CN116842171A - Article recommendation method, apparatus, computer device and storage medium - Google Patents
Article recommendation method, apparatus, computer device and storage medium Download PDFInfo
- Publication number
- CN116842171A CN116842171A CN202310831925.XA CN202310831925A CN116842171A CN 116842171 A CN116842171 A CN 116842171A CN 202310831925 A CN202310831925 A CN 202310831925A CN 116842171 A CN116842171 A CN 116842171A
- Authority
- CN
- China
- Prior art keywords
- article
- user
- information
- point data
- interest point
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012216 screening Methods 0.000 claims abstract description 20
- 238000005457 optimization Methods 0.000 claims abstract description 15
- 238000011160 research Methods 0.000 claims description 51
- 238000004590 computer program Methods 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 12
- 238000010586 diagram Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 68
- 239000000284 extract Substances 0.000 description 7
- 230000006870 function Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000002596 correlated effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000013475 authorization Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect 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
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
- G06F16/337—Profile generation, learning or modification
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to an article recommendation method, an article recommendation device, computer equipment and a storage medium. The application relates to the technical field of big data and artificial intelligence. The method comprises the following steps: acquiring personal information of a user and content information of a plurality of articles; extracting each interest point data of the content information of each article, identifying each feature point data of the personal information of the user, and establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on each interest point data of each article and each feature point data of the user; and determining the adjacency information of the user and each article in the structural relation graph network through a gradient descent optimization strategy to obtain an article recommendation graph network of the user and each article, calculating the similarity between each article and the user based on the article recommendation graph network, and screening the article with the highest similarity as the recommended article of the user. By adopting the method, the accuracy of recommending the articles required by the user to the user can be improved.
Description
Technical Field
The present application relates to the field of big data and artificial intelligence, and in particular, to an article recommendation method, apparatus, computer device, and storage medium.
Background
With the development of big data retrieval technology, article retrieval such as papers, journals and works becomes more and more convenient, but because various article contents are complex nowadays, a large number of similar thousands or tens of thousands of related articles exist in the same direction in the same field, so how to accurately retrieve articles required by users is the current research focus.
The conventional article retrieval method is to input a large amount of keyword (keywords including information such as author, title, content, abstract and the like) information by a user, so as to limit the retrieval range one by one, and obtain articles required by the user, but since the keywords of many similar articles basically coincide, for inexperienced researchers, the input keywords can still return thousands or millions of related articles, so that the accuracy of recommending articles required by the user to the user is low.
Disclosure of Invention
Based on this, it is necessary to provide an article recommendation method, apparatus, computer device, computer readable storage medium and computer program product in view of the above technical problems.
In a first aspect, the present application provides an article recommendation method. The method comprises the following steps:
Acquiring personal information of a user and content information of a plurality of articles;
extracting each interest point data of content information of each article, identifying each feature point data of personal information of the user, and establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on each interest point data of each article and each feature point data of the user;
and determining the adjacency information of the user and each article in the structural relation graph network through a gradient descent optimization strategy to obtain an article recommendation graph network of the user and each article, calculating the similarity between each article and the user based on the article recommendation graph network, and screening articles with highest similarity as recommended articles of the user.
Optionally, the content information includes article content and keyword information, and extracting each interest point data of the content information of each article includes:
identifying text information of the article content of each article aiming at each article, and inquiring text information with similarity greater than a preset similarity threshold value in the text information of the article based on the keyword information of the article to serve as first interest point data of the article;
And taking the first interest point data of the article and the keyword information of the article as the interest point data of the article.
Optionally, in the case that the personal information of the user includes the publication article information of the user, the identifying each feature point data of the personal information of the user includes:
identifying the published text information of the user, and extracting first characteristic data of the published text information of the user through a text feature extraction network;
and determining research content information of the user based on information of non-article information in personal information of the user, extracting research content data corresponding to the research content information, and taking each piece of first characteristic data and the research content data of the user as characteristic point data of the user.
Optionally, in a case where the personal information of the user does not include article information of the user, the identifying each feature point data of the personal information of the user includes:
acquiring article browsing history information of the user, and identifying content information of each piece of browsing article information corresponding to the article browsing history information;
And extracting each browsing interest point data of the content information of the browsing article information, identifying research content data in the personal information of the user, and taking each browsing interest point data and the research content data of the user as characteristic point data of the user.
Optionally, the establishing a structural relationship graph network between the user and each article through a heterogeneous information network based on the point-of-interest data of each article and the feature point data of the user includes:
for each article, calculating the inner product distance between each interest point data and each feature point data through a heterogeneous information network based on each interest point data of the article and each feature point data of the user, and determining meta-path information of each interest point data and each feature point data based on the inner product distance between each interest point data and each feature point data;
and determining a sub-relationship graph network of the articles and the user based on meta-path information of all the interest point data of the articles and all the characteristic point data of the user, and determining a structural relationship graph network of the user and each article based on the sub-relationship graph network of all the articles and the user.
