CN108776684A - Optimization method, device, medium, equipment and the system of side right weight in knowledge mapping - Google Patents
Optimization method, device, medium, equipment and the system of side right weight in knowledge mapping Download PDFInfo
- Publication number
- CN108776684A CN108776684A CN201810560627.0A CN201810560627A CN108776684A CN 108776684 A CN108776684 A CN 108776684A CN 201810560627 A CN201810560627 A CN 201810560627A CN 108776684 A CN108776684 A CN 108776684A
- Authority
- CN
- China
- Prior art keywords
- side right
- knowledge mapping
- right weight
- node
- optimization
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of optimization methods of side right weight in knowledge mapping, including：Knowledge mapping is defined, knowledge mapping includes the original side right weight of the directed edge and directed edge between node, node；Query Information is received from user；The candidate answers of the corresponding Query Information of search, are ranked up candidate answers using knowledge mapping and are pushed to user；The feedback information to the candidate answers after sequence is received from user；Signomial geometric programming problem is built, the constraint function of signomial geometric programming problem is set based on feedback information, and the object function of signomial geometric programming problem is the function for optimizing side right weight；It solves signomial geometric programming problem and obtains optimization side right weight.The side right re-optimization problem of knowledge mapping is converted to a signomial geometric programming problem by the present invention using the feedback information of user, efficiently can automatically improve the accuracy of side right weight and the quality of knowledge mapping.The invention further relates to optimization device, medium, equipment and the systems of side right weight in a kind of knowledge mapping.
Description
Technical field
The present invention relates to knowledge mappings to optimize field, and in particular to the optimization method of side right weight, dress in a kind of knowledge mapping
It sets, medium, equipment and system.
Background technology
In modern society, knowledge mapping has been widely used in various applications, such as question answering system (Q&A) system, pushes away
Recommend system, network search engines and accurate medicine etc..For example, knowledge based collection of illustrative plates carrys out the phase between computational problem and help document
Like degree, the method for effectively finding best document answer is had proved to be.
In knowledge mapping, side (edge) represents the related information between two entity nodes, and associated intensity
Usually indicated by the weight on side.The information of side right weight can show very high value in knowledge mapping application.For example,
In the application of one technical support question answering system, " computer crash " problem taste be as caused by " low memory ", then this two
The weight of incidence edge should bigger between a entity so that when user inquires computer crash, more by low memory and its correlation
Solution recommends user.For another example at one based in the diagnostic application of medical knowledge collection of illustrative plates, multiple diseases may cause together
One symptom, but the side right between them should be different again, it in this way being capable of the more accurate diagnosis of help system progress.
Obviously, it is the construction of knowledge mapping and the key challenge of maintenance that the weight on side, which how is arranged,.Existing determining knowledge
In collection of illustrative plates the method for side right weight be based primarily upon direct correlation relationship (such as hyperlink between webpage) between two entities or
Utilize relevant statistical information between two entities.However, these methods be highly susceptible to the existing mistake of source data itself or
The influence of mistake of statistics leads to the accuracy of side right weight and the poor quality of knowledge mapping.
Invention content
In order to solve the problems, such as above-mentioned all or part, the present invention provide the optimization method of side right weight in a kind of knowledge mapping,
Device, medium, equipment and system.
Embodiments of the present invention disclose a kind of optimization method of side right weight in knowledge mapping, and this method includes
Knowledge mapping is defined, knowledge mapping includes the original side right weight of the directed edge and directed edge between node, node；
Query Information is received from user；
The candidate answers of the corresponding Query Information of search, are ranked up candidate answers using knowledge mapping, and will be after sequence
Candidate answers be pushed to user；
The feedback information to the candidate answers after sequence is received from user；
Signomial geometric programming problem is built, the constraint function of signomial geometric programming problem is set based on feedback information, symbol
The object function of geometric programming problem is the function for optimizing side right weight；
It solves signomial geometric programming problem and obtains optimization side right weight.
In a demonstration example, using knowledge mapping to candidate answers be ranked up including：
Query Information is mapped on knowledge mapping and obtains query node；
Candidate answers are mapped on knowledge mapping and obtain candidate answers node；
Calculate the similarity between query node and candidate answers node；
Candidate answers are arranged according to the descending of similarity.
In a demonstration example, the similarity between query node and candidate answers node passes through following calculating：
Wherein, φ (vq,va) indicate query node vqWith candidate answers node vaBetween similarity, z:It indicates
From query node vqTo candidate answers node vaHop count is less than the path of threshold value, | z | indicate that the length of path z, P [z] they are path z
Probability, c is constant parameter, w (vi,vi+1) indicate knowledge mapping in viNode is directed toward vi+1The side right weight of node, w (vq,v1)
Indicate query node vqThe v being directed toward in knowledge mapping1The side right weight of node, w (vk,va) indicate knowledge mapping in vkNode is directed toward
Candidate answers node vaSide right weight.
