CN112000795A - Official document recommendation method and device - Google Patents

Abstract

The invention discloses a method and a device for recommending official documents, and relates to the technical field of computers. One embodiment of the method comprises: acquiring an incidence relation among a plurality of documents; determining the official documents to be recommended in the multiple official documents according to the incidence relation and the characteristic data of the target user; and recommending the document to be recommended to the target user. According to the embodiment, the official documents can be recommended to the target user according to the incidence relation among the official documents, so that higher recommendation accuracy is achieved.

Description

Official document recommendation method and device

technical field

The present invention relates to the field of computer technology, and in particular, to a method and device for recommending official documents.

Background technique

Official documents are obviously different from news, information and other content, and are characterized by relatively fixed content, relatively complete elements, and relatively clear structure. In the current document recommendation technology, the traditional recommendation method is mainly used, that is, the recommendation is made based on the similarity between the recommended objects or the similarity between users, but in fact, different documents will show obvious differences based on the content. Logical association, the above recommendation method does not consider this association, so the accuracy is low.

SUMMARY OF THE INVENTION

In view of this, the embodiments of the present invention provide a method and apparatus for recommending official documents, which can recommend official documents to target users according to the association relationship between the official documents, thereby having high recommendation accuracy.

To achieve the above object, according to an aspect of the present invention, a method for recommending official documents is provided.

The method for recommending official documents according to the embodiment of the present invention includes: acquiring an association relationship between a plurality of official documents; determining an official document to be recommended among the plurality of official documents according to the association relationship and the feature data of the target user; sending the official document to be recommended to the target user recommend.

Optionally, the association relationship between the multiple official documents includes: at least one official document that has an association relationship with any official document in the multiple official documents, and the type of the association relationship; and, according to the association relationship and the feature data of the target user to determine the official document to be recommended among the multiple official documents, including: for any official document in the multiple official documents: determining according to the official document that has an associated relationship with the any official document and the type of the association relationship The feature vector of the any official document; according to the feature vector of the any official document and the feature data of the target user, determine the first potential degree of attention of the target user to the any official document; The comprehensive potential attention degree of any official document; when the comprehensive potential attention degree meets the preset recommendation conditions, any official document is determined as the official document to be recommended.

Optionally, the determining the feature vector of the any official document according to the official document having an associated relationship with the any official document and the type of the association relationship includes: for any official document in the plurality of official documents: the any official document The original feature of the official document, the feature vector of each official document that has an association relationship with the official document, and the preset weight value corresponding to the type of the association relationship are input into a preset function for calculation to obtain the feature vector of any official document.

Optionally, the determining the feature vector of any official document according to the official document having an associated relationship with the any official document and the type of the association relationship includes: combining the original features of the multiple official documents and the difference between the multiple official documents. The correlation data between the two is input into a pre-trained feature extraction model, and a feature vector of any one of the multiple official documents is obtained.

Optionally, the feature data of the target user includes the current feature vector of the target user; and, the method further includes: obtaining the initial feature vector of the target user; determining according to the initial feature vector of the target user and the historical behavior data of the target user for the official document The current feature vector of the target user.

Optionally, the obtaining the initial feature vector of the target user includes: determining the initial feature vector of the target user according to the preferred document category provided by the target user.

Optionally, determining the current feature vector of the target user according to the initial feature vector of the target user and the historical behavior data of the target user for the official document includes: acquiring the feature vector of each official document corresponding to the historical behavior data; The feature vector of the official document is multiplied by the impact score of the official document to obtain the impact vector of the official document; wherein, the impact score is determined according to the type of historical behavior of the target user with respect to the official document; according to the impact vector of each official document and the target user's initial The feature vector determines the current feature vector of the target user.

Optionally, the determining the first potential degree of attention of the target user to the any official document according to the feature vector of the any official document and the characteristic data of the target user includes: for any official document in the plurality of official documents, calculating The similarity between the feature vector of any official document and the current feature vector of the target user; the first potential degree of attention of the target user to the any official document is determined according to the similarity.

Optionally, the determining the first potential degree of attention of the target user to the any official document according to the feature vector of the any official document and the characteristic data of the target user includes: converting the feature vector of any official document in the plurality of official documents and the current feature vector of the target user into the pre-trained intent recognition model to obtain the attention probability of the target user to any official document; determine the first potential degree of attention of the target user to any official document according to the attention probability.

Optionally, the acquiring the comprehensive potential degree of attention of the target user to any official document based on the first potential degree of interest includes: determining the first potential degree of interest of the target user to any official document among the plurality of official documents. It is the comprehensive potential attention of the target user to any official document.

Optionally, the method further includes: for any official document in the plurality of official documents, calculating the similarity between the original feature of the any official document and the feature data of the target user; determining the similarity between the target user and the target user according to the similarity. The second potential degree of attention of an official document; and the acquiring, based on the first potential degree of attention, the comprehensive potential degree of attention of the target user to any official document includes: combining the first potential degree of attention and the second degree of potential attention The potential attention degree determines the comprehensive potential attention degree of the target user to any official document.