Optionally, determining, by using a gradient descent optimization policy, proximity information between the user and each article in the structural relationship graph network, to obtain an article recommendation graph network between the user and each article, including:
identifying similarity information between interest point data corresponding to each meta-path information and feature point data corresponding to each meta-path information in the structural relation graph network, and taking the similarity information corresponding to each meta-path information as proximity information corresponding to the meta-path information;
and adding the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain an article recommendation graph network.
Optionally, the calculating the similarity between each article and the user based on the article recommendation graph network, and screening the article with the highest similarity as the recommended article of the user includes:
for each article, calculating sub-similarity of each interest point data of the article and each feature point data of the user based on each interest point data of the article and the proximity information of meta-path information between each feature point data of the user;
Identifying distribution information of content information of each interest point data in the article, and determining weight information of each interest point data based on the distribution information;
and carrying out weighted summation processing on each sub-similarity based on the weight information of each interest point data to obtain the similarity between the article and the user, and screening articles corresponding to the highest similarity from all articles to be used as recommended articles of the user.
In a second aspect, the application further provides an article recommending device. The device comprises:
the acquisition module is used for acquiring personal information of a user and content information of a plurality of articles;
the extraction module is used for extracting the interest point data of the content information of each article, identifying the characteristic point data of the personal information of the user, and establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on the interest point data of each article and the characteristic point data of the user;
and the recommending module is used for determining the proximity information of the user and each article in the structural relation diagram network through a gradient descent optimizing strategy to obtain an article recommending diagram network of the user and each article, calculating the similarity between each article and the user based on the article recommending diagram network, and screening articles with the highest similarity as recommending articles of the user.
Optionally, the content information includes article content and keyword information, and the extracting module is specifically configured to:
identifying text information of the article content of each article aiming at each article, and inquiring text information with similarity greater than a preset similarity threshold value in the text information of the article based on the keyword information of the article to serve as first interest point data of the article;
and taking the first interest point data of the article and the keyword information of the article as the interest point data of the article.
Optionally, in the case that the personal information of the user includes the publication information of the user, the extracting module is specifically configured to:
identifying the published text information of the user, and extracting first characteristic data of the published text information of the user through a text feature extraction network;
and determining research content information of the user based on information of non-article information in personal information of the user, extracting research content data corresponding to the research content information, and taking each piece of first characteristic data and the research content data of the user as characteristic point data of the user.
Optionally, in the case that the personal information of the user does not include article information of the user, the extracting module is specifically configured to:
acquiring article browsing history information of the user, and identifying content information of each piece of browsing article information corresponding to the article browsing history information;
and extracting each browsing interest point data of the content information of the browsing article information, identifying research content data in the personal information of the user, and taking each browsing interest point data and the research content data of the user as characteristic point data of the user.
Optionally, the extracting module is specifically configured to:
for each article, calculating the inner product distance between each interest point data and each feature point data through a heterogeneous information network based on each interest point data of the article and each feature point data of the user, and determining meta-path information of each interest point data and each feature point data based on the inner product distance between each interest point data and each feature point data;
and determining a sub-relationship graph network of the articles and the user based on meta-path information of all the interest point data of the articles and all the characteristic point data of the user, and determining a structural relationship graph network of the user and each article based on the sub-relationship graph network of all the articles and the user.
Optionally, the recommendation module is specifically configured to:
identifying similarity information between interest point data corresponding to each meta-path information and feature point data corresponding to each meta-path information in the structural relation graph network, and taking the similarity information corresponding to each meta-path information as proximity information corresponding to the meta-path information;
and adding the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain an article recommendation graph network.
Optionally, the recommendation module is specifically configured to:
for each article, calculating sub-similarity of each interest point data of the article and each feature point data of the user based on each interest point data of the article and the proximity information of meta-path information between each feature point data of the user;
identifying distribution information of content information of each interest point data in the article, and determining weight information of each interest point data based on the distribution information;
and carrying out weighted summation processing on each sub-similarity based on the weight information of each interest point data to obtain the similarity between the article and the user, and screening articles corresponding to the highest similarity from all articles to be used as recommended articles of the user.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the steps of the method of any of the first aspects when the processor executes the computer program.
In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method of any of the first aspects.
In a fifth aspect, the present application also provides a computer program product. The computer program product comprising a computer program which, when executed by a processor, implements the steps of the method of any of the first aspects.