In a demonstration example, feedback information includes the optimum answer that user selects from the candidate answers after sequence；
Similarity of the constraint function between the corresponding optimum answer node of optimum answer and query node is waited more than other
Select the similarity between answer node and query node；
Object function is so that the quadratic sum of the side right weight variable quantity of each directed edge is minimum, and the side right of every directed edge becomes again
Change amount is the difference of the optimization side right weight and original side right weight of this directed edge.
In a demonstration example, this method further includes：
Clustering processing is carried out to feedback information and obtains multiple clusters；
The corresponding optimization side right of each cluster, which is merged, according to fusion rule restores final optimization side right weight, it is each to cluster
Corresponding optimization side right is obtained by solving the signomial geometric programming problem for the feedback information structure for including based on the cluster again.
In a demonstration example, the set for the directed edge being related to based on feedback information carries out clustering processing to feedback information, instead
The collection for the directed edge that feedforward information is related to is combined into each candidate answers that the query node being related to from feedback information is related to feedback information
Node hop count is less than the set of the directed edge of all paths process of threshold value.
In a demonstration example, fusion rule is, for a directed edge,
If an optimization side right weight of a corresponding cluster is only existed, as the optimization side right of this directed edge
Weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and to be all higher than this again oriented for multiple optimization side rights
The maximum optimization side right recast of side right weight incrementss for making this directed edge is then this directed edge by the original side right weight on side
Optimization side right weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and respectively less than this is oriented again for multiple optimization side rights
The maximum optimization side right recast of side right weight decrement for making this directed edge is then this directed edge by the original side right weight on side
Optimization side right weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and only side right is great has in this for part optimization
To the original side right weight on side, then the weighted average and weight that calculate the side right weight variable quantity of this directed edge are each cluster
Including feedback information number, if weighted average is just, then the side right weight incrementss of this directed edge will be made maximum
Optimization side right recast be this directed edge optimization side right weight, if weighted average is negative, then will make this directed edge
Side right weight decrement it is maximum optimization side right recast be this directed edge optimization side right weight.
In a demonstration example, using the computational methods of text similarity, the time of corresponding Query Information is searched for from corpus
Select answer.
Embodiments of the present invention also disclose a kind of optimization device of side right weight in knowledge mapping, which includes：
Knowledge mapping definition module, defines knowledge mapping, and knowledge mapping includes directed edge between node, node and has
To the original side right weight on side；
Query Information receiving module receives Query Information from user；
Candidate answers pushing module searches for the candidate answers of corresponding Query Information, using knowledge mapping to candidate answers into
Row sequence, and the candidate answers after sequence are pushed to user；
Feedback information receiving module receives the feedback information to the candidate answers after sequence from user；
Signomial geometric programming problem builds module, builds signomial geometric programming problem, the constraint of signomial geometric programming problem
Function is set based on feedback information, and the object function of signomial geometric programming problem is the function for optimizing side right weight；
Signomial geometric programming problem solver module solves signomial geometric programming problem and obtains optimization side right weight.
Embodiments of the present invention also disclose a kind of non-volatile memory medium, are stored with knowledge graph on a storage medium
The optimization program of side right weight in spectrum, the optimization program of side right weight is computer-executed to implement side in knowledge mapping in knowledge mapping
The optimization method of weight, the program include：
Knowledge mapping definition instruction, defines knowledge mapping, and knowledge mapping includes directed edge between node, node and has
To the original side right weight on side；
Query Information receives instruction, and Query Information is received from user；
Candidate answers push instruction, searches for the candidate answers of corresponding Query Information, using knowledge mapping to candidate answers into
Row sequence, and the candidate answers after sequence are pushed to user；
Feedback information receives instruction, and the feedback information to the candidate answers after sequence is received from user；
The structure instruction of signomial geometric programming problem, builds signomial geometric programming problem, the constraint of signomial geometric programming problem
Function is set based on feedback information, and the object function of signomial geometric programming problem is the function for optimizing side right weight；
Signomial geometric programming problem solving instructs, and solves signomial geometric programming problem and obtains optimization side right weight.
Embodiments of the present invention also disclose a kind of optimization equipment of side right weight in knowledge mapping, including：
Memory is stored with the optimization program of side right weight in the knowledge mapping that computer can execute；And
Processor is connected to memory, and be configured as execute knowledge mapping in side right weight optimization program with：
Knowledge mapping is defined, knowledge mapping includes the original side right weight of the directed edge and directed edge between node, node；
Query Information is received from user；
The candidate answers of the corresponding Query Information of search, are ranked up candidate answers using knowledge mapping, and will be after sequence
Candidate answers be pushed to user；
The feedback information to the candidate answers after sequence is received from user；
Signomial geometric programming problem is built, the constraint function of signomial geometric programming problem is set based on feedback information, symbol
The object function of geometric programming problem is the function for optimizing side right weight；
It solves signomial geometric programming problem and obtains optimization side right weight.