Optionally, the method further includes: inputting the original features of any official document among the multiple official documents and the comprehensive characteristics of the target user into a pre-trained intent discrimination model, to obtain the attention probability of the target user to any official document, according to The attention probability determines the second potential degree of attention of the target user to any official document; wherein, the comprehensive characteristics of the target user are determined by the historical behavior data of the target user on the official document and the attribute characteristics of the target user; Obtaining the comprehensive potential attention degree of the target user to any official document by a potential degree of attention includes: combining the first potential attention degree and the second potential attention degree to determine the comprehensive potential attention degree of the target user to any official document.

Optionally, the method further includes: for any official document in the plurality of official documents, determining a first time penalty item according to the time length between the release moment of the any official document and the current moment, and using the first time penalty item to determine. the third potential degree of attention of the target user to any official document; and the acquiring, based on the first potential degree of interest, the comprehensive potential degree of interest of the target user to the any official document includes: combining the first potential degree of interest , the second potential attention degree and the third potential attention degree determine the comprehensive potential attention degree of the target user to any official document.

Optionally, the method further includes: for any official document in the plurality of official documents: when the target user has browsed the any official document at a historical moment, according to the time when the target user browsed the any official document last time and the current The time length between the moments determines the second time penalty item; and, the use of the first time penalty item to determine the third potential degree of attention of the target user to any official document includes: using the first time penalty item and the second time penalty item determines the target user's third potential degree of attention to any official document.

Optionally, the method further includes: for any official document in the plurality of official documents, determining a fourth potential degree of attention of the any official document according to the number of views of the any official document in at least one historical time interval; and, The acquiring, based on the first potential degree of attention, the target user's comprehensive potential degree of attention to any official document includes: combining the first potential degree of attention, the second potential degree of attention, and the third potential degree of attention and the fourth potential degree of attention to determine the comprehensive potential degree of attention of the target user to any official document.

Optionally, the method further includes: when any official document in the plurality of official documents is related to a predetermined hot event, increasing the fourth potential degree of attention of the any official document according to a preset strategy.

Optionally, the method further includes: when any official document in the plurality of official documents and the target user satisfy at least one preset rule, determining the No. 1 position of the any official document according to the recommendation weight set in the preset rule. Five potential degrees of attention; the acquiring, based on the first potential degree of attention, the target user's comprehensive potential degree of attention to any official document includes: combining the first potential degree of attention, the second potential degree of attention, the The third potential interest degree, the fourth potential interest degree and the fifth potential interest degree determine the comprehensive potential interest degree of the target user on any official document.

Optionally, determining the fifth potential degree of attention of any official document according to the recommendation weight set in the preset rule includes: when the preset rule is one, recommending the recommendation set in the preset rule. The weight is determined as the fifth potential degree of attention of any official document; when there are multiple preset rules, the maximum recommended weight among the multiple preset rules is determined as the fifth potential degree of interest of the any official document.

Optionally, the acquiring the comprehensive potential degree of attention of the target user to any official document based on the first potential degree of attention includes: combining the first potential degree of attention, the second potential degree of attention, the first The weighted sum of the three potential attention degrees, the fourth potential attention degree and the fifth potential attention degree is determined as the comprehensive potential attention degree of the target user to any official document.

Optionally, when the comprehensive potential degree of attention meets a preset recommendation condition, determining any of the official documents as the official document to be recommended includes: sorting the plurality of official documents according to the comprehensive potential degree of interest from large to small. Arrange in order; determine the pre-arranged official documents as the official documents to be recommended.

Optionally, the characteristic data of the target user includes: official documents that the target user has browsed, favorited or subscribed to in a preset historical time period; The official documents to be recommended in the document include: determining the official document among the plurality of official documents that the target user browsed, favorited or subscribed to in the preset historical time period as the target official document; determining at least one official document that has an associated relationship with the target official document For the recommendation letter.

Optionally, the types of the association relationship include: revision relationship, revocation relationship, based on relationship, and mention relationship; and the types of the historical behavior include: browse, favorite, subscribe, ignore, unfavorite, and unsubscribe.

To achieve the above object, according to another aspect of the present invention, a device for recommending official documents is provided.

The apparatus for recommending official documents according to the embodiment of the present invention may include: an information extraction unit, configured to acquire an association relationship between a plurality of official documents; an official document determination unit, configured to determine the multiple official documents according to the association relationship and characteristic data of a target user The document to be recommended in ; the recommendation unit is used to recommend the document to be recommended to the target user.

To achieve the above object, according to yet another aspect of the present invention, an electronic device is provided.

An electronic device of the present invention comprises: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more programs make the One or more processors implement the document recommendation method provided by the present invention.

To achieve the above object, according to yet another aspect of the present invention, a computer-readable storage medium is provided.

A computer-readable storage medium of the present invention stores a computer program thereon, and when the program is executed by a processor, the method for recommending official documents provided by the present invention is implemented.

According to the technical solution of the present invention, the embodiments in the above-mentioned invention have the following advantages or beneficial effects:

First, first extract the association relationship between multiple official documents such as revision, revocation, basis, and mention, then determine the official document to be recommended based on this association relationship and the characteristic data of the target user, and finally recommend the official document to be recommended to the target user. . Based on the above settings, it is possible to fully consider the logical relationship between the official documents to determine the official document matching the target user, thereby improving the recommendation accuracy.

