CN110442618A - Merge convolutional neural networks evaluation expert's recommended method of expert info incidence relation - Google Patents
Merge convolutional neural networks evaluation expert's recommended method of expert info incidence relation Download PDFInfo
- Publication number
- CN110442618A CN110442618A CN201910677191.8A CN201910677191A CN110442618A CN 110442618 A CN110442618 A CN 110442618A CN 201910677191 A CN201910677191 A CN 201910677191A CN 110442618 A CN110442618 A CN 110442618A
- Authority
- CN
- China
- Prior art keywords
- expert
- info
- incidence relation
- vector
- neural networks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 27
- 238000011156 evaluation Methods 0.000 title claims abstract description 22
- 239000011159 matrix material Substances 0.000 claims abstract description 36
- 230000004927 fusion Effects 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 3
- 230000006870 function Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 4
- 238000003780 insertion Methods 0.000 claims description 4
- 230000037431 insertion Effects 0.000 claims description 4
- 230000008447 perception Effects 0.000 claims description 3
- 238000007476 Maximum Likelihood Methods 0.000 claims description 2
- 238000000547 structure data Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 7
- 238000005516 engineering process Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 238000013507 mapping Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 5
- 238000001914 filtration Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 210000004209 hair Anatomy 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Fuzzy Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to convolutional neural networks evaluation expert's recommended methods of fusion expert info incidence relation, belong to technical field of data processing.The present invention becomes expert correlation matrix by constructing the incidence relation matrix between several expert infos, and splicing with the expert info vector for having been converted into vector, and evaluation expert's recommended models of fusion expert's incidence relation are constructed finally by convolutional neural networks.By the score relationship between project and expert, carry out expert's recommended models to train.The experimental results showed that expert's recommended models proposed by the present invention achieve preferable effect in actual task, it has a certain upgrade in recommendation effect compared to the method for not considering expert's relationship.
Description
Technical field
The present invention relates to convolutional neural networks evaluation expert's recommended methods of fusion expert info incidence relation, belong to number
According to processing technology field.
Background technique
In recent years, China's science and technology item such as declares, sets up the project at movable increasing, China's science and technology item at different levels and the science and technology
Talent Management also realizes information system successively, and expert plays always crucial effect in evaluation.In order to protect
Objectivity, fairness and the fairness of project appraisal work are demonstrate,proved, the work of selecting of evaluation expert is particularly important.
Expert recommend (Expert Recommendation) be based on information of expert database abundant, using data mining,
The technologies such as machine learning, which are found, looks for expert with what task to be recommended matched.From task to be recommended, expert recommends to see
Work is one of personalized recommendation of Cempetency-based education special shape.In e-commerce field, proposed algorithm and technology
Research is relatively more, this recommends to provide sufficient theoretical basis for expert.
Currently, mainly proposing following several method in recommended method research, 1. be the side based on content respectively
Method, 2. method and 3. methods based on deep learning based on collaborative filtering.In the method based on content, need to pass through digging
The content characteristic of pick project and user calculates similarity using machine learning method and is recommended.Content-based recommendation method
Dependent on the characteristic information about user preference and project, the record that largely scores is not needed, therefore there is no score data is dilute
Thin problem.Simultaneously for new projects, it is only necessary to which carrying out feature extraction can recommend to user, solve new projects
Cold start-up problem, but can usually suffer from the problem of feature extraction difficulty.Method main thought based on collaborative filtering is benefit
With the method between similar users with similar interests preference, potential preference of the Lai Faxian user to project.Collaborative filtering is only
The history score data using user is needed, therefore simple and effective.But since user is to the score data relative term of project
The problem of purpose total quantity is considerably less, frequently suffers from Sparse, further for new user or project, due to not scoring
Data and can not be recommended, there are problems that cold start-up.Although deep learning does not need the engineer as conventional method special
Sign, but how the incidence relation between expert to be dissolved into deep learning model is second critical issue.