Embodiments of the present invention also disclose a kind of optimization of side right weight in system, including knowledge mapping as described above
Device.
Compared with prior art, the main distinction and its effect are embodiment of the present invention：Utilize the feedback information of user
The side right re-optimization problem of knowledge mapping is converted to a signomial geometric programming problem, efficiently can automatically improve side
The accuracy of weight and the quality of knowledge mapping.
Description of the drawings
Fig. 1 is the structural schematic diagram according to the optimization device of side right weight in the knowledge mapping of embodiment of the present invention；
Fig. 2 is the flow diagram according to the optimization method of side right weight in the knowledge mapping of embodiment of the present invention；
Fig. 3 is the exemplary schematic diagram of side right re-optimization according to embodiment of the present invention.
Specific implementation mode
In the following description, in order to make the reader understand this application better, many technical details are proposed.But this
The those of ordinary skill in field is appreciated that even if without these technical details and many variations based on the following respective embodiments
And modification, each claim of the application technical solution claimed can also be realized.
To make the object, technical solutions and advantages of the present invention clearer, the implementation below in conjunction with attached drawing to the present invention
Mode is described in further detail.
Fig. 1 is according to the structural schematic diagram of the optimization device of side right weight in the knowledge mapping of embodiment of the present invention, side right
Re-optimization device 100 include knowledge mapping definition module 101, Query Information receiving module 102, candidate answers pushing module 103,
Feedback information receiving module 104, signomial geometric programming problem structure module 105 and signomial geometric programming problem solver module
106.Fig. 2 is according to the flow diagram of the optimization method of side right weight in the knowledge mapping of embodiment of the present invention, such as Fig. 2 institutes
Show, the optimization method of side right weight specifically includes in knowledge mapping：
Step 201, knowledge mapping definition module 101 defines knowledge mapping, and knowledge mapping includes having between node, node
To side and the original side right weight of directed edge；
Step 202, Query Information receiving module 102 receives Query Information from user；
Step 203, the candidate answers of the corresponding Query Information of the search of candidate answers pushing module 103, utilize knowledge mapping pair
Candidate answers are ranked up, and the candidate answers after sequence are pushed to user；
Step 204, feedback information receiving module 104 receives the feedback information to the candidate answers after sequence from user；
Step 205, signomial geometric programming problem structure module 105 builds signomial geometric programming problem, signomial geometric programming
The constraint function of problem is set based on feedback information, and the object function of signomial geometric programming problem is the function for optimizing side right weight；
Step 206, signomial geometric programming problem solver module 106 solves signomial geometric programming problem and obtains optimization side right
Weight.
Wherein, in step 201, knowledge mapping can be defined as to oriented to have weight map G=(V, E, W), wherein V be node
Collection, E is directed edge collection, and W is the weight sets of directed edge.
Wherein, the computational methods that text similarity can be utilized in step 203 search for corresponding Query Information from corpus
Candidate answers；It can specifically include in addition, being ranked up to candidate answers using knowledge mapping：
1) Query Information is mapped on knowledge mapping and obtains query node；
2) candidate answers are mapped on knowledge mapping and obtain candidate answers node；
3) similarity between query node and each candidate answers node is calculated；
4) the corresponding each candidate answers of each candidate answers node are arranged according to the descending of similarity.
Similarity between query node and candidate answers node can be calculated by following equation:
Wherein, φ (vq,va) indicate query node vqWith candidate answers node vaBetween similarity, z:It indicates
From query node vqTo candidate answers node vaHop count be less than threshold value L path, | z | indicate path z length (process it is oriented
The item number on side), P [z] is the probability of path z, and c is the constant parameter in personalization PageRank, w (vi,vi+1) indicate knowledge graph
V in spectrumiNode is directed toward vi+1The side right weight of node, w (vq,v1) indicate query node vqThe v being directed toward in knowledge mapping1Node
Side right weight, w (vk,va) indicate knowledge mapping in vkNode is directed toward candidate answers node vaSide right weight；For example, working as | z |=2
When, P [z]=w (vq,v1)w(vk,va)。
For step 205 and step 206, can be advised using the negative feedback information of single user to build symbol geometry
The problem of drawing and solving-optimizing side right weight, specifically：
Assuming that the ordered list for being pushed to the candidate answers of user isWhereinFor optimum answer, FtopTo seek the function of the element of ranking first in ordered list, then by user's
Negative feedback information tnIt is defined as from ordered list AkTo ordered listMatching tn:Wherein For the optimum answer of user feedback.