Second, when determining an official document to be recommended based on the relationship between the official documents, the embodiment of the present invention may first calculate the feature vector of any official document, and then determine the target user for the official document according to the feature vector of the official document and the feature data of the target user. The first potential degree of attention (that is, the potential degree of attention based on the relationship of the official document), and finally the official document to be recommended is determined based on the first potential degree of interest. When calculating the feature vector of any official document, the original feature of the official document, the feature vector of the official document that has an associated relationship with the official document, and the weight value corresponding to the type of the association relationship can be input into the preset function for direct calculation, or the official document can be directly calculated. The correlation data between the original features and official documents is input into the pre-trained feature extraction model for calculation. When calculating the first potential degree of attention, it can be based on the similarity between the feature vector of the official document and the feature data of the target user, or the focus of the target user on the official document obtained by inputting the feature vector of the official document and the feature data of the target user into the pre-trained intent recognition model. probability. Through the above settings, document feature calculation suitable for a variety of specific scenarios and document and user matching are realized.

Third, in the embodiment of the present invention, the feature data of the target user may be the current feature vector of the target user, and the current feature vector may follow the target user's browsing, collection, subscription, etc. for official documents on the basis of the initial feature vector of the target user. The behavior changes dynamically, so that when the target user browses, bookmarks or subscribes to a certain official document, its current feature vector may change towards the feature vector of the official document. A higher matching degree (the matching degree can be the similarity of the feature vector, the probability of attention, or the first potential degree of attention) will form a higher degree of matching between official documents that have an associated relationship, so that accurate and real-time recommendation of related official documents can be achieved.

Fourth, in the embodiment of the present invention, in addition to considering the relationship between official documents, collaborative recommendation can also be performed in combination with one or more of the following methods. These methods include: recommendation based on user behavior, time penalty item based , recommendation based on the number of official document browsing, and recommendation based on preset rules, which can further improve the recommendation accuracy.

Further effects of the above non-conventional alternatives will be described below in conjunction with specific embodiments.

Description of drawings

The accompanying drawings are used for better understanding of the present invention and do not constitute an improper limitation of the present invention. in:

1 is a schematic diagram of the main steps of a method for recommending official documents in an embodiment of the present invention;

2 is a schematic diagram of a knowledge map constructed by multiple official documents in an embodiment of the present invention;

3 is a schematic diagram of specific implementation steps of a method for recommending official documents in an embodiment of the present invention;

4 is a schematic diagram of the components of the official document recommendation device in the embodiment of the present invention;

5 is an exemplary system architecture diagram to which an embodiment of the present invention may be applied;

FIG. 6 is a schematic structural diagram of an electronic device used to implement the method for recommending official documents in an embodiment of the present invention.

Detailed ways

Exemplary embodiments of the present invention are described below with reference to the accompanying drawings, which include various details of the embodiments of the present invention to facilitate understanding and should be considered as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted from the following description for clarity and conciseness.

It should be pointed out that the embodiments of the present invention and the technical features in the embodiments may be combined with each other without conflict.

FIG. 1 is a schematic diagram of main steps of a method for recommending official documents according to an embodiment of the present invention.

As shown in FIG. 1 , the method for recommending official documents according to the embodiment of the present invention may be specifically performed according to the following steps:

Step S101: Acquire the association relationship between multiple official documents.

In the embodiment of the present invention, official documents refer to written materials formed by various types of institutions or organizations in a certain manner, including various policies, issued documents and internal documents of state agencies and government agencies at all levels, as well as public institutions, Various issued documents and internal documents of enterprise units. Documents can be displayed in various formats, such as text, pictures, audio, and video. Official documents generally have a specific form and are formed through certain processing procedures. At the same time, official documents often have content and logical associations between them. The types of associations can include revision relations, revocation relations, basis relations, and mention relations. For example, when a lower-level management functional department needs to form an official document B based on the official document A issued by the superior management functional department, there is a basis relationship between the official document B and the official document A, that is, the official document B is based on the official document A; when the official document F abolishes the official document B, Official document F and official document B have an annulment relationship; when official document F revises official document E, official document F and official document E have a revision relationship; when official document C refers to official document A, official document C and official document A have a reference to each other. relation. It should be noted that the relationship between official documents is not limited to the above four types, and can be expanded according to the actual situation.

FIG. 2 is a schematic diagram of a knowledge graph constructed by multiple documents in an embodiment of the present invention. In FIG. 2 , circular nodes represent documents, and edges between nodes represent associations between documents. As shown in Figure 2, document A and document B have a basis relationship (document B is based on document A), document E and document B have a basis relationship (document E is based on document B), and the relationship between document C and document A There is a reference relationship between official documents (document C mentions official document A), there is a revision relationship between official document F and official document E (official document F revised official document E), and official document F and official document C have a basis relationship (official document F is based on official document E). C), there is a basis relationship between official document D and official document B (official document D is based on official document B), there is a revision relationship between official document D and official document C (official document D revised official document C), and official document F and official document B have a nullification relationship Relationship (Official Document F repealed Official Document B).

It can be seen that the relationship between any two official documents is directional, and the directional edge with arrows is used in Figure 2 to represent this direction. Generally, for any two official documents with an associated relationship, the sending time of the official document pointed to by the arrow of the directional edge is earlier than the sending time of the other official document. It can be understood that the above-mentioned posting moment may correspond to any time precision, such as year, month, day, hour, minute or second. In addition, in specific scenarios, different correlations between official documents often represent different degrees of correlation between official documents. For example, the revocation relationship and the revision relationship represent a higher degree of correlation than the based relationship, and the based relationship represents a higher degree of correlation than the mention relationship. .