To solve the above-mentioned problems, the invention proposes the convolutional neural networks evaluations of research fusion expert's incidence relation
Expert recommendation method.
Summary of the invention
The present invention provides convolutional neural networks evaluation expert's recommended method of fusion expert info incidence relation, this hairs
The model of bright middle building is indicated by the vector of the expertise attribute information in Project evaluation information and existing database, is utilized
Convolutional neural networks, which are realized, calculates the matching degree score between expert and project.Particularly, it has incorporated in a model herein more
Information association relationship between name expert, the recommendation performance to lift scheme.
The technical scheme is that the convolutional neural networks evaluation expert recommendation side of fusion expert info incidence relation
Method, specific step is as follows for method:
Step1, the information for collecting expert, project information, and the information of the two is separately converted to specially by embeding
Family's hidden hidden vector of vector sum project information of information;
As a preferred solution of the present invention, it in the step Step1, is encoded by one-hot by expert and project structure
Change data information and be expressed as the vector expression that computer can recognize and handle, is distinguished by one layer of Embedding layers of insertion
It is mapped to dense hidden vector.
Step2, the thought using markov network, be expressed as the relationship between expert info for expert info it
Between incidence relation matrix;
As a preferred solution of the present invention, in the step Step2, the nonoriented edge of markov network is used to construct specially
Correlation between family's information;The incidence relation between expert is determined first with the part attribute information of expert, it is then sharp
The correlation between expert info is calculated with a variety of different incidence relations between expert info, to generate expert info
Incidence relation matrix.
As a preferred solution of the present invention, the calculation formula of the correlation between the expert info is as follows:
Wherein, if there are this relationship characteristic, h between expert infom(ei,ej)=1, otherwise hm(ei,ej)=0;
Wherein m is expressed as m-th of characteristic function, and M indicates incidence relation feature sum, in the present invention M=4, λmIndicate character pair
The weight of function,Indicate all and node eiThe set on connected side, characteristic function weight λm, using maximal possibility estimation
Method estimates it;As Sim (ei,ej) value when being greater than given threshold value beta, then it is assumed that expert info eiAnd expert info
ejRelevant, namely in expert's Markov network between the two nodes there are a nonoriented edge, the weight on side is Sim
(ei,ej)。
Step3, the expert info incidence relation matrix in the hidden vector of expert info obtained in Step1 and Step2 is closed
And group is combined into expert info correlation matrix;
As a preferred solution of the present invention, in the step Step3, in order to make full use of expert attribute information and specially
Family between incidence relation solves the problems, such as project appraisal expert recommend, consider by expert info incidence relation matrix with
Expert info Vector Fusion is prepared for convolution extraction expert's linked character of next step.
Step4, the expert info correlation matrix in Step3 is subjected to convolution sum Chi Huacao by convolutional neural networks
Make, extract the hidden vector of relationship characteristic in several experts, the hidden vector of output is closed with the hidden vector of project information in Step1 again
And and the matrix-vector after merging is input to MLP layers of multi-layer perception (MLP), study obtains expert to the matching degree of Project evaluation
Score, to train model.
As a preferred solution of the present invention, in the step Step4, for attribute information and the expert for making full use of expert
Between incidence relation solve the problems, such as that the multi-expert for project appraisal is recommended, using convolutional neural networks from having merged expert
Feature is extracted in the expert correlation matrix of information.The groundwork of the layer of MLP is to operate obtained expert to convolution sum pondization
The hidden vector of information is carried out with the vector after only being merged by the hidden vector of project information that insertion (Embedding) layer obtains
Softmax operation, thus marking of the implementation model to expert is recommended.
The beneficial effects of the present invention are:
1, the present invention has merged expert info incidence relation matrix in a network,
2, using CNN convolutional neural networks extract expert info relational matrix feature, facilitate in this way excavate expert it
Between existing hiding relationship, thus improve expert recommendation accuracy.