The article recommending method, the article recommending device, the computer equipment and the storage medium are used for acquiring personal information of a user and content information of a plurality of articles; extracting each interest point data of content information of each article, identifying each feature point data of personal information of the user, and establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on each interest point data of each article and each feature point data of the user; and determining the adjacency information of the user and each article in the structural relation graph network through a gradient descent optimization strategy to obtain an article recommendation graph network of the user and each article, calculating the similarity between each article and the user based on the article recommendation graph network, and screening articles with highest similarity as recommended articles of the user. The method comprises the steps of directly obtaining personal information of a user and content information of articles, constructing each interest point data of the content information of each article and each characteristic point data of the personal information of the user, obtaining a correlation network of each article and the user, optimizing the proximity information of the user and each article paper in a structural relationship graph network through a gradient descent optimization strategy, obtaining an article recommendation graph network between the user and each article, improving the accuracy of the correlation between the user and each article, and finally calculating the similarity between the user and each article based on the article recommendation graph network, so that the recommended article of the user is screened, the situation that the user cannot screen articles required by the user due to insufficient screening experience is avoided, and constructing each interest point data of the content information of each article and each characteristic point data of the personal information of the user, so that the global property of determining the correlation between the user and each article is improved, and the accuracy of recommending articles required by the user to the user is improved.
Drawings
FIG. 1 is a flow chart of a chapter recommendation method in one embodiment;
FIG. 2 is a flow diagram of an example chapter recommendation in one embodiment;
FIG. 3 is a block diagram showing a structure of a seal recommending apparatus in one embodiment;
fig. 4 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
The article recommending method provided by the embodiment of the application can be applied to a terminal, a server and a system comprising the terminal and the server and is realized through interaction of the terminal and the server. Wherein the terminal communicates with the server through a network. The data storage system may store data that the server needs to process. The data storage system may be integrated on a server or may be placed on a cloud or other network server. The server may be implemented as a stand-alone server or as a server cluster formed by a plurality of servers. The terminal may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, etc. The terminal directly acquires personal information of the user and content information of the articles, constructs each interest point data of the content information of each article and each characteristic point data of the personal information of the user, so as to obtain a correlation network of each article and the user, optimizes the proximity information of the user and each article paper in the structural relationship graph network through a gradient descent optimization strategy, obtains an article recommendation graph network between the user and each article, improves the accuracy of the correlation between the user and each article, and finally calculates the similarity between the user and each article based on the article recommendation graph network, so that the recommended article of the user is screened, the situation that the user cannot screen the articles required by the user due to insufficient screening experience is avoided, and constructs each interest point data of the content information of each article and each characteristic point data of the personal information of the user, so that the global feature for determining the correlation between the user and each article is improved, and the accuracy of the articles required by the user is improved.
In one embodiment, as shown in fig. 1, an article recommendation method is provided, and the method is applied to a terminal for illustration, and includes the following steps:
step S101, acquiring personal information of a user and content information of a plurality of articles.
In this embodiment, the terminal responds to the search operation of the user, acquires keyword information input by the user, and screens articles containing all the keyword information from a plurality of history articles in the article database, so as to obtain content information of the plurality of articles. Then, the terminal obtains personal information of the user storage and the terminal under the condition of obtaining the authorization of the user. The personal information comprises information such as publication information of a user, an academy, professional information, industry, working field and the like. The content information of the article comprises article content of the article and keyword information of the article.
Step S102, extracting the data of each interest point of the content information of each article, identifying the data of each feature point of the personal information of the user, and establishing a structural relationship graph network between the user and each article through a heterogeneous information network based on the data of each interest point of each article and the data of each feature point of the user.
In this embodiment, the terminal identifies text data related to the keyword information in the article content of the article through the similarity identification network based on the keyword information of the article, and obtains the data of each interest point of the article. And then the terminal identifies the characteristic point data in the personal information of the user through a word characteristic identification strategy. The interest point data and the characteristic point data are characterized as text information. The terminal constructs a structural relation graph network which is changed into each article based on the data of each interest point of each article and the data of each characteristic point of the user through a heterogeneous information network. The heterogeneous information network is a heterogeneous information network recommendation (HGRec) neural network, the word feature recognition strategy is a word feature recognition algorithm generated based on a pre-trained Doc2vec technology, and the similarity recognition network is a similarity recognition neural network generated by the pre-trained Doc2vec technology. The specific process of building the structural relationship graph network between the user and each article will be described in detail later.
Step S103, establishing the adjacency information of the user and each article in the structural relation graph network through a gradient descent optimization strategy, obtaining an article recommendation graph network of the user and each article, calculating the similarity between each article and the user based on the article recommendation graph network, and screening articles with highest similarity as recommended articles of the user.
In this embodiment, a terminal presets a gradient descent optimization policy, and establishes proximity information between a user and each article in the structural relationship graph network based on the gradient descent optimization policy, so as to obtain an article recommendation graph network between the user and each article. The proximity information is used for representing the strength of the association degree between the characteristic point data of the user and the interest point data of each article, and the higher the proximity is, the higher the association strength is, the lower the proximity is, and the lower the association strength is.