Negative feedback information collection T based on usernThe algorithm of solving-optimizing side right weight is represented by as follows：
Input：Negative feedback collection Tn, figure G to be optimized
Output：Figure G is optimized*
Algorithm description：
Wherein function GenerateConstraints (tn) flow it is as follows：
The t of user feedbacknHis optimum answer thought for containing user's selection, so the corresponding candidate of the optimum answer
Answer nodeWith query node vqSimilarity should be more than other candidate answers nodesWith query node vq
Between similarity, accordingly generate constraint equation：
Function GenerateObjective (tn) flow it is as follows：
Object function in signomial geometric programming problem can freely be set, and in embodiments of the present invention, set target
Function is so that the side right weight variable quantity of knowledge mapping is minimum, and formal definitions are：
NormalizeEdges functions are used for that figure is normalized, that is, from some node to its neighbor node
All directed edges weight summation be 1.
Fig. 3 is an exemplary schematic diagram using the above-mentioned negative feedback Advance data quality side right weight based on single user,
The candidate answers ordered list that user is pushed to according to original knowledge collection of illustrative plates is<d1,d2,d3>, wherein d1For optimum answer, and use
Think d in family2Should be optimum answer, solving optimization side right based on this negative feedback information weighs and update knowledge mapping, when again
When receiving identical Query Information from user, the candidate answers ordered list of user is pushed to according to newer knowledge mapping to be become
<d2,d1,d3>。
In order to further increase the quality of side right re-optimization, can utilize the negative feedback information that is received from multiple users with
And feedback information carrys out solving-optimizing side right weight certainly, wherein feedback information is defined as certainly：
Assuming that the ordered list for being pushed to the candidate answers of user isWhereinFor optimum answer, then by the affirmative feedback information t of userpIt is defined as from ordered list AkTo ordered list
A′kMatching tp:WhereinFor the optimum answer of user feedback.
Feedback information solving-optimizing side right based on multiple users can specifically include again：
1) set for the directed edge being related to based on feedback information is carried out clustering processing to feedback information and obtains multiple clusters
The collection for the directed edge that feedback information is related to is combined into what the query node being related to from feedback information was related to feedback information
The set for the directed edge that all paths of each candidate answers node hop count less than threshold value L are passed through.
When carrying out clustering processing so that the coincidence between the set for the directed edge that each feedback information is related in each cluster
Degree is as high as possible, and the registration between the set for the directed edge that difference cluster internal feedback information is related to is as low as possible.
2) it for each cluster, builds signomial geometric programming problem and solves
For each cluster, the constraint side of signomial geometric programming problem is set based on each feedback information in each cluster
Journey, then the corresponding constraint equation of all feedback informations is formed an equation group as the signomial geometric programming problem of the cluster
Constraint function；The object function in signomial geometric programming problem is freely set, constraint function is finally based on and object function solves
The corresponding optimization side right weight of the cluster.
3) it merges the corresponding optimization side right of each cluster according to fusion rule and restores final optimization side right weight
Specifically fusion criterion can be, for example,：For a directed edge,
If an optimization side right weight of a corresponding cluster is only existed, as the optimization side right of this directed edge
Weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and to be all higher than this again oriented for multiple optimization side rights
The maximum optimization side right recast of side right weight incrementss for making this directed edge is then this directed edge by the original side right weight on side
Optimization side right weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and respectively less than this is oriented again for multiple optimization side rights
The maximum optimization side right recast of side right weight decrement for making this directed edge is then this directed edge by the original side right weight on side
Optimization side right weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and only side right is great has in this for part optimization
To the original side right weight on side, then the weighted average and weight that calculate the side right weight variable quantity of this directed edge are each cluster
Including feedback information number, if weighted average be just, then will make directed edge side right weight incrementss it is maximum excellent
Change the optimization side right weight that side right recast is this directed edge, if weighted average is negative, then the side right weight that will make directed edge
The maximum optimization side right recast of decrement is the optimization side right weight of this directed edge.
In the present invention, the side right re-optimization problem of knowledge mapping is converted to a symbol using the feedback information of user
Geometric programming problem efficiently can automatically improve the accuracy of side right weight and the quality of knowledge mapping.
Embodiments of the present invention also provide a kind of non-volatile memory medium, are stored with knowledge mapping on a storage medium
The optimization program of middle side right weight, the optimization program of side right weight is computer-executed to implement side right in knowledge mapping in knowledge mapping
The optimization method of weight, the program include：
Knowledge mapping definition instruction, defines knowledge mapping, and knowledge mapping includes directed edge between node, node and has
To the original side right weight on side；
Query Information receives instruction, and Query Information is received from user；
Candidate answers push instruction, searches for the candidate answers of corresponding Query Information, using knowledge mapping to candidate answers into
Row sequence, and the candidate answers after sequence are pushed to user；
Feedback information receives instruction, and the feedback information to the candidate answers after sequence is received from user；
The structure instruction of signomial geometric programming problem, builds signomial geometric programming problem, the constraint of signomial geometric programming problem
Function is set based on feedback information, and the object function of signomial geometric programming problem is the function for optimizing side right weight；
Signomial geometric programming problem solving instructs, and solves signomial geometric programming problem and obtains optimization side right weight.