3, the experimental results showed that expert's recommended models proposed by the present invention achieve preferable effect in actual task,
It has a certain upgrade in recommendation effect compared to the method for not considering expert's relationship.
Detailed description of the invention
Fig. 1 is total flow chart in the present invention;
Fig. 2 is evaluation expert's data correlation relation figure based on Markov network in invention;
Fig. 3 is the convolutional neural networks flow chart in invention.
Specific embodiment
Embodiment 1: as shown in Figure 1-3, the convolutional neural networks evaluation expert recommendation side of fusion expert info incidence relation
Method, specific step is as follows for method:
Expert architecture data information in Step1, extraction Science and Technology Department, Yunnan Province talent's expert management system, needs to evaluate
Project structure data information, the sparse vector of binaryzation is converted by one-hot coding, passes through one layer of insertion
Embedding layers are mapped to it dense hidden vector respectively.
Step2, the thought using markov network, be expressed as the relationship between expert info for expert info it
Between incidence relation matrix;
Before constructing expert info incidence relation matrix, need to construct expert using Markov (markov) network
Between relational graph, recycle relational graph go generate expert's incidence relation matrix.From fig. 2 it can be seen that in expert Ma Erke
Expert info is by node e in husband's networkjIt indicates, the nonoriented edge that the correlation between expert info is constituted, the weight on side depends on
Correlation between expert info.In expert's Markov network, needs to merge expert's incidence relation and calculate between expert
Correlation.In order to merge the expert info incidence relation of these types, the definition of these expert info incidence relations is characterized letter
Number hm(ei,ej), and weight λ is distributed for each characteristic functionm, these linked characters are merged using log-linear model.Assuming that
Correlation Sim (e between expert infoi,ej) indicate, correlation calculations formula are as follows:Wherein, if there are this relationship characteristic, h between expert infom
(ei,ej)=1, otherwise hm(ei,ej)=0.Wherein m is expressed as m-th of characteristic function, and M indicates incidence relation feature sum, In
M=4 herein, λmIndicate the weight of character pair function,Indicate all and node eiThe set on connected side.Characteristic function power
Weight λm, it is estimated using maximum Likelihood.As Sim (ei,ej) value when being greater than given threshold value beta, then recognize
For expert info eiWith expert info ejIt is relevant, namely there are one between the two nodes in expert's Markov network
Nonoriented edge, the weight on side are Sim (ei,ej).Expert's relational matrix is constructed according to above method.
Table 1 is expert info incidence relation feature
Step3, the expert info incidence relation matrix in the hidden vector of expert info obtained in Step1 and Step2 is closed
And group is combined into expert info correlation matrix;
As a preferred solution of the present invention, in the step Step3, in order to make full use of expert attribute information and specially
Family between incidence relation solves the problems, such as project appraisal expert recommend, consider by expert info incidence relation matrix with
Expert info Vector Fusion is prepared for convolution extraction expert's linked character of next step.
Specifically, n expert's vector corresponding to project then first passes through one layer of Embedding layers of mapping and then vertical
The matrix of n × k size is connected into, wherein k is the dimension of the hidden vector of expert.Afterwards by the matrix of obtained n × k size with
The expert info incidence matrix horizontal connection of this n expert's n × n size is combined into new expert's confederate matrix.
Step4, the expert info correlation matrix in Step3 is subjected to convolution sum Chi Huacao by convolutional neural networks
Make, extract the hidden vector of relationship characteristic in several experts, the hidden vector of output is closed with the hidden vector of project information in Step1 again
And and the matrix-vector after merging is input to MLP layers of multi-layer perception (MLP), study obtains expert to the matching degree of Project evaluation
Score, to train model.
Building process is specific as follows:
First layer: input layer: from obtained in the project information vector sum step Step2 obtained in the step Step1
Expert's confederate matrix of expert's related information is merged as input.
