CN104035917B - A kind of knowledge mapping management method and system based on semantic space mapping - Google Patents
A kind of knowledge mapping management method and system based on semantic space mapping Download PDFInfo
- Publication number
- CN104035917B CN104035917B CN201410253673.8A CN201410253673A CN104035917B CN 104035917 B CN104035917 B CN 104035917B CN 201410253673 A CN201410253673 A CN 201410253673A CN 104035917 B CN104035917 B CN 104035917B
- Authority
- CN
- China
- Prior art keywords
- semantic
- vector
- edge
- node
- model
- 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.)
- Expired - Fee Related
Links
- 238000013507 mapping Methods 0.000 title claims abstract description 35
- 238000007726 management method Methods 0.000 title claims abstract description 33
- 239000013598 vector Substances 0.000 claims abstract description 123
- 238000000034 method Methods 0.000 claims abstract description 25
- 238000010276 construction Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 11
- 238000002372 labelling Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 2
- 230000000717 retained effect Effects 0.000 claims 1
- 238000009825 accumulation Methods 0.000 abstract description 4
- 238000011524 similarity measure Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 241000251468 Actinopterygii Species 0.000 description 1
- 241000282373 Panthera pardus Species 0.000 description 1
- 241000282376 Panthera tigris Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to text semantic treatment, semantic network technology field, specially a kind of knowledge mapping management method and system based on semantic space mapping.The inventive method includes：Semantic vector builds, semantic space maps, knowledge mapping management；Knowledge mapping management is divided into including three again：Semantic Clustering, semantic duplicate removal, semantic tagger.For the side/node of knowledge mapping, its text unit will be described first and is projected to semantic space, and its vector representation on semantic space is obtained by vector accumulation；On this basis, the multinomial management role of knowledge mapping is realized；System includes that corresponding semantic vector builds, semantic space maps, knowledge mapping manages 3 modules.Carried out when semanteme compares to sensitive shortcomings of factor such as word deformation, synonym change, grammatical form changes instant invention overcomes traditional knowledge collection of illustrative plates management method, and the mode of vector accumulation can easily tackle the difference of word number, it is easy to accomplish the knowledge mapping management role such as further Semantic Clustering, semantic duplicate removal, semantic tagger.
Description
Technical Field
The invention belongs to the technical field of text semantic processing and semantic web, and particularly relates to a knowledge graph management method and system based on semantic space mapping.
Background
The construction of the knowledge graph is an important project in the big data era, and can correlate disordered data and arrange the data into structured knowledge to be provided for users, and the characteristic determines that the knowledge graph has important application in many fields, for example, the current search causes that the search is carried out based on keyword matching, and after the knowledge graph is established, after a certain keyword is input, the related information such as the attribute, the category, the relationship with other entities and the like of the keyword can be returned, so that the required information can be provided for the users more accurately and perfectly. The knowledge graph is a base stone for realizing a series of applications such as semantic search, machine automatic question answering, internet advertisement recommendation, personalized electronic reading and the like, and the size of the knowledge graph playing a role in the fields is directly determined if the knowledge graph can be effectively managed.
However, what is finally extracted in the current knowledge graph construction is a deterministic relational representation, and the deterministic description has poor adaptability under the conditions of word deformation, synonym change, grammar form change and the like, for example, two edges with similar semantics can be regarded as two completely different edges because the two edges are described by different words, and the processing mode is not only unreasonable, but also brings great difficulty to the management of the knowledge graph, such as edge/node clustering, edge/node deduplication, edge/node labeling and the like, thereby affecting the effective application of the knowledge graph.
Disclosure of Invention
The invention provides a knowledge graph management method and system based on semantic space mapping, aiming at the defects of the current knowledge graph management technical method.
For the edges/nodes (namely the relation/entity between the entities) of the knowledge graph, firstly projecting and accumulating the text units describing the edges/nodes to a semantic space so as to obtain the vector representation of the edges/nodes on the semantic space; then on the basis of text semantic vectorization, a plurality of management tasks of the knowledge graph can be further realized: the clustering method can be used and vector similarity measurement is combined to conveniently carry out semantic clustering of edges/nodes, so that relationships/entities with similar semantics are excavated; semantic duplication removal can be realized by calculating a typical edge/typical node substitution class set on the basis of semantic clustering; the automatic labeling of the relation/entity can be realized according to the semantic distance between the newly added edge/node and the labeled edge/node model.