Embodiments of the present invention also provide a kind of optimization equipment of side right weight in knowledge mapping, including：
Memory is stored with the optimization program of side right weight in the knowledge mapping that computer can execute；And
Processor is connected to memory, and be configured as execute knowledge mapping in side right weight optimization program with：
Knowledge mapping is defined, knowledge mapping includes the original side right weight of the directed edge and directed edge between node, node；
Query Information is received from user；
The candidate answers of the corresponding Query Information of search, are ranked up candidate answers using knowledge mapping, and will be after sequence
Candidate answers be pushed to user；
The feedback information to the candidate answers after sequence is received from user；
Signomial geometric programming problem is built, the constraint function of signomial geometric programming problem is set based on feedback information, symbol
The object function of geometric programming problem is the function for optimizing side right weight；
It solves signomial geometric programming problem and obtains optimization side right weight.
Embodiments of the present invention also provide a kind of optimization device of side right weight in system, including above-mentioned knowledge mapping.
It should be noted that in the claim and specification of this patent, such as first and second or the like relationship
Term is only used to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying
There are any actual relationship or orders between these entities or operation.Moreover, the terms "include", "comprise" or its
Any other variant is intended to non-exclusive inclusion so that including the processes of a series of elements, method, article or
Equipment includes not only those elements, but also includes other elements that are not explicitly listed, or further include for this process,
Method, article or the intrinsic element of equipment.In the absence of more restrictions, being wanted by what sentence " including one " limited
Element, it is not excluded that there is also other identical elements in the process, method, article or apparatus that includes the element.
Although by referring to some of the preferred embodiment of the invention, the present invention is shown and described,
It will be understood by those skilled in the art that can to it, various changes can be made in the form and details, without departing from this hair
Bright spirit and scope.
Claims (12)
1. the optimization method of side right weight in a kind of knowledge mapping, which is characterized in that the method includes：
Knowledge mapping is defined, the knowledge mapping includes the original of the directed edge and the directed edge between node, the node
Initial line weight；
Query Information is received from user；
The candidate answers of the corresponding Query Information of search, are ranked up the candidate answers using the knowledge mapping, and
Candidate answers after the sequence are pushed to the user；
The feedback information to the candidate answers after the sequence is received from the user；
Signomial geometric programming problem is built, the constraint function of the signomial geometric programming problem is set based on the feedback information,
The object function of the signomial geometric programming problem is the function for optimizing side right weight；
It solves the signomial geometric programming problem and obtains the optimization side right weight.
2. according to the method described in claim 1, it is characterized in that, being arranged the candidate answers using the knowledge mapping
Sequence includes：
The Query Information is mapped on the knowledge mapping and obtains query node；
The candidate answers are mapped on the knowledge mapping and obtain candidate answers node；
Calculate the similarity between the query node and the candidate answers node；
The candidate answers are arranged according to the descending of the similarity.
3. according to the method described in claim 2, it is characterized in that, between the query node and the candidate answers node
The similarity passes through following calculating：
Wherein, φ (vq, va) indicate the query node vqWith the candidate answers node vaBetween the similarity, z：It indicates from the query node vqTo the candidate answers node vaHop count is less than the path of threshold value, | z | indicate path
The length of z, P [z] are the probability of path z, and c is constant parameter, w (vi, vi+1) indicate v in the knowledge mappingiNode is directed toward
vi+1The side right weight of node, w (vq, v1) indicate the query node vqThe v being directed toward in the knowledge mapping1The side right weight of node, w
(vk, va) indicate v in the knowledge mappingkNode is directed toward the candidate answers node vaSide right weight.
4. according to the method described in claim 2, it is characterized in that, the feedback information includes the user after the sequence
Candidate answers in the optimum answer that selects；
The constraint function is described similar between the corresponding optimum answer node of the optimum answer and the query node
Degree is more than the similarity between other candidate answers nodes and the query node；
The object function be so that the quadratic sum of the side right weight variable quantity of each directed edge is minimum, every directed edge
Side right weight variable quantity is the difference of the optimization side right weight and the original side right weight of this directed edge.
5. according to the method described in claim 2, it is characterized in that, the method further includes：
Clustering processing is carried out to the feedback information and obtains multiple clusters；
The corresponding optimization side right of each cluster, which is merged, according to fusion rule restores the final optimization side right weight,
Each the corresponding optimization side right of the cluster is again by solving based on the feedback information structure that cluster includes described in this
The signomial geometric programming problem and obtain.
6. according to the method described in claim 5, it is characterized in that, the collection for the directed edge being related to based on the feedback information
It closes and clustering processing is carried out to the feedback information, the collection for the directed edge that the feedback information is related to is combined into from the feedback letter
It ceases each candidate answers node hop count that the query node being related to is related to the feedback information and is less than all of threshold value
The set for the directed edge that path is passed through.