The second layer: convolutional layer, the groundwork of this layer are that spy is extracted from expert's confederate matrix using convolutional neural networks
Sign.Enable xi∈RnIndicate the n-dimensional vector of i-th of expert in expert's confederate matrix, then the matrix comprising m expert may be expressed as:Wherein,Indicate attended operation, such as Convolution
Operation refers to using the sliding window that length is h from x1:mInput of the vector as convolutional layer activation primitive is intercepted, to activation letter
After number is calculated, obtained output is feature to be extracted.It is shown below: ci=f (Wxi:i+h-1+b).Wherein, letter
Number f indicates activation primitive, selects ReLU function as activation primitive herein.W and b indicates weight matrix and bias unit, for mind
Parameter through network, can be by being trained to obtain to neural network.ciIt indicates according to activation primitive f and input vector xi:i+h-1
Calculated feature.Above-mentioned formula describe length be h sliding window in position xi:i+h-1The operation carried out when upper, when
When sliding window slides into ending from the beginning of sentence, one group of input can be obtained, be expressed as (x1:h,x2:1+h,...,xm-h+1:m), into
And one group of output can be obtained according to the input, it is expressed as (c1,c2,...,cm-h+1), the output of this group is known as Feature Mapping.
Third layer: max-pooling layers, the main purpose of this layer is handled upper one layer of Feature Mapping, is obtained
Final feature vector.It specifically, is the maximum value extracted in Feature Mapping, the thought done so is to think each spy
Maximum value in sign mapping represents most important feature in the mapping.In addition, max-pooling layers of a major advantage
It is that can obtain the feature vector of a regular length by means of the operation of this layer, is originally inputted vector without being concerned about
Length.
4th layer: output layer, the groundwork of this layer are that the expert info that will learn by convolutional neural networks is hidden
Vector and the hidden vector of project information are attached and learn and export scores.The layer is MLP layers, is made using Relu function
For activation primitive, and softmax is carried out to result and handles to obtain probability.If softmax layers of categorical measure, i.e. neuron number
For k, then expression formula isIn, z indicates that softmax layers of input vector, W indicate softmax layer network
Parameter, yiIndicate that the output valve of i-th of neuron of output layer, the result of final output are that k dimensional vector indicates that k type is other
Probability value.The output layer of the model determines the quantity of neuron according to the categorical measure of classification task.N is a specially in order to obtain
Family needs to export n using n softmax in MLP layers of the last layer for the probability value of the score size of same project
The corresponding scores of a expert.
In addition, in order to verify the validity for incorporating expert info incidence relation, while in view of the special of experimental study problem
Property, use the matrix not comprising expert info incidence relation that must input as model and tests as a comparison, which is denoted as M1,
The convolutional neural networks recommended models of fusion expert info incidence relation proposed in this paper are denoted as M2.Respectively using identical
The above two model of training set training, and respectively obtain two kinds of different models using identical test set and pushed away for single project
The average NDCG value of identical five experts is recommended, as shown in Table 2 below, wherein K is every class in expert or the expression of contents of a project attribute
Size of the attribute after Embedding
Table 2 is comparative test result
K | M1 | M2 |
8 | 0.931 | 0.934 |
16 | 0.935 | 0.945 |
32 | 0.942 | 0.948 |
From Table 2, it can be seen that the effect for using expert info incidence relation to obtain as the model of auxiliary input wants excellent
In the model without using expert info incidence relation as auxiliary input.The results show, herein in recommender system sequence
Preferable effect is achieved in NDCG index.
Above in conjunction with attached drawing, the embodiment of the present invention is explained in detail, but the present invention is not limited to upper
Embodiment is stated, within the knowledge of a person skilled in the art, present inventive concept can also not departed from
Under the premise of various changes can be made.