The invention provides a knowledge graph management method based on semantic space mapping, which comprises the following specific steps: semantic vector construction, semantic space mapping and knowledge map management; wherein:
(1) the specific steps of semantic vector construction are as follows:
the semantic vector library is constructed based on the corpus, so that the text units are mapped to vectors in a semantic space, and the semantic similarity between the text units can be compared according to the distance of corresponding vectors in the semantic space, and the spatial distance of semantic vectors corresponding to words with similar semantics can be very close, so that the influences of word deformation, synonym change and grammar form change during direct comparison among the words are overcome.
The semantic vector can be calculated by various methods, such as the Word2Vec method, the esa (explicit semantic analysis) method, the lsa (late semantic analysis) method, the co-occurrence Word frequency characteristics, and so on, preferably, the Word2Vec method (https:// code.
The selection principle of the training data for constructing the semantic vector is to ensure high coverage rate and field independence by using a large-scale encyclopedic corpus, preferably, a Wikipedia knowledge base (http:// www.wikipedia. org /) is used as the corpus for training the semantic vector by using a Word2Vec method, and the semantic vector base is constructed by using the training result for other modules to use in semantic mapping.
(2) Semantic space mapping
The method maps texts representing edge nodes in a knowledge graph into vectors in a semantic space, and comprises the following specific steps:
(2.1) filtering words in edges/nodes (relationships between entities/entities) in the knowledge graph to remove stop words without semantics;
(2.2) acquiring the projection vector of each word in the semantic space from the constructed semantic vector library for each word which is reserved after the last operation processing, and then accumulating the semantic vectors corresponding to the words to further obtain the overall semantic vector representing the edge/node.
(3) The knowledge graph management is divided into four sub-steps: semantic clustering, semantic duplication removal and semantic labeling;
and (3.1) semantic clustering, which is further semantic mining based on knowledge graph construction, is very important for managing the knowledge graph and specifically comprises edge clustering (relational clustering) and node clustering (entity clustering). For edge clustering, the edges connecting different node pairs can be clustered, entity pairs with similar semantic relations are found, a plurality of edges of one node can be clustered, main related entity classes of the node are excavated, even a plurality of edges connecting the same pair of nodes can be clustered, and main relation classes among the edges are excavated; for node clustering, entities with similar semantics can be found.
The specific steps of semantic clustering are as follows:
and (3) carrying out semantic space mapping on the edge/node set to be clustered based on the constructed semantic vector library, and then further clustering the obtained semantic vectors. The clustering method may employ various methods such as a hierarchical clustering method, a Kmeans method, etc., and preferably, a hierarchical clustering method is employed. The similarity measure can be a variety of measures such as Cosine, Cityblock, Euclidean, Mahalanobis, Minkowski, Chebychev, etc., preferably using Cosine similarity.
Wherein, x and y are two vectors to be compared respectively, and Sim is a Cosine similarity result obtained by calculation.
(3.2) semantic deduplication
This is the case for knowledge maps constructed based on big data: although the specific representation forms (texts describing relations/entities) are different, the semantic contents represented by the different edges/nodes are very close to or even consistent, which results in that the knowledge graph is increased in size and is accompanied by the increase of redundant information. From the data cleaning perspective, if the edges/nodes are uniformly expressed and semantic deduplication (edge deduplication and node deduplication) is realized, the number of semantic edges/nodes (namely the number of relations/entities) is reduced and the simplified expression of the knowledge graph is realized at the same time.
The specific steps of semantic deduplication are as follows:
for the result of semantic clustering, for the edge/node set clustered in the same class, the redundancy of semantic information is reduced by calculating typical edge/typical node to replace the original class set elements, and the selection basis is as follows:
here, ViIs the semantic vector of the corresponding ith relation/entity in the set to be merged, V is the accumulated semantic vector of all relations/entities in the set to be merged, Sim (a, b) represents the similarity of the vector a and the vector b, the similarity measure can adopt various measures such as Cosine, Cityblock, Euclidean, Mahalanobis, Minkowski, Chebychev, etc., preferably, Cosine phase is adoptedSimilarity.
By selecting typical edges/typical nodes by calculation to perform relationship/entity duplication removal, the storage space of the knowledge graph is effectively reduced, and the simplified representation of the knowledge graph is realized without losing representativeness.
(3.3) semantic annotation
The semantic similarity between the input edge/node and the known edge/node model is compared, the corresponding model is judged, and then the corresponding label in the predefined known type range is attached to the model, so that the edge/node in the knowledge graph can be conveniently and uniformly expressed and managed. The semantic annotation specifically comprises the following steps:
(3.3.1) edge/node model construction:
for the clustered edges/nodes, an edge/node model (namely a relation/entity model) is constructed based on the corresponding semantic vector set, the model can be constructed by using various methods such as a mean vector model, a Gaussian model, an artificial neural network, a support vector machine and the like, and preferably, the mean vector model is used; meanwhile, the type label corresponding to each type of relation/entity is marked manually.