In particular, the association described above is a direct association between official documents. Besides, there may also be an indirect association between official documents. For example, official document F in FIG. 2 abolishes official document B, and official document B is based on official document A. Official document F and official document A have a second-level indirect association relationship (which can be referred to as a second-level association relationship), in which the first-level association relationship is abolished, and the second-level association relationship is the basis. This second-level association relationship can be Expressed as annulment - by relationship. In this way, the first-level related official documents (that is, the official documents that have a direct relationship with the any official document) and the multi-level related official documents (that is, the official documents that have a multi-level relationship with the any official document) can be obtained from the knowledge graph. . For example, the first-level related official documents of official document A are official documents B and C, and the second-level related official documents are official documents D, E, and F. If document A is the root node, documents B and C are at the first level, and documents D, E, and F are at the second level.

In step S101, a pre-trained information extraction model may be used to obtain the pre-acquired association relationship between multiple official documents for subsequent recommendation, and the above-mentioned association relationship may include: having an association with any one of the multiple official documents At least one document of the relationship and the type of the relationship. In practical applications, the text information of any one of the multiple official documents can be input into the information extraction model, so as to obtain the official document that has an association relationship with the official document and the type of the association relationship.

Step S102: Determine the official document to be recommended among the plurality of official documents according to the association relationship and the characteristic data of the target user.

In this step, the target user refers to the user for which the recommendation is directed, the official document to be recommended refers to the official document determined by calculation and to be recommended to the user, and the characteristic data of the target user refers to the data that can reflect the characteristics of the user. It may include one or more of the following aspects: the target user's gender, age, occupation, address and other attributes, and the preferred document category provided by the target user (document category can be divided into various dimensions, such as commands, decisions, resolutions, instructions, etc.), historical behavior data of target users for official documents. In practical applications, users can have a variety of behavior types for official documents, such as browsing, favorites, subscriptions, searching, watching (can be for official documents presented in the form of videos), feedback, ignoring, canceling favorites, and unsubscribing. The above historical behavior data can include: : The relevant data of the official documents that the target users browsed, favorited, subscribed, un-favorites and unsubscribed at historical moments, the duration of browsing a certain official document, the percentage of progress in viewing a certain official document, the search terms entered in the search box, the feedback information (such as evaluation information). Exemplarily, the relevant data of the official document may include the title, abstract, content, original feature and feature vector of the official document; wherein, the original feature of the official document refers to the use of TF-IDF (Term Frequency-Inverse File) based on the title, abstract or content of the official document. frequency), word2vec (a word vector model), BERT (a natural language processing model) and other algorithms or models to obtain official document features, the feature vector of official documents refers to the original features of official documents after considering the relationship between official documents. Document features.

In this embodiment of the present invention, step S102 may be specifically performed according to the following steps: for any official document among the plurality of official documents acquired in advance, first determine the official document with an associated relationship with the any official document and the type of the association relationship. A feature vector of an official document, then according to the feature vector of any official document and the feature data of the target user to determine the first potential degree of attention of the target user to any official document, and then based on the first potential degree of interest to obtain the target user for any one The comprehensive potential attention degree of the official document, and finally, when the comprehensive potential attention degree meets the preset recommendation conditions, any official document is determined as the official document to be recommended. Among them, the potential degree of attention can represent the probability that the user is interested in the official document, the first potential degree of interest is the potential degree of interest obtained after considering the relationship between the official document, and the comprehensive potential degree of interest is the probability of evaluating the user's interest in the official document from at least one aspect Potential attention indicators obtained afterward.

Preferably, in some embodiments, two methods can be used to calculate the feature vector of the official document. In the first method, for any official document among the pre-acquired multiple official documents, the original feature of the any official document, the feature vector of each official document that has an associated relationship with the any official document, and the association relationship may be The preset weight value corresponding to the type of is input into the preset function for calculation, so as to obtain the feature vector of any official document. Before performing the above calculation, you can determine which levels of related documents need to be selected for calculation according to the actual scenario, and set weight values for different related relationships according to the degree of correlation. Set a weight value of 0.3, a weight value of 0.1 for the mention relation, and a weight value of 0.03 for the based-mention relation.

In addition, in an optional implementation manner, an associated official document whose time of posting the document is earlier than any one of the official documents may be selected for calculation. For example, when it is necessary to select the related official documents of the first level for calculation, for official document F, official documents B, C, and E with earlier issuance time can be selected for calculation; for official document B, official document A with earlier issuance time can be selected. Calculation. When it is necessary to select the first-level and second-level related documents for calculation, for the official document F, the first-level related documents B, C, and E and the second-level related documents A and D can be selected for calculation. In another optional implementation manner, any related official documents of any official document can be selected for calculation, and the calculation is not limited to the official documents whose time of issuance is earlier than any official document. calculate. The following will take the case of selecting the first level of related documents for calculation as an example. It can be understood that when calculating the feature vector of a certain official document, you can select any level of related documents for calculation as needed, and the above example does not form any limit.