Claims (6)
1. merging convolutional neural networks evaluation expert's recommended method of expert info incidence relation, it is characterised in that: the tool of method
Steps are as follows for body:
Step1, the information for collecting expert, project information, and the information of the two is separately converted to expert's letter by embeding
Cease the hidden hidden vector of vector sum project information;
Step2, the thought using markov network, make the relationship between expert info be expressed as the pass between expert info
Join relational matrix;
Step3, the hidden vector of expert info obtained in Step1 is merged with the expert info incidence relation matrix in Step2, group
It is combined into expert info correlation matrix;
Step4, the expert info correlation matrix in Step3 is subjected to the operation of convolution sum pondization by convolutional neural networks, mentioned
The hidden vector of relationship characteristic in several experts is taken, the hidden vector of output is merged with the hidden vector of the project information in Step1 again, and will
Matrix-vector after merging is input to MLP layers of multi-layer perception (MLP), and study obtains expert to the matching degree score of Project evaluation, thus
Train model.
2. convolutional neural networks evaluation expert's recommended method of fusion expert info incidence relation according to claim 1,
It is characterized by: being encoded by one-hot in the step Step1 and being expressed as counting by expert and project structure data information
Calculation machine can recognize and the vector of processing indicates, is embedded in Embedding layers by one layer and it is mapped to dense hidden vector respectively.
3. convolutional neural networks evaluation expert's recommended method of fusion expert info incidence relation according to claim 1,
It is characterized by: the nonoriented edge of markov network is used to construct the correlation between expert info in the step Step2;
The incidence relation between expert is determined first with the part attribute information of expert, then using a variety of between expert info
Different incidence relations calculates the correlation between expert info, to generate expert info incidence relation matrix.
4. convolutional neural networks evaluation expert's recommended method of fusion expert info incidence relation according to claim 3,
It is characterized by: the calculation formula of the correlation between the expert info is as follows:
Wherein, if there are this relationship characteristic, h between expert infom(ei,ej)=1, otherwise hm(ei,ej)=0;Wherein m
It is expressed as m-th of characteristic function, M indicates incidence relation feature sum, in the present invention M=4, λmIndicate character pair function
Weight,Indicate all and node eiThe set on connected side, characteristic function weight λm, using maximum Likelihood to it
Estimated;As Sim (ei,ej) value when being greater than given threshold value beta, then it is assumed that expert info eiWith expert info ejIt is related
, namely in expert's Markov network between the two nodes there are a nonoriented edge, the weight on side is Sim (ei,ej)。
5. convolutional neural networks evaluation expert's recommended method of fusion expert info incidence relation according to claim 1,
It is characterized by: in the step Step3, in order to make full use of the incidence relation between the attribute information of expert and expert to solve
Recommend problem for the expert of project appraisal, considers by expert info incidence relation matrix and expert info Vector Fusion, under
The convolution of one step extracts expert's linked character and prepares.
6. convolutional neural networks evaluation expert's recommended method of fusion expert info incidence relation according to claim 1,
It is characterized by: in the step Step4, to make full use of the incidence relation between the attribute information of expert and expert to solve needle
Problem is recommended to the multi-expert of project appraisal, using convolutional neural networks from the expert correlation matrix for having merged expert info
Feature is extracted, the groundwork of the layer of MLP is the hidden vector of expert info obtained to the operation of convolution sum pondization and only passes through insertion