Wherein,m i,j is shown asiClass I the firstjThe number of the vectors is such that,n i for the number of samples in the class,is the mean vector.
After the model is built, it is added to the edge/node model library.
(3.3.2) edge/node identification
For the edge/node to be queried, after obtaining the semantic vector representation according to the steps of the semantic space mapping module, sequentially comparing the vector with the edge/node models in the relational model library, for example: for the mean vector model and the Gaussian model, the similarity between vectors can be directly compared or the probability value of the input vector belonging to the model is calculated, and the category corresponding to the highest value is taken as output after traversal; for the artificial neural network and the support vector machine, the corresponding categories are directly output.
Class of output, taking mean vector model as an exampleClassComprises the following steps:
v is the semantic vector to be identified,for the mean vector corresponding to class i edges/nodes, i ∈ { 1,2, …, N }, N is the number of models in the edge/node model library, Sim (a, b) represents the similarity of vector a and vector b, where the similarity measure can be various measures such as Cosine, Cityblock, Euclidean, Mahalanobis, Minkowski, Chebychev, etc., preferably Cosine similarity.
(3.3.3) edge/node semantic annotation
And for the category output in the last step, taking out the corresponding type label which is labeled in advance from the edge/node model library and assigning the label to the input edge/node, thereby completing the semantic labeling process.
The invention also provides a knowledge graph management system based on semantic space mapping corresponding to the method. The system consists of three modules: the system comprises a semantic vector construction module, a semantic space mapping module and a knowledge map management module. Wherein, the knowledge map management module comprises three sub-modules: the semantic clustering submodule, the semantic duplication removing submodule and the semantic labeling submodule.
The specific contents are as follows:
(1) the semantic vector construction module:
the module is used for constructing a semantic vector library based on a corpus so that text units are mapped to vectors in a semantic space, and has the advantages that semantic similarity among the text units can be compared according to the distance of corresponding vectors in the semantic space, and the distance of semantic vectors corresponding to words with similar semantics in the semantic space is very short, so that the influences of word deformation, synonym change and grammar form change during direct comparison among the words are overcome.
The selection principle of the training data for constructing the semantic vector is to ensure high coverage rate and field independence by using a large-scale encyclopedic corpus, preferably, a Wikipedia knowledge base (http:// www.wikipedia. org /) is used as the corpus for training the semantic vector by using a Word2Vec method, and the semantic vector base is constructed by using the training result for other modules to use in semantic mapping.
(2) The semantic space mapping module specifically comprises the following contents:
the module maps texts representing edge nodes in the knowledge graph into vectors in a semantic space:
(2.1) filtering words in edges/nodes (relationships between entities/entities) in the knowledge graph to remove stop words without semantics;
(2.2) acquiring the projection vector of each word in the semantic space from the constructed semantic vector library for each word which is reserved after the last operation processing, and then accumulating the semantic vectors corresponding to the words to further obtain the overall semantic vector representing the edge/node.
(3) The knowledge graph management module comprises the following specific contents:
the module is responsible for completing the management of the knowledge graph and comprises three sub-modules: the semantic clustering submodule, the semantic duplication removing submodule and the semantic labeling submodule. Respectively corresponding to 3 substeps of the knowledge-graph management step;
(3.1) semantic clustering submodule
Semantic clustering is further semantic mining based on knowledge graph construction, which is important for managing the knowledge graph and specifically comprises edge clustering (relational clustering) and node clustering (entity clustering). For edge clustering, the edges connecting different node pairs can be clustered, entity pairs with similar semantic relations are found, a plurality of edges of one node can be clustered, main related entity classes of the node are excavated, even a plurality of edges connecting the same pair of nodes can be clustered, and main relation classes among the edges are excavated; for node clustering, entities with similar semantics can be found;
(3.2) semantic deduplication submodule
This is the case for knowledge maps constructed based on big data: although the specific representation forms (texts describing relations/entities) are different, the semantic contents represented by the different edges/nodes are very close to or even consistent, which results in that the knowledge graph is increased in size and is accompanied by the increase of redundant information. From the data cleaning perspective, if the edges/nodes are uniformly expressed and semantic deduplication (edge deduplication and node deduplication) is realized, the number of semantic edges/nodes (namely the number of relations/entities) is reduced and the simplified expression of the knowledge graph is realized at the same time.