Based on the scheme, personal information of the user and content information of the articles are directly obtained, interest point data of the content information of the articles and feature point data of the personal information of the user are built, so that a correlation network of each article and the user is obtained, then the adjacency information of the user and paper of each article in the structural relation graph network is optimized through a gradient descent optimization strategy, an article recommendation graph network between the user and each article is obtained, accuracy of the correlation between the user and each article is improved, finally, similarity between the user and each article is calculated based on the article recommendation graph network, and therefore recommended articles of the user are screened, the situation that the user cannot screen the articles required by the user due to insufficient screening experience is avoided, interest point data of the content information of each article and feature point data of the personal information of the user are built, and global performance of determining the correlation between the user and each article is improved, and therefore accuracy of the articles required by the user recommendation is improved.
Optionally, the content information includes article content and keyword information, and extracting each interest point data of the content information of each article includes: for each article, identifying the text information of the article content of the article, and based on the keyword information of the article, inquiring the text information with the similarity to the keyword information of the article being greater than a preset similarity threshold value in the text information of the article, wherein the text information is used as first interest point data of the article; and taking the first interest point data of the article and the keyword information of the article as the interest point data of the article.
In this embodiment, the terminal extracts, for each article, text information corresponding to the article content of the article. Then, the terminal traverses the article content of the article based on the keyword information of the article content of the article, and calculates the text similarity between the keyword information of the article and each text information corresponding to the article content. And then, presetting a similarity threshold by the terminal, and screening text information corresponding to the similarity larger than the similarity threshold as first interest point data of the article. Wherein the network for computing similarity is a text similarity recognition neural network generated by a pre-trained Doc2vec technique. Finally, the terminal takes all keyword information of the article and the first interest point data as the interest point data of the article
Based on the scheme, text information similar to the keywords is screened to serve as a first interest point of the article, and then the keyword information of the article is combined to obtain interest point data of the article. The global feature of the article keyword extraction is improved, so that a more comprehensive data base is provided for the subsequent establishment of the structural relationship graph network.
Optionally, in a case where the personal information of the user includes the publication information of the user, identifying each feature point data of the personal information of the user includes: identifying the published text information of the user, and extracting first characteristic data of the published text information of the user through a text feature extraction network; based on the information of the non-article information in the personal information of the user, the research content information of the user is determined, the research content data corresponding to the research content information is extracted, and the first characteristic data and the research content data of the user are used as the characteristic point data of the user.
In this embodiment, when the personal information of the user includes the published article information of the user, the terminal extracts all the published text information in the published article information of the user through the text extraction policy. And then, the terminal extracts keyword information related to the publication article in all the publication text information of the user through a text feature extraction network to obtain first feature data corresponding to the publication text information. And then, the terminal inquires the corresponding research content of the user in a research relation database based on the information of non-article information (namely, information of the affiliated institution, affiliated professional information, affiliated industry, affiliated working field and the like) in the personal information of the user, and obtains the research content information of the user. The research content is the content such as the research direction corresponding to the user and the research angle corresponding to the user. Finally, the terminal extracts research content data corresponding to the research content information, and takes the first characteristic data and the research content data of the user as characteristic point data of the user.
Based on the scheme, the characteristic point data of the user is determined through the published article information of the user and the research content information of the user, so that the global property of identifying the requirement information of the user for the article is improved, and a more comprehensive data basis is provided for the subsequent establishment of the structural relation graph network.
Optionally, in a case where the article information of the user is not included in the personal information of the user, identifying each feature point data of the personal information of the user includes: acquiring article browsing history information of a user, and identifying content information of each piece of browsing article information corresponding to the article browsing history information; and extracting each browsing interest point data of the content information of the browsing article information, identifying research content data in the personal information of the user, and taking each browsing interest point data and the research content data of the user as characteristic point data of the user.
In this embodiment, when personal information of a user does not include article information of the user, the terminal obtains a history search record of the user when the terminal obtains authorization of the user, and obtains browsing information of articles searched by the user in the past. Then, the terminal extracts the content information of the article corresponding to each piece of user browsing information, and then extracts each piece of browsing interest point data corresponding to the content information of each piece of article information in the manner of step S102. And then the terminal inquires the research content corresponding to the user in a research relation database based on the personal information of the user to obtain the research content information of the user. The research content is the content such as the research direction corresponding to the user and the research angle corresponding to the user. And finally, the terminal extracts research content data corresponding to the research content information, and takes the browsing interest point data and the research content data of the user as the characteristic point data of the user.
Based on the scheme, the characteristic point data of the user is determined through the article browsing information of the user and the research content information of the user, so that the global property of identifying the requirement information of the user for the articles is improved, and a more comprehensive data basis is provided for the subsequent establishment of the structural relation graph network.
Optionally, based on the data of each interest point of each article and the data of each feature point of the user, a structural relationship graph network between the user and each article is established through a heterogeneous information network, which comprises: aiming at each article, calculating the inner product distance between each interest point data and each feature point data based on each interest point data of the article and each feature point data of a user through a heterogeneous information network, and determining meta-path information of each interest point data and each feature point data based on the inner product distance between each interest point data and each feature point data; and determining a sub-relationship graph network of the articles and the users based on meta-path information of all the interest point data of the articles and all the characteristic point data of the users, and determining a structural relationship graph network of the users and each article based on the sub-relationship graph network of all the articles and the users.