Vector after the hidden vector of Embedding layers of obtained project information merges carries out softmax operation, so that implementation model is to recommendation
The marking of expert.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910677191.8A CN110442618B (en) | 2019-07-25 | 2019-07-25 | Convolutional neural network review expert recommendation method fusing expert information association relation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910677191.8A CN110442618B (en) | 2019-07-25 | 2019-07-25 | Convolutional neural network review expert recommendation method fusing expert information association relation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110442618A true CN110442618A (en) | 2019-11-12 |
CN110442618B CN110442618B (en) | 2023-04-18 |
Family
ID=68431499
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910677191.8A Active CN110442618B (en) | 2019-07-25 | 2019-07-25 | Convolutional neural network review expert recommendation method fusing expert information association relation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110442618B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111782797A (en) * | 2020-07-13 | 2020-10-16 | 贵州省科技信息中心 | Automatic matching method for scientific and technological project review experts and storage medium |
CN112100370A (en) * | 2020-08-10 | 2020-12-18 | 淮阴工学院 | Picture examination expert combined recommendation method based on text convolution and similarity algorithm |
CN113343666A (en) * | 2021-06-29 | 2021-09-03 | 深圳前海微众银行股份有限公司 | Method, device and equipment for determining confidence degree of score and storage medium |
CN113516094A (en) * | 2021-07-28 | 2021-10-19 | 中国科学院计算技术研究所 | System and method for matching document with review experts |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103631859A (en) * | 2013-10-24 | 2014-03-12 | 杭州电子科技大学 | Intelligent review expert recommending method for science and technology projects |
CN107330115A (en) * | 2017-07-12 | 2017-11-07 | 广东工业大学 | A kind of information recommendation method and device |
CN108647206A (en) * | 2018-05-04 | 2018-10-12 | 重庆邮电大学 | Chinese spam filtering method based on chaotic particle swarm optimization CNN networks |
CN108804689A (en) * | 2018-06-14 | 2018-11-13 | 合肥工业大学 | The label recommendation method of the fusion hidden connection relation of user towards answer platform |
CN109062961A (en) * | 2018-06-27 | 2018-12-21 | 淮阴工学院 | A kind of expert's combination recommended method of knowledge based map |
CN109241424A (en) * | 2018-08-29 | 2019-01-18 | 陕西师范大学 | A kind of recommended method |
CN109308315A (en) * | 2018-10-19 | 2019-02-05 | 南京理工大学 | A kind of collaborative recommendation method based on specialist field similarity and incidence relation |
CN109657140A (en) * | 2018-12-13 | 2019-04-19 | 平安普惠企业管理有限公司 | Information-pushing method, device, computer equipment and storage medium |
US20190138806A1 (en) * | 2017-11-03 | 2019-05-09 | Tata Consultancy Services Limited | Signal analysis systems and methods for features extraction and interpretation thereof |
CN109902201A (en) * | 2019-03-08 | 2019-06-18 | 天津理工大学 | A kind of recommended method based on CNN and BP neural network |
-
2019
- 2019-07-25 CN CN201910677191.8A patent/CN110442618B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103631859A (en) * | 2013-10-24 | 2014-03-12 | 杭州电子科技大学 | Intelligent review expert recommending method for science and technology projects |
CN107330115A (en) * | 2017-07-12 | 2017-11-07 | 广东工业大学 | A kind of information recommendation method and device |
US20190138806A1 (en) * | 2017-11-03 | 2019-05-09 | Tata Consultancy Services Limited | Signal analysis systems and methods for features extraction and interpretation thereof |
CN108647206A (en) * | 2018-05-04 | 2018-10-12 | 重庆邮电大学 | Chinese spam filtering method based on chaotic particle swarm optimization CNN networks |
CN108804689A (en) * | 2018-06-14 | 2018-11-13 | 合肥工业大学 | The label recommendation method of the fusion hidden connection relation of user towards answer platform |
CN109062961A (en) * | 2018-06-27 | 2018-12-21 | 淮阴工学院 | A kind of expert's combination recommended method of knowledge based map |
CN109241424A (en) * | 2018-08-29 | 2019-01-18 | 陕西师范大学 | A kind of recommended method |
CN109308315A (en) * | 2018-10-19 | 2019-02-05 | 南京理工大学 | A kind of collaborative recommendation method based on specialist field similarity and incidence relation |