The specific content of semantic deduplication is as follows:
for the result of semantic clustering, for the edge/node set clustered in the same class, the redundancy of semantic information is reduced by calculating typical edge/typical node to replace the original class set elements, and the selection basis is as follows:
here, ViIs a semantic vector corresponding to the ith relation/entity in the set to be merged, V is a cumulative semantic vector of all relations/entities in the set to be merged, Sim (a, b) represents the similarity of the vector a and the vector b, the similarity measure here can adopt various measures such as Cosine, Cityblock, Euclidean, Mahalanobis, Minkowski, Chebychev, etc., preferably, Cosine similarity is adopted;
by selecting typical edges/typical nodes by calculation to carry out relationship/entity duplication removal, the storage space of the knowledge graph is effectively reduced, the simplified representation of the knowledge graph is realized, and the representativeness is not lost;
(3.3) semantic annotation submodule
The module judges the corresponding model by comparing the semantic similarity of the input edge/node with the known edge/node model, and then attaches the corresponding label in the predefined known type range to the model, which is beneficial to the uniform representation and management of the edges/nodes in the knowledge graph. The specific contents of the sub-modules are as follows:
(3.3.1) edge/node model construction:
for the clustered edges/nodes, an edge/node model (namely a relation/entity model) is constructed based on the corresponding semantic vector set, the model can be constructed by using various methods such as a mean vector model, a Gaussian model, an artificial neural network, a support vector machine and the like, and preferably, the mean vector model is used; meanwhile, the type label corresponding to each type of relation/entity is marked manually.
Wherein,m i,j is shown asiClass I the firstjThe number of the vectors is such that,n i for the number of samples in the class,is the mean vector.
After the model is built, it is added to the edge/node model library.
(3.3.2) edge/node identification
For the edge/node to be queried, after obtaining the semantic vector representation according to the steps of the semantic space mapping module, sequentially comparing the vector with the edge/node models in the relational model library, for example: for the mean vector model and the Gaussian model, the similarity between vectors can be directly compared or the probability value of the input vector belonging to the model is calculated, and the category corresponding to the highest value is taken as output after traversal; for the artificial neural network and the support vector machine, the corresponding categories are directly output.
Class of output, taking mean vector model as an exampleClassComprises the following steps:
v is the semantic vector to be identified,for the mean vector corresponding to class i edges/nodes, i ∈ { 1,2, …, N }, N is the number of models in the edge/node model library, Sim (a, b) represents the similarity of vector a and vector b, where the similarity measure can be various measures such as Cosine, Cityblock, Euclidean, Mahalanobis, Minkowski, Chebychev, etc., preferably Cosine similarity.
(3.3.3) edge/node semantic annotation
And for the category output in the last step, taking out the corresponding type label which is labeled in advance from the edge/node model library and assigning the label to the input edge/node, thereby completing the semantic labeling process.
The invention has the advantages of
The invention overcomes the defect that the traditional knowledge graph management method is sensitive to factors such as word deformation, synonym change, grammar form change and the like when semantic comparison is carried out by mapping the text representing the knowledge graph edges/nodes into semantic vectors, and the vector accumulation mode enables the knowledge graph management method to easily cope with the difference of the number of words, thereby being easy to realize further knowledge graph management tasks such as semantic clustering, semantic duplication removal and semantic annotation, and improving the accuracy of semantic comparison while enhancing the processing flexibility.
Drawings
FIG. 1: a system block diagram.
FIG. 2: hierarchical clustering result graphs (edge clustering). The abscissa is the serial number of the entity pair and the ordinate is the inter-class distance.
FIG. 3: hierarchical clustering result graphs (node clustering). The abscissa is the sequence number of the entity and the ordinate is the inter-class distance.
FIG. 4: semantic deduplication-typical edge selection. The abscissa is the serial number of the entity and the ordinate is the similarity.
Detailed Description
The following examples are provided to illustrate the specific embodiments of the present invention, and the system modules are processed as follows:
(1) semantic vector construction
Based on the text corpus of the whole English wiki (http:// www.wikipedia.org /), the Word2Vec is used for training, and the vector dimension output by training is 500 dimensions.
(2) Semantic space mapping
And for the words on the edge/node, after the stop words are removed, taking out the corresponding semantic vector from the trained semantic vector library, and then performing vector accumulation to obtain the semantic vector representation of the edge/node.