The above-mentioned preset function for calculating the feature vector of official documents can be set according to actual needs, for example, it can be set as follows:

Among them, M _i represents the feature vector of the official document i (i is the official document identification), M _i0 represents the original feature of the official document i, M _j represents the official document j that has an associated relationship with the official document i, and j is a positive integer starting from 1, indicating an association The serial number of the official document, k _j represents the weight value of the relationship between the official document j and the official document i, and p and q are preset parameters. It can be understood that the above formula is only an example, and does not form any limitation on the form of the above preset function.

In the second method of calculating the feature vector of official documents, the pre-acquired original features of multiple official documents and the correlation data between the official documents can be input into the pre-trained feature extraction model, so as to obtain the feature of any one of the multiple official documents. vector. Wherein, the association relationship data between the official documents may include the identification of the associated official document of any of the above-mentioned multiple official documents and the type of the corresponding association relationship, and the feature extraction model may be CNN (Convolutional Neural Networks, Convolutional Neural Networks), RNN ( Recurrent Neural Network, recurrent neural network) and other neural network models. In one embodiment, the following formula can be used inside the feature extraction model to calculate the feature vector of the official document:

表示特征提取模型中公文节点的特征向量，下标v表示公文节点的索引号，上标k表示该公文节点处在特征提取模型中的第几层，σ表示激活函数(例如线性整流函数ReLU)，W_k、B_k均为权重矩阵，N(v)表示公文节点v的关联公文集合，AGG(·)为聚合函数，可以通过计算平均值、加权和、卷积池化等方式收集关联公文的特征，AGG(·)中可以含有关联公文相应关联关系的权重值。在以上两种计算公文特征向量的方法中，均使用了公文的原始特征，这样能够解决系统的冷启动问题。in,

Represents the feature vector of the document node in the feature extraction model, the subscript v indicates the index number of the document node, the superscript k indicates the layer of the document node in the feature extraction model, and σ indicates the activation function (such as the linear rectification function ReLU) , W _k , B _k are weight matrices, N(v) represents the set of related documents of document node v, AGG( ) is an aggregation function, which can collect related documents by calculating the average value, weighted sum, convolution pooling, etc. The feature of AGG(·) can contain the weight value of the corresponding relationship of the related official document. In the above two methods for calculating the feature vector of the official document, the original features of the official document are used, which can solve the problem of cold start of the system.

In this embodiment of the present invention, the feature data of the target user may be the current feature vector of the target user, and the current feature vector may be dynamically updated according to the historical behavior data of the target user on official documents. It is calculated according to the following steps: First, the initial feature vector of the target user is obtained. In one embodiment, the initial feature vector of the target user can be determined according to the preferred document category provided by the target user. For example, the initial feature vector of the target user can be obtained by averaging the feature vectors of the preferred document category entered by the target user during registration. . Thereafter, the current feature vector of the target user can be determined according to the initial feature vector of the target user and the historical behavior data of the target user with respect to the official document. Specifically, the feature vector of each official document corresponding to the historical behavior data of the target user can be obtained first, and then the feature vector of any official document is multiplied by the impact score of the official document to obtain the impact vector of the official document, wherein the impact score can be It is determined according to the type of historical behavior of the target user on the official document. For example, the impact score of browsing can be set to 0.1, the impact score of favorites can be set to 0.2, the impact score of subscription can be set to 0.3, and the impact score of ignored can be set to -0.1, sets the impact score for unfavorites to -0.2, and for unsubscribes to -0.3. Finally, the current feature vector of the target user can be determined according to the influence vector of each official document and the initial feature vector of the target user. For example, the current feature vector of the target user can be obtained by adding the impact vector of each official document and the initial feature vector of the target user. .

According to the above settings, the user's current feature vector can be updated in real time with the user's behavior. In practical applications, after the target user browses, bookmarks or subscribes to a certain official document, the current feature vector of the target user may change towards the feature vector of the official document. Such changes will result in a higher degree of matching between the target user and the related official documents of the official document, so that accurate and real-time recommendation of related official documents can be achieved.

After the feature vector of the official document and the current feature vector of the target user are obtained, the first potential degree of attention of the target user to any one of the pre-acquired multiple official documents can be calculated. Two calculation methods are provided below. In a calculation method, for any official document among the plurality of official documents, the similarity between the feature vector of the any official document and the current feature vector of the target user can be calculated, and the target user can determine the similarity of the any official document according to the similarity. For example, multiply the similarity by a preset coefficient to obtain the first potential degree of attention, or directly use the similarity as the first potential degree of attention. Wherein, the above similarity may be cosine similarity, Jaccard similarity, Pearson correlation coefficient or adjusted cosine similarity,

In the second calculation method, the feature vector of any official document among the multiple official documents and the current feature vector of the target user can be input into the pre-trained intent recognition model, so as to obtain the attention probability of the target user to any official document, and finally based on The attention probability determines the first potential attention degree of the target user to any official document, for example, the attention probability is multiplied by a preset coefficient to obtain the first potential attention degree, or the attention probability is directly used as the first potential attention degree. The intent recognition model may be a classification model such as Naive Bayes and logistic regression. In particular, the above-mentioned intent recognition model can be connected with the aforementioned feature extraction model and trained together. For example, the output data of the feature extraction model and the current feature vector of the target user can be input into the intent recognition model, thereby realizing the integration between the feature extraction model and the intent recognition model. Connect, and determine the positive and negative samples according to the user’s feedback effect on the recommended official documents (such as whether to browse, browsing time, and evaluation information) at historical moments, and construct a training set based on the positive and negative positives. train.