CN109657140A (en) * | 2018-12-13 | 2019-04-19 | 平安普惠企业管理有限公司 | Information-pushing method, device, computer equipment and storage medium |
CN109902201A (en) * | 2019-03-08 | 2019-06-18 | 天津理工大学 | A kind of recommended method based on CNN and BP neural network |
Non-Patent Citations (4)
Title |
---|
SHENGXIANG GAO ET AL.: "A Method to Review Expert Recommendation Using Topic Relevance and Expert Relationship", 《INTERNATIONAL JOURNAL OF COOPERATIVE INFORMATION SYSTEMS》 * |
SHENGXIANG GAO ET AL.: "Review Expert Collaborative Recommendation Algorithm Based on Topic Relationship", 《IEEE/CAA JOURNAL OF AUTOMATICA SINICA》 * |
何孝胥: "融合关联关系的卷积神经网络项目评审专家推荐方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
余峰 等: "基于主题信息的项目评审专家推荐方法", 《计算机工程》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111782797A (en) * | 2020-07-13 | 2020-10-16 | 贵州省科技信息中心 | Automatic matching method for scientific and technological project review experts and storage medium |
CN112100370A (en) * | 2020-08-10 | 2020-12-18 | 淮阴工学院 | Picture examination expert combined recommendation method based on text convolution and similarity algorithm |
CN112100370B (en) * | 2020-08-10 | 2023-07-25 | 淮阴工学院 | Picture-trial expert combination recommendation method based on text volume and similarity algorithm |
CN113343666A (en) * | 2021-06-29 | 2021-09-03 | 深圳前海微众银行股份有限公司 | Method, device and equipment for determining confidence degree of score and storage medium |
CN113516094A (en) * | 2021-07-28 | 2021-10-19 | 中国科学院计算技术研究所 | System and method for matching document with review experts |
CN113516094B (en) * | 2021-07-28 | 2024-03-08 | 中国科学院计算技术研究所 | System and method for matching and evaluating expert for document |
Also Published As
Publication number | Publication date |
---|---|
CN110442618B (en) | 2023-04-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110442618A (en) | Merge convolutional neural networks evaluation expert's recommended method of expert info incidence relation | |
CN108492200B (en) | User attribute inference method and device based on convolutional neural network | |
CN111488474B (en) | Fine-grained freehand sketch image retrieval method based on attention enhancement | |
CN111753024B (en) | Multi-source heterogeneous data entity alignment method oriented to public safety field | |
CN113704546B (en) | Video natural language text retrieval method based on space time sequence characteristics | |
CN109299657B (en) | Group behavior identification method and device based on semantic attention retention mechanism | |
CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
CN111444367B (en) | Image title generation method based on global and local attention mechanism | |
CN103778227A (en) | Method for screening useful images from retrieved images | |
CN110599592B (en) | Three-dimensional indoor scene reconstruction method based on text | |
CN109271539A (en) | A kind of image automatic annotation method and device based on deep learning | |
CN112084373B (en) | Graph embedding-based multi-source heterogeneous network user alignment method | |
CN110390017A (en) | Target sentiment analysis method and system based on attention gate convolutional network | |
CN111611367B (en) | Visual question-answering method introducing external knowledge | |
CN112182156B (en) | Aspect-level interpretable deep network scoring prediction recommendation method based on text processing | |
CN107688870A (en) | A kind of the classification factor visual analysis method and device of the deep neural network based on text flow input | |
CN113822340A (en) | Image-text emotion recognition method based on attention mechanism | |
CN112115253A (en) | Depth text ordering method based on multi-view attention mechanism | |
CN111753207A (en) | Collaborative filtering model of neural map based on comments | |
CN112417063A (en) | Heterogeneous relation network-based compatible function item recommendation method | |
CN111506760B (en) | Depth integration measurement image retrieval method based on difficult perception | |
CN110443574A (en) | Entry convolutional neural networks evaluation expert's recommended method | |
Li et al. | Combining local and global features into a Siamese network for sentence similarity | |
CN110569355B (en) | Viewpoint target extraction and target emotion classification combined method and system based on word blocks | |
CN114004220A (en) | Text emotion reason identification method based on CPC-ANN |
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 |