(3) Semantic clustering
(3.1) edge semantic clustering
An input example, the format is:
sequence number: { node 1}, { side }, { node 2}
1：{Shanghai}, {large city}, {China}
2：{ipad}, {product}, {Apple}
3：{Barack Obama}, {president}, {USA}
4：{Kindle}, {manufacture}, {Amazon}
5：{New York}, {metropolis}, {USA}
6：{Dmitry Medvedev}, {Prime Minister}, {Russia}
The hierarchical clustering result graph (edge clustering) is shown in fig. 2.
Taking the threshold value as 0.8, the clustering result is as follows:
the first type: 2. 4. the following examples of the present invention
The second type: 1. 5. mu.l of
In the third category: 3. 6. mu.l of
The clustering result is correct;
(3.2) node semantic clustering
Inputting 6 nodes:
1：{tuna}
2：{tiger}
3：{leopard}
4：{car}
5：{fish}
6：{train}
the hierarchical clustering result graph (node clustering) is shown in fig. 3.
Taking the threshold value as 0.8, the clustering result is as follows:
the first type: 1. 5. mu.l of
The second type: 2. 3. mu.l
In the third category: 4. 6. mu.l of
The clustering result is correct.
(4) Semantic deduplication
For example, two nodes in the knowledge-graph: { Bill Gates }, { Microsoft }, the following edges between them are clustered in the same class after semantic clustering:
1：{CEO}
2：{executives}
3：{president}
4：{chief executive officer}
5：{current chairman}
6：{chairman}
7：{chair}
semantic deduplication-typical edge selection, as shown in fig. 4.
Accumulating the semantic vectors of all the edges to obtain a total semantic representation vector, then sequentially calculating the similarity between each edge and the total semantic representation vector, and selecting the typical edge with the maximum similarity and the serial number of 6, namely { chairman }, so that 7 edges originally gathered into the same class are replaced by only 1 typical edge, and the purposes of simplifying and representing the knowledge graph, reducing the storage space and keeping the representativeness are achieved.
(5) Semantic annotation
For example, for a set of edges of a class of relationships that complete a cluster:
1：{large city}
2：{metropolis}
3：{megacity}
4：{major city}
5：{big cities}
6：{megacities}
7：{mega cities}
and constructing a mean vector model according to the corresponding semantic vector set, and marking the type label of the model as 'metropolitan area'.
For a newly input edge { big city }, calculating the similarity between the corresponding semantic vector and the edge model,
Sim = 0.8434
and if the threshold value is 0.8, the input edge is considered to have the same meaning as the meaning represented by the class edge, so that a model type label 'metropolian area' is assigned to the input edge, thereby completing the semantic annotation process.
Reference to the literature
[1]Tomas Mikolov, et al. Efficient Estimation of WordRepresentations in Vector Space. In Proceedings of Workshop at ICLR, 2013.
[2]Tomas Mikolov, et al. Distributed Representations of Words andPhrases and their Compositionality. In Proceedings of NIPS, 2013.
[3]Tomas Mikolov, et al. Linguistic Regularities in Continuous SpaceWord Representations. In Proceedings of NAACL HLT, 2013。
Claims (3)
1. A knowledge graph management method based on semantic space mapping is characterized by comprising the following specific steps: semantic vector construction, semantic space mapping and knowledge map management; wherein:
(1) the specific steps of semantic vector construction are as follows:
constructing a semantic vector library based on the corpus so that the text units are mapped to vectors on a semantic space;
adopting a wikipedia knowledge base as a corpus for training semantic vectors by using a Word2Vec method to construct training data of the semantic vectors, and constructing a semantic vector base by using training results;
(2) semantic space mapping
Mapping texts representing edge nodes in the knowledge graph into vectors in a semantic space, and specifically comprising the following steps:
(2.1) filtering words in edges/nodes in the knowledge graph to remove stop words without semantics;
(2.2) acquiring a projection vector of each word retained after the operation processing in the step (2.1) in a semantic space from a constructed semantic vector library, and then accumulating semantic vectors corresponding to the words to further obtain a total semantic vector representing the edge/node;
(3) knowledge graph management is divided into three sub-steps: semantic clustering, semantic duplication removal and semantic labeling;
(3.1) the specific steps of semantic clustering are as follows:
firstly, semantic space mapping is carried out on the edge/node set to be clustered based on a constructed semantic vector library, and then the obtained semantic vectors are further clustered;