In this embodiment, the terminal calculates, for each article, an inner product distance between each feature point data of the user and each interest point data of the article through the heterogeneous information network.
Specifically, it is assumed that Mj and Mk are two nodes (i.e., point-of-interest data, and feature point data, mj is feature point data, mk is point-of-interest data), and λj and λk are feature data of the above nodes. Thus, the similarity of nodes Mj and Mk can be represented by calculating the inner product distance between two data points, i.e., sjk=λtjλk. In the above equation, the larger Sjk, the more similar they are. Given another node Mt, it is assumed that Mt is less similar to Mj than Mk. Intuitively, sjk will be greater than Sjt. The distance between nodes Mj and Mk should be smaller than the distance between nodes Mj and Mt. Specifically, sjk > Sjt is modeled using a logical function σ (x) =11+ex. An objective function describing the relationship of nodes in a feature representation space may be defined as follows:
the objective function is designed for homogeneous graphs. To measure similarity between nodes in the heterograms, two meta-path based proximity metrics are included in the definition of the objective function. Specifically, S "jk represents a first-order proximity based on the meta path, and S" jk represents a second-order proximity based on the meta path. Minimizing the sum of negative log-likelihood objective functions, defined as follows:
OBJ M ＝min-lnσ(S′ jk -S′ jt )+γ 1 Reg(M)
OBJ N ＝min-lnσ(S″ jk -S″ jt )+γ 2 Reg(N)
where γ1Reg (M) and γ2Reg (N) are l2 norm regularization terms used to avoid overfitting. γ1 and γ2 are penalty coefficients for two proximity metrics. Reg (M) and Reg (N) are set to/M/2F and/N/2F, respectively.
Then, the terminal determines meta-path information of each point of interest data and each feature point data based on an inner product distance of each point of interest data and each feature point data. The meta-path information is characterized by a feature vector between one point of interest data and one feature point data, and a set of inner product distances between one point of interest data and one feature point data. And then, the terminal determines a sub-relationship graph network of the article and the user based on the meta-path information of all the interest point data of the article and all the characteristic point data of the user. Likewise, through the scheme, the terminal obtains the sub-relationship graph network of all articles and the user. And finally, the terminal determines the structural relationship graph network of the user and each article based on the sub-relationship graph network of all articles and the user.
Based on the scheme, the sub-relationship graph network of each article and the user is determined by calculating the inner product distance between each interest point data and each characteristic point data, so that the accuracy of analyzing the relevance between the user and each article is improved.
Optionally, establishing the proximity information of the user and each article in the structural relationship graph network by using a gradient descent optimization strategy to obtain an article recommendation graph network of the user and each article, including: identifying similarity information between interest point data corresponding to each meta-path information and feature point data corresponding to each meta-path information in the structural relation graph network, and taking the similarity information corresponding to each meta-path information as proximity information corresponding to the meta-path information; and adding the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain the article recommendation graph network.
In this embodiment, the terminal presets updated path information of the meta path, calculates, based on each updated path information, interest point data corresponding to each meta path information and proximity between feature point data corresponding to each meta path information in the structural relationship graph network, and then uses the proximity between the interest point data corresponding to each meta path information and the feature point data corresponding to each meta path information as similarity information between the interest point data corresponding to each meta path information and the feature point data corresponding to each meta path information.
Specifically, the terminal presets update path information of two element paths, respectively, (MNM), (Mj-Nj-Mk), for example, the terminal will be used to update the node Mj (feature point data) based on (Mj-Nj-Mk). Where Mk is the point of interest data positively correlated with the similarity of the feature point data, and Mt is the point of interest data negatively correlated with the similarity of the feature point data. In the update process λMk should be larger than λMt. Assume that (Mj-Nk-Mk) and (Mj 1-Nt-Mk 1) are two new instances of the meta-path (MNM). Similarly, nodes Nk and Nt are positively and negatively correlated point of interest data for node Nj. It is necessary to update λnj, λnk, λnt so that they follow the second order adjacency in the original iso-composition (i.e., the similarity information between the point of interest data corresponding to each meta-path information, the feature point data corresponding to each meta-path information). The first order proximity update based on the meta-path is as follows: where α is the learning rate. The gradient of λmj, λmk, λmt can be calculated as follows: also, the second order proximity update based on the meta-path is as follows:
The terminal takes the similarity information corresponding to each meta-path information as the proximity information corresponding to the meta-path information. And then, the terminal adds the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain the article recommendation graph network.
Based on the scheme, the adjacency information between the feature point data and the interest point data in the meta-path information is determined through the random gradient descent optimization strategy, so that the accuracy of the determined adjacency information is improved.