7. according to the method described in claim 5, it is characterized in that, the fusion rule is, for a directed edge,
If an optimization side right weight of a corresponding cluster is only existed, as this directed edge
The optimization side right weight；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and the multiple optimization side right is all higher than again
The original side right weight of this directed edge will then make the side right weight incrementss of this directed edge maximum described
Optimize the optimization side right weight that side right recast is this directed edge；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and the multiple optimization side right is respectively less than again
The original side right weight of this directed edge will then make the side right weight decrement of this directed edge maximum described
Optimize the optimization side right weight that side right recast is this directed edge；
If there is multiple optimization side right weights of the multiple clusters of correspondence, and the only part optimization side right is great
In the original side right weight of this directed edge, then the weighted average of the side right weight variable quantity of this directed edge is calculated
Value and weight are each number for clustering the feedback information for including, if the weighted average is just, then will
So that the maximum optimization side right recast of the side right weight incrementss of this directed edge is the described of this directed edge
Optimize side right weight, if the weighted average is negative, then by making the side right weight decrement of this directed edge maximum
The optimization side right recast is the optimization side right weight of this directed edge.
8. method according to any one of claims 1 to 7, which is characterized in that the computational methods for utilizing text similarity, from
The candidate answers of the corresponding Query Information are searched in corpus.
9. the optimization device of side right weight in a kind of knowledge mapping, which is characterized in that described device includes：
Knowledge mapping definition module, defines knowledge mapping, the knowledge mapping include the directed edge between node, the node with
And the original side right weight of the directed edge；
Query Information receiving module receives Query Information from user；
Candidate answers pushing module, the candidate answers of the corresponding Query Information of search, using the knowledge mapping to the time
It selects answer to be ranked up, and the candidate answers after the sequence is pushed to the user；
Feedback information receiving module receives the feedback information to the candidate answers after the sequence from the user；
Signomial geometric programming problem builds module, builds signomial geometric programming problem, the constraint of the signomial geometric programming problem
Function is set based on the feedback information, and the object function of the signomial geometric programming problem is the function for optimizing side right weight；
Signomial geometric programming problem solver module solves the signomial geometric programming problem and obtains the optimization side right weight.
10. a kind of non-volatile memory medium, which is characterized in that be stored with side right weight in knowledge mapping on said storage
Optimization program, the optimization program of side right weight, which is computer-executed, in the knowledge mapping is weighed with implementing side right in knowledge mapping
Optimization method, described program include：
Knowledge mapping definition instruction, defines knowledge mapping, the knowledge mapping include the directed edge between node, the node with
And the original side right weight of the directed edge；
Query Information receives instruction, and Query Information is received from user；
Candidate answers push instruction, the candidate answers of the corresponding Query Information of search, using the knowledge mapping to the time
It selects answer to be ranked up, and the candidate answers after the sequence is pushed to the user；
Feedback information receives instruction, and the feedback information to the candidate answers after the sequence is received from the user；
The structure instruction of signomial geometric programming problem, builds signomial geometric programming problem, the constraint of the signomial geometric programming problem
Function is set based on the feedback information, and the object function of the signomial geometric programming problem is the function for optimizing side right weight；
Signomial geometric programming problem solving instructs, and solves the signomial geometric programming problem and obtains the optimization side right weight.
11. the optimization equipment of side right weight in a kind of knowledge mapping, which is characterized in that including：
Memory is stored with the optimization program of side right weight in the knowledge mapping that computer can execute；And
Processor is connected to the memory, and be configured as executing the optimization program of the weight of side right in the knowledge mapping with：
Knowledge mapping is defined, the knowledge mapping includes the original of the directed edge and the directed edge between node, the node
Initial line weight；
Query Information is received from user；
The candidate answers of the corresponding Query Information of search, are ranked up the candidate answers using the knowledge mapping, and
Candidate answers after the sequence are pushed to the user；
The feedback information to the candidate answers after the sequence is received from the user；
Signomial geometric programming problem is built, the constraint function of the signomial geometric programming problem is set based on the feedback information,
The object function of the signomial geometric programming problem is the function for optimizing side right weight；
It solves the signomial geometric programming problem and obtains the optimization side right weight.
