 广东工业大学 | A kind of knowledge graph generation method, apparatus, equipment and readable storage medium |
| CN110119355A (en) * | 2019-04-25 | 2019-08-13 | 天津大学 | A kind of knowledge based map vectorization reasoning common software defect modeling method |
| CN112559734A (en) * | 2019-09-26 | 2021-03-26 | 中国科学技术信息研究所 | Presentation generation method and device, electronic equipment and computer readable storage medium |
| CN112559734B (en) * | 2019-09-26 | 2023-10-17 | 中国科学技术信息研究所 | Brief report generating method, brief report generating device, electronic equipment and computer readable storage medium |
| CN110851620A (en) * | 2019-10-29 | 2020-02-28 | 天津大学 | Knowledge representation method based on combination of text embedding and structure embedding |
| CN111046187A (en) * | 2019-11-13 | 2020-04-21 | 山东财经大学 | One-sample knowledge graph relation learning method and system based on adversarial attention mechanism |
| CN111046187B (en) * | 2019-11-13 | 2023-04-18 | 山东财经大学 | Sample knowledge graph relation learning method and system based on confrontation type attention mechanism |
| CN110955764A (en) * | 2019-11-19 | 2020-04-03 | 百度在线网络技术（北京）有限公司 | Scene knowledge graph generation method, man-machine conversation method and related equipment |
| CN110955764B (en) * | 2019-11-19 | 2021-04-06 | 百度在线网络技术（北京）有限公司 | Scene knowledge graph generation method, man-machine conversation method and related equipment |
| CN111198950A (en) * | 2019-12-24 | 2020-05-26 | 浙江工业大学 | Knowledge graph representation learning method based on semantic vector |
| CN111198950B (en) * | 2019-12-24 | 2021-10-15 | 浙江工业大学 | A Knowledge Graph Representation Learning Method Based on Semantic Vectors |
| CN113536742A (en) * | 2020-04-20 | 2021-10-22 | 阿里巴巴集团控股有限公司 | Method and device for generating description text based on knowledge graph and electronic equipment |
| CN111680163A (en) * | 2020-04-21 | 2020-09-18 | 国网内蒙古东部电力有限公司 | A knowledge graph visualization method for electric power scientific and technological achievements |
| CN111581392A (en) * | 2020-04-28 | 2020-08-25 | 电子科技大学 | An automatic composition score calculation method based on sentence fluency |
| CN111581392B (en) * | 2020-04-28 | 2022-07-05 | 电子科技大学 | An automatic composition score calculation method based on sentence fluency |
| CN112288091A (en) * | 2020-10-30 | 2021-01-29 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Knowledge inference method based on multi-mode knowledge graph |
| CN112288091B (en) * | 2020-10-30 | 2023-03-21 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Knowledge inference method based on multi-mode knowledge graph |
| CN112541589A (en) * | 2020-12-21 | 2021-03-23 | 福州大学 | Text knowledge embedding method based on AHE alignment hyperplane |
| CN113032415A (en) * | 2021-03-03 | 2021-06-25 | 西北工业大学 | Personalized product description generation method based on user preference and knowledge graph |
| CN113032415B (en) * | 2021-03-03 | 2024-04-19 | 西北工业大学 | Personalized product description generation method based on user preference and knowledge graph |
| CN112784066A (en) * | 2021-03-15 | 2021-05-11 | 中国平安人寿保险股份有限公司 | Information feedback method, device, terminal and storage medium based on knowledge graph |
| CN112784066B (en) * | 2021-03-15 | 2023-11-03 | 中国平安人寿保险股份有限公司 | Knowledge graph-based information feedback method, device, terminal and storage medium |
| CN113792544B (en) * | 2021-07-06 | 2023-08-29 | 中国地质大学（武汉） | Text sentiment classification method and device considering geographic space distribution |
| CN113792544A (en) * | 2021-07-06 | 2021-12-14 | 中国地质大学（武汉） | Text emotion classification method and device considering geographic spatial distribution |
| CN113488165B (en) * | 2021-07-26 | 2023-08-22 | 平安科技（深圳）有限公司 | Text matching method, device, equipment and storage medium based on knowledge graph |
| CN113488165A (en) * | 2021-07-26 | 2021-10-08 | 平安科技（深圳）有限公司 | Text matching method, device and equipment based on knowledge graph and storage medium |
| CN114329234A (en) * | 2022-03-04 | 2022-04-12 | 深圳佑驾创新科技有限公司 | Collaborative filtering recommendation method and system based on knowledge graph |
| CN114817424A (en) * | 2022-05-27 | 2022-07-29 | 中译语通信息科技(上海)有限公司 | Graph characterization method and system based on context information |
| CN115099504A (en) * | 2022-06-29 | 2022-09-23 | 中南民族大学 | Identification method of cultural relics security risk factors based on knowledge graph completion model |
| CN115099504B (en) * | 2022-06-29 | 2024-06-25 | 中南民族大学 | A method for identifying risk factors of cultural relics security based on knowledge graph completion model |
| CN116306924A (en) * | 2023-02-08 | 2023-06-23 | 西北工业大学 | Associated information mining method based on space-time hypergraph |
| CN116306924B (en) * | 2023-02-08 | 2024-08-23 | 西北工业大学 | Associated information mining method based on space-time hypergraph |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109299284B (en) | 2021-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109299284A (en) | A Knowledge Graph Representation Learning Method Based on Structural Information and Text Description | |
| CN105512209B (en) | A Biomedical Event Trigger Word Recognition Method Based on Automatic Feature Learning | |
| CN111538835B (en) | Social media emotion classification method and device based on knowledge graph | |
| CN118606900B (en) | An open vocabulary three-dimensional scene understanding method based on bimodal interaction | |
| Garg et al. | Model-driven visual analytics | |
| Li | [Retracted] Intelligent Environmental Art Design Combining Big Data and Artificial Intelligence | |
| CN116401376A (en) | Knowledge graph construction method and system for manufacturability inspection | |
| Ordonez et al. | Predicting entry-level categories | |
| CN113010690B (en) | A Method of Enhancing Entity Embedding Based on Text Information | |
| CN116561272A (en) | Open domain visual language question answering method, device, electronic equipment and storage medium | |
| CN119621921B (en) | Multi-mode knowledge question-answering retrieval method and system for specific professional field | |
| Liang et al. | A comprehensive survey and guide to multimodal large language models in vision-language tasks | |
| CN114860869A (en) | A controllable general dialogue model with intent generalization | |
| Zeng et al. | Application of human computing in image captioning under deep learning | |
| CN114372454A (en) | Text information extraction method, model training method, device and storage medium | |
| Lou | [Retracted] Evaluation of College English Teaching Quality Based on Improved BT‐SVM Algorithm | |
| Tsai et al. | An intelligent recommendation system for animation scriptwriters’ education | |
| CN115525754A (en) | A question answering method and system based on deep learning and knowledge graph | |
| Liang et al. | A semantic‐driven generation of 3D Chinese opera performance scenes | |
| Guo et al. | Research on mathematical formula knowledge base for formula recognition | |
| Guo et al. | Image Captioning for Nantong Blue Calico Through Stacked Local-Global Channel Attention Network | |
| Zhou | [Retracted] Construction of English Intelligent Translation Software Framework Based on Data Analysis Algorithm | |
| CN116661940B (en) | Component identification method, device, computer equipment and storage medium | |
| CN116737964B (en) | Artificial intelligence brain system | |
| Xu et al. | Coordinated and specific autoencoder for cross-modal retrieval |
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 |