Optionally, based on the article recommendation graph network, calculating the similarity between each article and the user, screening the article with the highest similarity as the recommended article of the user, including: aiming at each article, calculating sub-similarity of each interest point data of the article and each feature point data of the user based on the interest point data of the article and the proximity information of meta-path information between the feature point data of the user; identifying distribution information of content information of each interest point data in the article, and determining weight information of each interest point data based on the distribution information; and carrying out weighted summation processing on the sub-similarity based on the weight information of each interest point data to obtain the similarity between the articles and the user, and screening articles corresponding to the highest similarity from all the articles to be used as recommended articles of the user.
In this embodiment, the terminal calculates sub-similarities corresponding to each of the interest point data of the article and each of the feature point data of the user, respectively, for each of the articles based on the interest point data of the article and the proximity information of the meta-path information between the feature point data of the user. And then the terminal determines the distribution information of each interest point data in the content information of the article by calculating the ratio of the number of text information of each interest point data in the content information of the article. And the terminal normalizes the ratio of the number of the text information of each interest point data in the article content information to obtain the weight information of each interest point data. And finally, the terminal performs weighted summation processing on the sub-similarity corresponding to each interest point data based on the weight information of each interest point data to obtain the similarity between the article and the user. Similarly, through the scheme, the terminal obtains the similarity between all articles and the user. And finally, the terminal screens articles corresponding to the highest similarity from all the articles to be used as recommended articles of the user.
Based on the scheme, the terminal identifies the similarity between each article and the user based on the article recommendation graph network, and accuracy of the identified similarity is improved, so that accuracy of recommending articles for the user is improved.
In one embodiment, as shown in FIG. 2, an article recommendation example is provided, the example comprising the steps of:
step S201, acquiring personal information of a user and content information of a plurality of articles.
Step S202, identifying text information of the text content of the text for each text, and based on the text keyword information of the text, inquiring the text information with similarity greater than a preset similarity threshold value in the text information of the text, wherein the similarity is greater than a preset similarity threshold value, and the text information is used as first interest point data of the text.
In step S203, the first interest point data of the article and the keyword information of the article are used as the interest point data of the article.
Step S204, identifying the published text information of the user, and extracting the first characteristic data of the published text information of the user through a text feature extraction network.
Step S205, based on the information of the non-article information in the personal information of the user, the study content information of the user is determined, the study content data corresponding to the study content information is extracted, and each of the first feature data and the study content data of the user is used as the feature point data of the user.
Step S206, for each article, calculating the inner product distance between each interest point data and each feature point data based on each interest point data of the article and each feature point data of the user through a heterogeneous information network, and determining meta-path information of each interest point data and each feature point data based on the inner product distance between each interest point data and each feature point data.
Step S207, determining a sub-relationship graph network of the articles and the users based on the meta-path information of all the interest point data of the articles and all the feature point data of the users, and determining a structural relationship graph network of the users and each article based on the sub-relationship graph network of all the articles and the users.
Step S208, identifying the similarity information between the interest point data corresponding to each meta-path information and the feature point data corresponding to each meta-path information in the structural relation graph network, and taking the similarity information corresponding to each meta-path information as the proximity information corresponding to the meta-path information.
And step S209, adding the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain the article recommendation graph network.
Step S210, for each article, calculating sub-similarity of each interest point data of the article and each feature point data of the user based on each interest point data of the article and the proximity information of meta-path information between each feature point data of the user.
Step S211, identifying distribution information of content information of each interest point data in the article, and determining weight information of each interest point data based on the distribution information.
Step S212, based on the weight information of each interest point data, weighting and summing the sub-similarity to obtain the similarity between the articles and the user, and screening the articles corresponding to the highest similarity from all the articles to be used as recommended articles of the user.
It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiment of the application also provides an article recommending device for realizing the article recommending method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitation in the embodiments of the article recommendation device or articles provided below may refer to the limitation of the article recommendation method hereinabove, and will not be repeated here.
In one embodiment, as shown in fig. 3, there is provided an article recommendation apparatus, including: an acquisition module 310, an extraction module 320, and a recommendation module 330, wherein:
an acquisition module 310, configured to acquire personal information of a user and content information of a plurality of articles;
the extracting module 320 is configured to extract each interest point data of the content information of each article, identify each feature point data of the personal information of the user, and establish a structural relationship graph network between the user and each article through a heterogeneous information network based on each interest point data of each article and each feature point data of the user;
and the recommendation module 330 is configured to determine, according to a gradient descent optimization policy, proximity information between the user and each article in the structural relationship graph network, obtain an article recommendation graph network of the user and each article, calculate similarity between each article and the user based on the article recommendation graph network, and screen an article with the highest similarity as a recommendation article of the user.
Optionally, the content information includes article content and keyword information, and the extracting module 320 is specifically configured to:
Identifying text information of the article content of each article aiming at each article, and inquiring text information with similarity greater than a preset similarity threshold value in the text information of the article based on the keyword information of the article to serve as first interest point data of the article;
and taking the first interest point data of the article and the keyword information of the article as the interest point data of the article.