(3.2) the specific steps of semantic deduplication are as follows:
for the result of semantic clustering, for the edge/node set clustered in the same class, the redundancy of semantic information is reduced by calculating typical edge/typical node to replace the original class set elements, and the selection basis is as follows:
the meaning of the formula is to choose the functionTaking k corresponding to the maximum value as Typical, wherein the Typical refers to a selected Typical edge or a Typical node;
here, VkIs the semantic vector of the corresponding kth relation/entity in the set to be merged, V is the cumulative semantic vector of all relations/entities in the set to be merged, Sim (a, b) represents the direction toSimilarity of the quantity a and the vector b;
(3.3) the semantic annotation specifically comprises the following steps:
(3.3.1) edge/node model construction:
for the clustered edges/nodes, constructing an edge/node model based on the corresponding semantic vector set;
meanwhile, manually marking a type label corresponding to each type of relation/entity;
i ∈ { 1,2, …, N } which is the mean vector corresponding to the class i edges/nodes, N being the number of models in the edge/node model library;
wherein,m i,j is shown asiClass I the firstjThe number of the vectors is such that,n i for the number of samples in the class,is a mean vector;
after the model is built, adding the edge/node model into an edge/node model library;
(3.3.2) edge/node identification
For the edge/node to be inquired, after the edge/node semantic vector representation is obtained by mapping the steps according to the semantic space, the vector is sequentially compared with edge/node models in a relational model library, wherein for an average value vector model and a Gaussian model, the similarity among the vectors can be directly compared or the probability value of the input vector belonging to the model is calculated, and the category corresponding to the highest value is taken as output after traversal; for the artificial neural network and the support vector machine, the corresponding categories are directly output;
(3.3.3) edge/node semantic annotation
And (3) for the category output in the step (3.3.2), taking out the corresponding type label which is labeled in advance from the edge/node model library and assigning the label to the input edge/node, thereby completing the semantic labeling process.
2. The knowledge-graph management method based on semantic space mapping according to claim 1, characterized in that in step (3.3.2), for the mean vector model, the output categories are:
the meaning of the formula is to choose the functionTaking i corresponding to the maximum value as Class;
v is a semantic vector to be identified, and Sim (a, b) represents the similarity of the vector a and the vector b.
3. The knowledge-graph management system based on semantic space mapping based on the method of claim 1 is characterized by comprising the following three modules: the semantic vector construction module is used for executing the step (1), the semantic space mapping module is used for executing the step (2), and the knowledge graph management module is used for executing the step (3), wherein: the knowledge graph management module comprises three sub-modules: the semantic clustering submodule is used for executing the step (3.1), the semantic duplication removing submodule is used for executing the step (3.2), and the semantic labeling submodule is used for executing the step (3.3).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410253673.8A CN104035917B (en) | 2014-06-10 | 2014-06-10 | A kind of knowledge mapping management method and system based on semantic space mapping |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410253673.8A CN104035917B (en) | 2014-06-10 | 2014-06-10 | A kind of knowledge mapping management method and system based on semantic space mapping |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104035917A CN104035917A (en) | 2014-09-10 |
| CN104035917B true CN104035917B (en) | 2017-07-07 |
Family
ID=51466688
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410253673.8A Expired - Fee Related CN104035917B (en) | 2014-06-10 | 2014-06-10 | A kind of knowledge mapping management method and system based on semantic space mapping |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104035917B (en) |
Families Citing this family (44)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10409910B2 (en) | 2014-12-12 | 2019-09-10 | Omni Ai, Inc. | Perceptual associative memory for a neuro-linguistic behavior recognition system |
| US10409909B2 (en) | 2014-12-12 | 2019-09-10 | Omni Ai, Inc. | Lexical analyzer for a neuro-linguistic behavior recognition system |