In some embodiments, after obtaining the first potential degree of interest of the target user on any official document, the first potential degree of interest may be determined as the comprehensive potential degree of interest of the target user on any official document for subsequent recommendation.

Preferably, in the embodiment of the present invention, the first potential attention degree can be combined with the potential attention degrees of other aspects to calculate the comprehensive potential attention degree, so that a more accurate official document recommendation can be realized. In a specific application, for any official document among a plurality of official documents obtained in advance, the similarity between the original feature of the official document and the feature data of the target user can be calculated, and the similarity of the target user to the any official document can be determined according to the similarity. For the second potential degree of attention, for example, multiplying the similarity by a preset coefficient to obtain the second potential degree of attention, or directly using the similarity as the second potential degree of attention. The characteristic data of the target user can be calculated according to the following steps: first, determine the initial characteristic vector of the target user according to the preferred document category provided by the target user, then obtain the original characteristic of each document corresponding to the historical behavior data of the target user, and use The original feature of any official document is multiplied by the impact score of the official document to obtain the specific vector of the official document. Finally, the characteristic data of the target user can be obtained by adding the specific vector of each official document and the initial feature vector of the target user.

In practical applications, the original features of any official document and the comprehensive features of the target user can also be input into the pre-trained intent discrimination model, so as to obtain the probability of the target user's attention to any official document, and determine the target user's attention to any official document according to the probability of interest. For the second potential attention degree of an official document, for example, multiplying the attention probability by a preset coefficient to obtain the second potential attention degree, or directly using the attention probability as the second potential attention degree. Among them, the comprehensive characteristics of the target user can be determined by the historical behavior data of the target user for the official document and the attribute characteristics of the target user. The above comprehensive characteristics can be spliced into the following parts: Perform word vector calculation on the search word sequence of the target user behavior data And the data formed by the average pooling, the title and abstract of each official document viewed in the target user behavior data are calculated through the word vector and the pooled data is averaged, and the target user's name, gender, occupation and other attribute characteristic data. In a specific scenario, the above-mentioned intent discrimination model may be various deep learning models. When training the above-mentioned intent discrimination model, the positive or negative value may be determined according to the user's feedback effect on the recommended official document (for example, whether to browse or not, browsing time, and evaluation information) at historical moments. Samples, and then build a training set based on positive and negative positives to train the above intent discrimination model at the same time.

After obtaining the second potential degree of attention of the target user to any official document, the comprehensive potential degree of interest of the target user to any official document can be determined by combining the first potential degree of attention and the second potential degree of attention, for example, calculating the first potential degree of interest according to a preset method. The sum or weighted sum of the first potential attention degree and the second potential attention degree is regarded as the comprehensive potential attention degree, and the present invention does not impose any limitation on the method of calculating the comprehensive potential attention degree.

In practical application scenarios, relatively speaking, people pay more attention to newly released official documents. The longer a certain official document is published, the less people are interested in it. On the other hand, there is obviously no need for repeated recommendations for official documents that people have just read. Based on the above considerations, it is possible to establish a time penalty item based on the time of publication of the official document to determine the corresponding potential degree of attention. Specifically, for any official document, the first time penalty item can be determined according to the time length between the release time of the official document and the current moment, and the first time penalty item can be used to determine the third potential concern of the target user for the any official document degree (for example, the first time penalty item is directly used as the third potential attention degree). For example, the first time penalty term can be expressed as:

-log(t _publish ( _pi ))*punish _publish

Among them, _pi represents any official document, t _publish ( _pi ) represents the duration between the publication time of the official document and the current time, and punish _publish is an adjustable parameter.

In an optional implementation manner, for any official document, when the target user has browsed the any official document at a historical moment, the second can be determined according to the time length between the last time the target user browsed the any official document and the current moment. The time penalty item is calculated, and the third potential degree of attention of the target user to any official document is calculated based on the first time penalty item and the second time penalty item. The second time penalty term can be expressed as:

log ( _{Tt watch} (pi , _{u j )} +1)*punish _watch

Among them, T is a preset time constant, u _j represents the target user, and punch _watch is an adjustable parameter.

Exemplarily, calculating the third potential degree of attention of the target user to any official document based on the first time penalty item and the second time penalty item may be: adding the first time penalty item and the second time penalty item to the third potential Attention. After the third potential degree of attention is obtained, the first potential degree of attention, the second potential degree of attention and the third potential degree of attention can be combined to determine the comprehensive potential degree of attention of the target user to any official document, for example, calculating the first potential degree of attention according to a preset method. The sum or weighted sum of the first potential interest degree, the second potential interest degree and the third potential interest degree is taken as the comprehensive potential interest degree.

In some embodiments, for any official document among the plurality of official documents, the fourth potential degree of attention of any official document may also be determined according to the number of viewings of the any official document in at least one historical time interval, for example, the following formula can be used to perform calculate:

w ₄ (pi , _{u j )} =log(v _1h ( _pi ))*a _1h +log(v _1d ( _pi ))*a _1d +log(v _7d ( _pi ))*a _7d

Among them, pi represents the official document, _{u j represents the target user, w 4 (pi , u j ) represents the fourth} potential degree of attention, v _1h represents the number of views of the official document in the last hour, and v _1d represents the official document in the last hour. The number of views in one day, v _7d represents the number of views of the document in the last week, and a _1h , a _1d , and a _7d are preset parameters.