Optionally, in the case that the personal information of the user includes the publication information of the user, the extracting module 320 is specifically configured to:
identifying the published text information of the user, and extracting first characteristic data of the published text information of the user through a text feature extraction network;
and determining research content information of the user based on information of non-article information in personal information of the user, extracting research content data corresponding to the research content information, and taking each piece of first characteristic data and the research content data of the user as characteristic point data of the user.
Optionally, in the case that the personal information of the user does not include article information of the user, the extracting module 320 is specifically configured to:
Acquiring article browsing history information of the user, and identifying content information of each piece of browsing article information corresponding to the article browsing history information;
and extracting each browsing interest point data of the content information of the browsing article information, identifying research content data in the personal information of the user, and taking each browsing interest point data and the research content data of the user as characteristic point data of the user.
Optionally, the extracting module 320 is specifically configured to:
for each article, calculating the inner product distance between each interest point data and each feature point data through a heterogeneous information network based on each interest point data of the article and each feature point data of the user, and determining meta-path information of each interest point data and each feature point data based on the inner product distance between each interest point data and each feature point data;
and determining a sub-relationship graph network of the articles and the user based on meta-path information of all the interest point data of the articles and all the characteristic point data of the user, and determining a structural relationship graph network of the user and each article based on the sub-relationship graph network of all the articles and the user.
Optionally, the recommendation module 330 is specifically configured to:
identifying similarity information between interest point data corresponding to each meta-path information and feature point data corresponding to each meta-path information in the structural relation graph network, and taking the similarity information corresponding to each meta-path information as proximity information corresponding to the meta-path information;
and adding the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain an article recommendation graph network.
Optionally, the recommendation module 330 is specifically configured to:
for each article, calculating sub-similarity of each interest point data of the article and each feature point data of the user based on each interest point data of the article and the proximity information of meta-path information between each feature point data of the user;
identifying distribution information of content information of each interest point data in the article, and determining weight information of each interest point data based on the distribution information;
and carrying out weighted summation processing on each sub-similarity based on the weight information of each interest point data to obtain the similarity between the article and the user, and screening articles corresponding to the highest similarity from all articles to be used as recommended articles of the user.
The respective modules in the article recommendation apparatus described above may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure of which may be as shown in fig. 4. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement an article recommendation method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by persons skilled in the art that the architecture shown in fig. 4 is merely a block diagram of some of the architecture relevant to the present inventive arrangements and is not limiting as to the computer device to which the present inventive arrangements are applicable, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
In an embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method of any of the first aspects when the computer program is executed.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the steps of the method of any of the first aspects.
In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method of any of the first aspects.
The user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.
Claims (11)
1. An article recommendation method, the method comprising:
acquiring personal information of a user and content information of a plurality of articles;
extracting each interest point data of content information of each article, identifying each feature point data of personal information of the user, and establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on each interest point data of each article and each feature point data of the user;
And determining the adjacency information of the user and each article in the structural relation graph network through a gradient descent optimization strategy to obtain an article recommendation graph network of the user and each article, calculating the similarity between each article and the user based on the article recommendation graph network, and screening articles with highest similarity as recommended articles of the user.
2. The method of claim 1, wherein the content information includes article content, and keyword information, and the extracting the respective interest point data of the content information of each article includes:
identifying text information of the article content of each article aiming at each article, and inquiring text information with similarity greater than a preset similarity threshold value in the text information of the article based on the keyword information of the article to serve as first interest point data of the article;
and taking the first interest point data of the article and the keyword information of the article as the interest point data of the article.
3. The method according to claim 1, wherein in the case where the personal information of the user includes the publication information of the user, the identifying each feature point data of the personal information of the user includes:
Identifying the published text information of the user, and extracting first characteristic data of the published text information of the user through a text feature extraction network;
and determining research content information of the user based on information of non-article information in personal information of the user, extracting research content data corresponding to the research content information, and taking each piece of first characteristic data and the research content data of the user as characteristic point data of the user.
4. The method according to claim 1, wherein in a case where article information of the user is not included in the personal information of the user, the identifying each feature point data of the personal information of the user includes:
acquiring article browsing history information of the user, and identifying content information of each piece of browsing article information corresponding to the article browsing history information;
and extracting each browsing interest point data of the content information of the browsing article information, identifying research content data in the personal information of the user, and taking each browsing interest point data and the research content data of the user as characteristic point data of the user.
5. The method of claim 1, wherein the establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on the respective point of interest data of each article and the respective feature point data of the user comprises:
for each article, calculating the inner product distance between each interest point data and each feature point data through a heterogeneous information network based on each interest point data of the article and each feature point data of the user, and determining meta-path information of each interest point data and each feature point data based on the inner product distance between each interest point data and each feature point data;
and determining a sub-relationship graph network of the articles and the user based on meta-path information of all the interest point data of the articles and all the characteristic point data of the user, and determining a structural relationship graph network of the user and each article based on the sub-relationship graph network of all the articles and the user.