12. a kind of system, which is characterized in that include the optimization device of side right weight in the knowledge mapping described in claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810560627.0A CN108776684B (en) | 2018-05-25 | 2018-05-25 | Optimization method, device, medium, equipment and system for edge weight in knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810560627.0A CN108776684B (en) | 2018-05-25 | 2018-05-25 | Optimization method, device, medium, equipment and system for edge weight in knowledge graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108776684A true CN108776684A (en) | 2018-11-09 |
| CN108776684B CN108776684B (en) | 2021-01-01 |
Family
ID=64024469
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810560627.0A Active CN108776684B (en) | 2018-05-25 | 2018-05-25 | Optimization method, device, medium, equipment and system for edge weight in knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108776684B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109582958A (en) * | 2018-11-20 | 2019-04-05 | 厦门大学深圳研究院 | A kind of disaster story line construction method and device |
| CN109670051A (en) * | 2018-12-14 | 2019-04-23 | 北京百度网讯科技有限公司 | Knowledge mapping method for digging, device, equipment and storage medium |
| CN109712704A (en) * | 2018-12-14 | 2019-05-03 | 北京百度网讯科技有限公司 | The recommended method and its device of scheme |
| CN110390003A (en) * | 2019-06-19 | 2019-10-29 | 北京百度网讯科技有限公司 | Question and answer processing method and system, computer equipment and readable medium based on medical treatment |
| CN111143573A (en) * | 2019-12-04 | 2020-05-12 | 华东师范大学 | Method for predicting target node of knowledge graph based on user feedback information |
| CN111177474A (en) * | 2019-06-27 | 2020-05-19 | 腾讯科技（深圳）有限公司 | Graph data processing method and related device |
| CN111191042A (en) * | 2019-12-10 | 2020-05-22 | 同济大学 | Knowledge graph path semantic relation-based search accuracy evaluation method |
| CN111523000A (en) * | 2020-04-23 | 2020-08-11 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for importing data |
| CN111736738A (en) * | 2020-06-30 | 2020-10-02 | 广州小鹏车联网科技有限公司 | Control object query method and device of vehicle-mounted system |
| CN112948567A (en) * | 2021-05-13 | 2021-06-11 | 华自科技股份有限公司 | Method and system for collecting equipment fault information |
| CN113095646A (en) * | 2021-03-31 | 2021-07-09 | 天津大学 | Shale gas exploitation step water cyclic utilization intelligent system |
| CN113468311A (en) * | 2021-07-20 | 2021-10-01 | 四川启睿克科技有限公司 | Knowledge graph-based complex question and answer method, device and storage medium |
| CN113918689A (en) * | 2021-09-17 | 2022-01-11 | 秒针信息技术有限公司 | Optimization method and device of knowledge graph question-answering system |
| CN116415564A (en) * | 2023-06-12 | 2023-07-11 | 中国电子科技集团公司信息科学研究院 | Functional point amplification method and system based on knowledge graph |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI456412B (en) * | 2011-10-11 | 2014-10-11 | Univ Ming Chuan | Method for generating a knowledge map |
| CN104965869A (en) * | 2015-06-09 | 2015-10-07 | 浙江大学 | Mobile application sorting and clustering method based on heterogeneous information network |
| CN107038508A (en) * | 2017-06-06 | 2017-08-11 | 海南大学 | The study point tissue and execution route of the learning ability modeling of knowledge based collection of illustrative plates and the target drives of dynamic self-adapting recommend method |
| CN107122399A (en) * | 2017-03-16 | 2017-09-01 | 中国科学院自动化研究所 | Combined recommendation system based on Public Culture knowledge mapping platform |
| CN107368468A (en) * | 2017-06-06 | 2017-11-21 | 广东广业开元科技有限公司 | A kind of generation method and system of O＆M knowledge mapping |
-
2018
- 2018-05-25 CN CN201810560627.0A patent/CN108776684B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI456412B (en) * | 2011-10-11 | 2014-10-11 | Univ Ming Chuan | Method for generating a knowledge map |
| CN104965869A (en) * | 2015-06-09 | 2015-10-07 | 浙江大学 | Mobile application sorting and clustering method based on heterogeneous information network |
| CN107122399A (en) * | 2017-03-16 | 2017-09-01 | 中国科学院自动化研究所 | Combined recommendation system based on Public Culture knowledge mapping platform |
| CN107038508A (en) * | 2017-06-06 | 2017-08-11 | 海南大学 | The study point tissue and execution route of the learning ability modeling of knowledge based collection of illustrative plates and the target drives of dynamic self-adapting recommend method |
| CN107368468A (en) * | 2017-06-06 | 2017-11-21 | 广东广业开元科技有限公司 | A kind of generation method and system of O＆M knowledge mapping |
Non-Patent Citations (1)
| Title |
|---|
| 杜泽宇 等: "基于中文知识图谱的电商领域问答系统", 《计算机应用与软件》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109582958B (en) * | 2018-11-20 | 2023-07-18 | 厦门大学深圳研究院 | Disaster story line construction method and device |
| CN109582958A (en) * | 2018-11-20 | 2019-04-05 | 厦门大学深圳研究院 | A kind of disaster story line construction method and device |
| CN109670051A (en) * | 2018-12-14 | 2019-04-23 | 北京百度网讯科技有限公司 | Knowledge mapping method for digging, device, equipment and storage medium |