In practical applications, when any one of the multiple official documents is related to a predetermined hot event, the fourth potential degree of attention of any one of the official documents can be increased according to a preset strategy, for example, a _1h and a _1d in the above formula can be increased. , the value of a _7d to improve the fourth potential attention degree. After the fourth potential degree of attention is obtained, the first potential degree of attention, the second potential degree of attention, the third potential degree of attention, and the fourth potential degree of attention may be combined to determine the comprehensive potential degree of attention of the target user to any official document, for example The sum or weighted sum of the first potential attention degree, the second potential attention degree, the third potential attention degree, and the fourth potential attention degree is calculated according to a preset method as the comprehensive potential attention degree.

In a specific application, in order to achieve specific recommendation requirements, it can support the recommendation of a specified official document or a specified category of official documents to users of a specified range or a specified category by confirming the task. Therefore, in this embodiment of the present invention, when any official document among the multiple official documents and the target user satisfy at least one preset rule, the fifth potential degree of attention of any official document is determined according to the recommendation weight set in the preset rule. Specifically, when there is one preset rule that is satisfied, the recommendation weight set in the preset rule may be determined as the fifth potential degree of attention of any official document; when there are multiple preset rules that are satisfied, you can The maximum value of the recommendation weights in the plurality of preset rules is determined as the fifth potential degree of attention of any official document.

After the fifth potential degree of attention is obtained, the first potential degree of attention, the second potential degree of attention, the third potential degree of attention, the fourth potential degree of attention, and the fifth potential degree of attention can be combined to determine the target user's interest in any official document Comprehensive potential attention degree, for example, calculating the sum or weighted sum of the first potential attention degree, the second potential attention degree, the third potential attention degree, the fourth potential attention degree, and the fifth potential attention degree according to a preset method as the comprehensive potential attention degree .

After obtaining the comprehensive potential degree of attention of the target user to any official document, the official document that meets the preset recommendation conditions may be determined as the official document to be recommended. For example, a plurality of official documents may be arranged in descending order of comprehensive potential attention degree, and a preset number of official documents arranged in the front may be determined as the official documents to be recommended.

In particular, in the embodiment of the present invention, the recommendation can be implemented without relying on the feature vector of the official document. Specifically, when the target user has browsed, bookmarked or subscribed to the official documents in the aforementioned official documents in a preset historical time period (for example, the last hour), these official documents are determined as the target official documents, and at least the official documents that have an associated relationship with the target official documents are determined. An official document is determined as the official document to be recommended. For example, in Fig. 2, when the target user has browsed document A in the last hour, the related documents B and C of the first level of document A can be determined as the documents to be recommended. Of course, the first level of document A can also be determined. The related official documents B and C of , and the second-level related official documents D, E, and F are determined as the official documents to be recommended. Through the above settings, simple and fast recommendation can be implemented based on the relationship between official documents.

Step S103: Recommend the official document to be recommended to the target user. In this step, the official document to be recommended may be pushed to the terminal corresponding to the user, thereby completing the official document recommendation.

FIG. 3 is a schematic diagram of specific implementation steps of a method for recommending official documents in an embodiment of the present invention. As shown in FIG. 3 , in an embodiment of the present invention, the first potential attention degree, the second potential attention degree, and the third potential attention degree can be calculated through five paths. The attention degree, the fourth potential attention degree, and the fifth potential attention degree are calculated, and the comprehensive potential attention degree is calculated based on the above potential attention degree, and finally the official document is recommended based on the comprehensive potential attention degree. When calculating the first potential degree of attention, first obtain the official document association formula, and use a preset function or feature extraction model to calculate the feature vector of the official document, and then calculate the similarity or use the intent recognition model to obtain the first potential degree of attention. . When calculating the second potential attention degree, the method of calculating the similarity degree or the intention discrimination model can be used. When calculating the third potential attention degree, the time penalty terms of the two aspects can be combined. And, the fourth potential degree of attention may be calculated using the number of official document views, and the fifth potential degree of interest may be calculated using a preset rule.

In the technical solution of the embodiment of the present invention, firstly, the association relationship between multiple official documents such as revision, revocation, basis, and mention is extracted, then the official document to be recommended is determined based on this association relationship and the characteristic data of the target user, and finally the official document to be recommended is determined. The official document to be recommended is recommended to the target user. Based on the above settings, it is possible to fully consider the logical relationship between the official documents to determine the official document matching the target user, thereby improving the recommendation accuracy. When determining the official document to be recommended based on the relationship between the official documents, the embodiment of the present invention may first calculate the feature vector of any official document, and then determine the target user's first feature vector of the official document according to the feature vector of the official document and the feature data of the target user. The potential attention degree, and finally the official document to be recommended is determined based on the first potential attention degree. When calculating the feature vector of any official document, the original feature of the official document, the feature vector of the official document that has an associated relationship with the official document, and the weight value corresponding to the type of the association relationship can be input into the preset function for direct calculation, or the official document can be directly calculated. The correlation data between the original features and official documents is input into the pre-trained feature extraction model for calculation. When calculating the first potential degree of attention, it can be based on the similarity between the feature vector of the official document and the feature data of the target user, or the focus of the target user on the official document obtained by inputting the feature vector of the official document and the feature data of the target user into the pre-trained intent recognition model. probability. Through the above settings, document feature calculation suitable for a variety of specific scenarios and document and user matching are realized. In this embodiment of the present invention, the feature data of the target user may be the current feature vector of the target user, and the current feature vector may dynamically change with the target user's behaviors such as browsing, saving, and subscribing to official documents on the basis of the initial feature vector of the target user. , in this way, after the target user browses, collects or subscribes to a certain document, its current feature vector may change towards the feature vector of the official document, and this change will make a higher relationship between the target user and the document associated with the official document. Matching degree, so that accurate related document recommendation and real-time recommendation can be achieved. In this embodiment of the present invention, in addition to considering the relationship between official documents, collaborative recommendation can also be performed in combination with one or more of the following methods. These methods include: recommendation based on user behavior, recommendation based on time penalty items, Recommendations based on the number of official document views and recommendations based on preset rules can further improve the accuracy of recommendations.