6. The method of claim 5, wherein determining the proximity information of the user and each article in the structural relationship graph network by the gradient descent optimization strategy to obtain the article recommendation graph network of the user and each article comprises:
Identifying similarity information between interest point data corresponding to each meta-path information and feature point data corresponding to each meta-path information in the structural relation graph network, and taking the similarity information corresponding to each meta-path information as proximity information corresponding to the meta-path information;
and adding the proximity information corresponding to all the meta-path information to each meta-path information in the structural relation graph network to obtain an article recommendation graph network.
7. The method of claim 6, wherein the calculating the similarity between each article and the user based on the article recommendation graph network, and screening the article with the highest similarity as the recommended article of the user comprises:
for each article, calculating sub-similarity of each interest point data of the article and each feature point data of the user based on each interest point data of the article and the proximity information of meta-path information between each feature point data of the user;
identifying distribution information of content information of each interest point data in the article, and determining weight information of each interest point data based on the distribution information;
And carrying out weighted summation processing on each sub-similarity based on the weight information of each interest point data to obtain the similarity between the article and the user, and screening articles corresponding to the highest similarity from all articles to be used as recommended articles of the user.
8. An article recommendation device, the device comprising:
the acquisition module is used for acquiring personal information of a user and content information of a plurality of articles;
the extraction module is used for extracting the interest point data of the content information of each article, identifying the characteristic point data of the personal information of the user, and establishing a structural relationship graph network of the user and each article through a heterogeneous information network based on the interest point data of each article and the characteristic point data of the user;
and the recommending module is used for determining the proximity information of the user and each article in the structural relation diagram network through a gradient descent optimizing strategy to obtain an article recommending diagram network of the user and each article, calculating the similarity between each article and the user based on the article recommending diagram network, and screening articles with the highest similarity as recommending articles of the user.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
11. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310831925.XA CN116842171A (en) | 2023-07-07 | 2023-07-07 | Article recommendation method, apparatus, computer device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310831925.XA CN116842171A (en) | 2023-07-07 | 2023-07-07 | Article recommendation method, apparatus, computer device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116842171A true CN116842171A (en) | 2023-10-03 |
Family
ID=88163047
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310831925.XA Pending CN116842171A (en) | 2023-07-07 | 2023-07-07 | Article recommendation method, apparatus, computer device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116842171A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118069828A (en) * | 2024-04-22 | 2024-05-24 | 曲阜师范大学 | Article recommendation method based on heterogeneous graph and semantic fusion |
-
2023
- 2023-07-07 CN CN202310831925.XA patent/CN116842171A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118069828A (en) * | 2024-04-22 | 2024-05-24 | 曲阜师范大学 | Article recommendation method based on heterogeneous graph and semantic fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kang et al. | Kernel-driven similarity learning | |
| Qi et al. | Spatial-temporal data-driven service recommendation with privacy-preservation | |
| Zhang et al. | Enabling kernel-based attribute-aware matrix factorization for rating prediction | |
| Zandkarimi et al. | A generic framework for trace clustering in process mining | |
| Xing et al. | Points-of-interest recommendation based on convolution matrix factorization | |
| Lu et al. | GLR: A graph-based latent representation model for successive POI recommendation | |
| CN105320724A (en) | New heuristic for optimizing non-convex function for learning to rank | |
| CN116249991A (en) | Neural network distillation method and device | |
| Lyu et al. | Weighted multi-information constrained matrix factorization for personalized travel location recommendation based on geo-tagged photos | |
| Peng et al. | Multi-level preference regression for cold-start recommendations | |
| Wang et al. | POI recommendation method using LSTM-attention in LBSN considering privacy protection | |
| CN116842171A (en) | Article recommendation method, apparatus, computer device and storage medium | |
| Ma et al. | Exploring multiple spatio-temporal information for point-of-interest recommendation | |
| CN115879508A (en) | Data processing method and related device | |
| Sun | Personalized music recommendation algorithm based on spark platform | |
| CN116911867A (en) | Problem processing method, device, computer equipment and storage medium | |
| CN114168804B (en) | Similar information retrieval method and system based on heterogeneous subgraph neural network | |
| CN115758271A (en) | Data processing method, data processing device, computer equipment and storage medium | |
| Doukas | Technologies for greener internet of things systems | |
| Begum et al. | A hybrid model for optimum gene selection of microarray datasets | |
| CN114528491A (en) | Information processing method, information processing device, computer equipment and storage medium | |
| Liu et al. | An online activity recommendation approach based on the dynamic adjustment of recommendation lists | |
| CN115455276A (en) | Method and device for recommending object, computer equipment and storage medium | |
| Wang et al. | Bit selection via walks on graph for hash-based nearest neighbor search | |
| CN116881297A (en) | Cache content ordering method, device, computer equipment and storage medium |
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 |