| CN109712704A (en) * | 2018-12-14 | 2019-05-03 | 北京百度网讯科技有限公司 | The recommended method and its device of scheme |
| CN110390003A (en) * | 2019-06-19 | 2019-10-29 | 北京百度网讯科技有限公司 | Question and answer processing method and system, computer equipment and readable medium based on medical treatment |
| CN111177474A (en) * | 2019-06-27 | 2020-05-19 | 腾讯科技（深圳）有限公司 | Graph data processing method and related device |
| CN111177474B (en) * | 2019-06-27 | 2022-12-02 | 腾讯科技（深圳）有限公司 | Graph data processing method and related device |
| CN111143573A (en) * | 2019-12-04 | 2020-05-12 | 华东师范大学 | Method for predicting target node of knowledge graph based on user feedback information |
| CN111143573B (en) * | 2019-12-04 | 2023-04-28 | 华东师范大学 | Method for predicting knowledge-graph target node based on user feedback information |
| CN111191042A (en) * | 2019-12-10 | 2020-05-22 | 同济大学 | Knowledge graph path semantic relation-based search accuracy evaluation method |
| CN111523000A (en) * | 2020-04-23 | 2020-08-11 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for importing data |
| CN111523000B (en) * | 2020-04-23 | 2023-06-23 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for importing data |
| CN111736738A (en) * | 2020-06-30 | 2020-10-02 | 广州小鹏车联网科技有限公司 | Control object query method and device of vehicle-mounted system |
| WO2022001682A1 (en) * | 2020-06-30 | 2022-01-06 | 广州橙行智动汽车科技有限公司 | Control object query method and apparatus for vehicle-mounted system |
| CN113095646A (en) * | 2021-03-31 | 2021-07-09 | 天津大学 | Shale gas exploitation step water cyclic utilization intelligent system |
| CN113095646B (en) * | 2021-03-31 | 2022-08-19 | 天津大学 | Shale gas exploitation step water cyclic utilization intelligent system |
| CN112948567B (en) * | 2021-05-13 | 2021-09-03 | 华自科技股份有限公司 | Method and system for collecting equipment fault information |
| CN112948567A (en) * | 2021-05-13 | 2021-06-11 | 华自科技股份有限公司 | Method and system for collecting equipment fault information |
| CN113468311A (en) * | 2021-07-20 | 2021-10-01 | 四川启睿克科技有限公司 | Knowledge graph-based complex question and answer method, device and storage medium |
| CN113468311B (en) * | 2021-07-20 | 2023-09-19 | 四川启睿克科技有限公司 | Knowledge graph-based complex question and answer method, device and storage medium |
| CN113918689A (en) * | 2021-09-17 | 2022-01-11 | 秒针信息技术有限公司 | Optimization method and device of knowledge graph question-answering system |
| CN116415564A (en) * | 2023-06-12 | 2023-07-11 | 中国电子科技集团公司信息科学研究院 | Functional point amplification method and system based on knowledge graph |
| CN116415564B (en) * | 2023-06-12 | 2023-08-25 | 中国电子科技集团公司信息科学研究院 | Functional point amplification method and system based on knowledge graph |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108776684B (en) | 2021-01-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108776684A (en) | Optimization method, device, medium, equipment and the system of side right weight in knowledge mapping | |
| CN108804641B (en) | Text similarity calculation method, device, equipment and storage medium | |
| US10706103B2 (en) | System and method for hierarchical distributed processing of large bipartite graphs | |
| US10157218B2 (en) | Author disambiguation and publication assignment | |
| US10133807B2 (en) | Author disambiguation and publication assignment | |
| Agarwal et al. | Statistical methods for recommender systems | |
| WO2013138516A1 (en) | Publishing product information | |
| WO2014148948A1 (en) | Method for selecting valid variants in search and recommendation systems (variants) | |
| US11574240B2 (en) | Categorization for a global taxonomy | |
| Kamal et al. | Evolutionary framework for coding area selection from cancer data | |
| Hung et al. | Answer validation for generic crowdsourcing tasks with minimal efforts | |
| US9740748B2 (en) | Similarity and ranking of databases based on database metadata | |
| Jaspers et al. | Machine learning techniques for the automation of literature reviews and systematic reviews in EFSA | |
| US20190066054A1 (en) | Accuracy of member profile retrieval using a universal concept graph | |
| WO2020040253A1 (en) | Prediction interpretation device and prediction interpretation method | |
| CN113343101B (en) | Object ordering method and system | |
| JP2017151926A (en) | Information processing program, information processing device, and information processing method | |
| US7734633B2 (en) | Listwise ranking | |
| Chen et al. | Improved interpretability of machine learning model using unsupervised clustering: predicting time to first treatment in chronic lymphocytic leukemia | |
| US20150095183A1 (en) | Methods and Apparatus for Generating Recommendations | |
| Gao et al. | A fuzzy logistic regression model based on the least squares estimation | |
| Park et al. | Using entropy as a measure of acceptance for multi-label classification | |
| Singh et al. | Optimized doctor recommendation system using supervised machine learning | |
| CN113312523B (en) | Dictionary generation and search keyword recommendation method and device and server | |
| US9928291B2 (en) | Author disambiguation and publication assignment |
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 |