| CN104462506A (en) * | 2014-12-19 | 2015-03-25 | 北京奇虎科技有限公司 | Method and device for establishing knowledge graph based on user annotation information |
| CN104794163B (en) * | 2015-03-25 | 2018-07-13 | 中国人民大学 | Entity sets extended method |
| CN104866593B (en) * | 2015-05-29 | 2018-05-22 | 中国电子科技集团公司第二十八研究所 | A kind of database search method of knowledge based collection of illustrative plates |
| CN105550190B (en) * | 2015-06-26 | 2019-03-29 | 许昌学院 | Cross-media retrieval system towards knowledge mapping |
| CN105335519B (en) * | 2015-11-18 | 2021-08-17 | 百度在线网络技术（北京）有限公司 | Model generation method and device and recommendation method and device |
| CN105740329B (en) * | 2016-01-21 | 2019-04-05 | 浙江万里学院 | A kind of contents semantic method for digging of unstructured high amount of traffic |
| CN105808931B (en) * | 2016-03-03 | 2019-05-07 | 北京大学深圳研究生院 | A kind of the acupuncture decision support method and device of knowledge based map |
| CN105787105B (en) * | 2016-03-21 | 2019-04-19 | 浙江大学 | A kind of Chinese encyclopaedic knowledge map classification system construction method based on iterative model |
| CN105824802B (en) * | 2016-03-31 | 2018-10-30 | 清华大学 | It is a kind of to obtain the method and device that knowledge mapping vectorization indicates |
| CN105956052A (en) * | 2016-04-27 | 2016-09-21 | 青岛海尔软件有限公司 | Building method of knowledge map based on vertical field |
| CN106446148B (en) * | 2016-09-21 | 2019-08-09 | 中国运载火箭技术研究院 | A kind of text duplicate checking method based on cluster |
| CN108108345B (en) * | 2016-11-25 | 2021-08-10 | 南京尚网网络科技有限公司 | Method and apparatus for determining news topic |
| CN106776564B (en) * | 2016-12-21 | 2020-04-24 | 张永成 | Semantic recognition method and system based on knowledge graph |
| CN106874378B (en) * | 2017-01-05 | 2020-06-02 | 北京工商大学 | Method for constructing knowledge graph based on entity extraction and relation mining of rule model |
| CN106909643B (en) * | 2017-02-20 | 2020-08-14 | 同济大学 | Knowledge graph-based social media big data topic discovery method |
| CN106933985B (en) * | 2017-02-20 | 2020-06-26 | 广东省中医院 | Analysis and discovery method of core party |
| CN108959328B (en) * | 2017-05-27 | 2021-12-21 | 株式会社理光 | Knowledge graph processing method and device and electronic equipment |
| CN107038261B (en) * | 2017-05-28 | 2019-09-20 | 海南大学 | A kind of processing framework resource based on data map, Information Atlas and knowledge mapping can Dynamic and Abstract Semantic Modeling Method |
| CN107103100B (en) * | 2017-06-10 | 2019-07-30 | 海南大学 | A kind of fault-tolerant intelligent semantic searching method based on map framework |
| CN107480125B (en) * | 2017-07-05 | 2020-08-04 | 重庆邮电大学 | Relation linking method based on knowledge graph |
| CN108090165B (en) * | 2017-12-13 | 2021-12-28 | 美林数据技术股份有限公司 | Method for acquiring map change difference based on embedded graph database |
| US10884865B2 (en) | 2018-01-26 | 2021-01-05 | International Business Machines Corporation | Identifying redundant nodes in a knowledge graph data structure |
| CN108334617A (en) * | 2018-02-07 | 2018-07-27 | 大连大学 | The method of semantic-based music retrieval |
| CN109145153B (en) * | 2018-07-02 | 2021-03-12 | 北京奇艺世纪科技有限公司 | Intention category identification method and device |
| CN110851610B (en) * | 2018-07-25 | 2022-09-27 | 百度在线网络技术（北京）有限公司 | Knowledge graph generation method and device, computer equipment and storage medium |
| CN109347798A (en) * | 2018-09-12 | 2019-02-15 | 东软集团股份有限公司 | Generation method, device, equipment and the storage medium of network security knowledge map |
| CN109408811B (en) * | 2018-09-29 | 2021-10-22 | 联想(北京)有限公司 | Data processing method and server |
| CN109684625B (en) * | 2018-10-31 | 2021-01-12 | 百度在线网络技术（北京）有限公司 | Entity processing method, device and storage medium |
| CN109522551B (en) * | 2018-11-09 | 2024-02-20 | 天津新开心生活科技有限公司 | Entity linking method and device, storage medium and electronic equipment |
| CN109753909B (en) * | 2018-12-27 | 2021-08-10 | 广东人啊人网络技术开发有限公司 | Resume analysis method based on content blocking and BilSTM model |
| US10997187B2 (en) | 2019-02-22 | 2021-05-04 | General Electric Company | Knowledge-driven federated big data query and analytics platform |
| US10963518B2 (en) | 2019-02-22 | 2021-03-30 | General Electric Company | Knowledge-driven federated big data query and analytics platform |
| CN110059271B (en) * | 2019-06-19 | 2020-01-10 | 达而观信息科技（上海）有限公司 | Searching method and device applying tag knowledge network |
| CN110457486A (en) * | 2019-07-05 | 2019-11-15 | 中国人民解放军战略支援部队信息工程大学 | The people entities alignment schemes and device of knowledge based map |
| CN110795527B (en) * | 2019-09-03 | 2022-04-29 | 腾讯科技（深圳）有限公司 | Candidate entity ordering method, training method and related device |
| CN111339267A (en) * | 2020-02-17 | 2020-06-26 | 京东方科技集团股份有限公司 | Knowledge graph-based question answering method and system, computer equipment and medium |
| CN112560499B (en) * | 2020-12-11 | 2024-01-09 | 北京百度网讯科技有限公司 | Pre-training method and device for semantic representation model, electronic equipment and storage medium |
| CN112580716B (en) * | 2020-12-16 | 2023-07-11 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for identifying edge types in atlas |
| CN113011152B (en) * | 2021-03-04 | 2022-08-26 | 腾讯科技（深圳）有限公司 | Text processing method, device and equipment and computer readable storage medium |
| CN113742494B (en) * | 2021-09-06 | 2024-03-15 | 湘潭大学 | Domain text similarity calculation method and system based on label graph conversion |
| CN113987152B (en) * | 2021-11-01 | 2022-08-12 | 北京欧拉认知智能科技有限公司 | Knowledge graph extraction method, system, electronic equipment and medium |
| CN114357114A (en) * | 2022-01-04 | 2022-04-15 | 新华智云科技有限公司 | Entity cleaning method and system based on unsupervised learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101079072A (en) * | 2007-06-22 | 2007-11-28 | 中国科学院研究生院 | Text clustering element study method and device |
| CN102646113A (en) * | 2012-02-17 | 2012-08-22 | 清华大学 | Method for semantic relativity between measurement concepts based on Wikipedia |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DK200301926A (en) * | 2003-12-23 | 2005-06-24 | Eskebaek Thomas | Knowledge management system with ontology-based methods for the dissemination of knowledge and the search for knowledge |
-
2014
- 2014-06-10 CN CN201410253673.8A patent/CN104035917B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101079072A (en) * | 2007-06-22 | 2007-11-28 | 中国科学院研究生院 | Text clustering element study method and device |
| CN102646113A (en) * | 2012-02-17 | 2012-08-22 | 清华大学 | Method for semantic relativity between measurement concepts based on Wikipedia |
Non-Patent Citations (2)
| Title |
|---|
| Conceptual graph based text classification;Yi Wan et al.;《The 2014 IEEE International Conference on Progress in Informatics and Computing》;20140518;第104-107页 * |
| 基于语义事务信息聚类的用户概貌构建;杨鹤标 等;《计算机工程与设计》;20101231;第31卷(第20期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104035917A (en) | 2014-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104035917B (en) | A kind of knowledge mapping management method and system based on semantic space mapping | |
| CN104699763B (en) | The text similarity gauging system of multiple features fusion | |
| CN110162644B (en) | Image set establishing method, device and storage medium | |
| CN107205016B (en) | Retrieval method of Internet of things equipment | |
| CN104573130B (en) | The entity resolution method and device calculated based on colony | |
| CN105719191B (en) | The social groups that behavior semanteme is not known under multiscale space find method | |
| CN104298776B (en) | Search-engine results optimization system based on LDA models | |
| CN112836487B (en) | Automatic comment method and device, computer equipment and storage medium | |
| CN104199965A (en) | Semantic information retrieval method | |
| CN104239553A (en) | Entity recognition method based on Map-Reduce framework | |
| US20200272674A1 (en) | Method and apparatus for recommending entity, electronic device and computer readable medium | |
| CN104008092A (en) | Method and system of relation characterizing, clustering and identifying based on the semanteme of semantic space mapping | |
| CN111291177A (en) | Information processing method and device and computer storage medium | |
| Yuan et al. | Discovering thematic objects in image collections and videos | |
| WO2023000764A1 (en) | Target retrieval method and apparatus, device and storage medium | |
| Papadopoulos et al. | Image clustering through community detection on hybrid image similarity graphs | |
| CN112329460A (en) | Text topic clustering method, device, equipment and storage medium | |
| CN104268192A (en) | Webpage information extracting method, device and terminal | |
| CN106649250A (en) | Method and device for identifying emotional new words | |
| CN107665221A (en) | The sorting technique and device of keyword | |
| CN116881430A (en) | Industrial chain identification method and device, electronic equipment and readable storage medium | |
| Nie et al. | Cross-domain semantic transfer from large-scale social media | |
| CN110019670A (en) | A kind of text searching method and device | |
| Medvet et al. | Brand-related events detection, classification and summarization on twitter | |
| CN103136221A (en) | Method capable of generating requirement template and requirement identification method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170707 Termination date: 20200610 |