It should be noted that, for the convenience of description, the foregoing method embodiments are expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, some The steps may in fact be performed in other orders or simultaneously. In addition, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the related actions and modules are not necessarily necessary to realize the present invention.

In order to better implement the above solutions of the embodiments of the present invention, related devices for implementing the above solutions are also provided below.

Referring to FIG. 4 , an apparatus for recommending official documents 400 provided in an embodiment of the present invention may include: an information extracting unit 401 , a document determining unit 402 , and a recommending unit 403 .

Wherein, the information extraction unit 401 can be used to obtain the association relationship between multiple official documents; the official document determination unit 402 can be used to determine the to-be-recommended official document among the multiple official documents according to the association relationship and the feature data of the target user; the recommendation unit 403 It is used to recommend official documents to be recommended to target users.

In this embodiment of the present invention, the association relationship between the multiple official documents includes: at least one official document that has an association relationship with any official document in the multiple official documents and the type of the association relationship; the official document determining unit 402 may further Used for: for any official document in the plurality of official documents: determine the feature vector of the any official document according to the official document that has an associated relationship with the any official document and the type of the association relationship; according to the feature vector of the any official document and The feature data of the target user determines the first potential degree of attention of the target user to any official document; based on the first potential degree of interest, the comprehensive potential degree of interest of the target user to the any official document is obtained; When the preset recommendation condition is used, any official document is determined as the official document to be recommended.

In some embodiments, the official document determining unit 402 may be further configured to: for any official document in the plurality of official documents: the original feature of the any official document and the characteristics of each official document that has an associated relationship with the any official document The vector and the preset weight value corresponding to the type of the association relationship are input into the preset function for calculation, and the feature vector of any official document is obtained.

Preferably, the official document determining unit 402 may be further configured to: input the original features of the multiple official documents and the correlation data between the multiple official documents into a pre-trained feature extraction model, and obtain any of the multiple official documents. The feature vector of a document.

In a specific application, the feature data of the target user includes the current feature vector of the target user, and the apparatus 400 may further include a target user feature calculation unit, which is used for: acquiring the initial feature vector of the target user; according to the initial feature vector of the target user and The target user determines the current feature vector of the target user according to the historical behavior data of the official document.

In practical applications, the target user feature calculation unit may be further configured to: determine the initial feature vector of the target user according to the preferred document category provided by the target user.

As a preferred solution, the target user feature calculation unit can be further used to: obtain the feature vector of each official document corresponding to the historical behavior data; multiply the feature vector of any official document by the impact score of the official document to obtain the impact of the official document The influence score is determined according to the type of historical behavior of the target user on the official document; the current feature vector of the target user is determined according to the influence vector of each official document and the initial feature vector of the target user.

In a specific application scenario, the official document determining unit 402 may be further configured to: for any official document in the plurality of official documents, calculate the similarity between the feature vector of the any official document and the current feature vector of the target user; determine the similarity according to the similarity The first potential degree of attention of the target user to any official document.

In some embodiments, the official document determining unit 402 may be further configured to: input the feature vector of any official document among the plurality of official documents and the current feature vector of the target user into the pre-trained intent recognition model, and obtain the target user for any one of the official documents. The attention probability of the official document; the first potential attention degree of the target user to any official document is determined according to the attention probability.

In the embodiment of the present invention, the official document determining unit 402 may be further configured to: determine the first potential degree of attention of the target user to any official document among the plurality of official documents as the comprehensive potential degree of interest of the target user to the any official document.

In an optional implementation manner, the apparatus 400 may further include a second attention degree calculation unit, which is used for: for any official document in the plurality of official documents, calculate the original feature of the any official document and the target user's original feature. The similarity of the feature data; the second potential degree of attention of the target user to any official document is determined according to the similarity; the official document determination unit 402 may be further configured to: combine the first potential degree of attention and the second potential degree of attention Determine the comprehensive potential attention of the target user to any official document.

In some embodiments, the second attention degree calculation unit may be further configured to: input the original feature of any official document among the plurality of official documents and the comprehensive feature of the target user into the pre-trained intention discrimination model, and obtain the target user's interest in any one of the official documents. The attention probability of an official document, the second potential degree of attention of the target user to any official document is determined according to the attention probability; wherein, the comprehensive feature of the target user is determined by the historical behavior data of the target user on the official document and the attribute characteristics of the target user; the official document The determining unit 402 may be further configured to: combine the first potential degree of interest and the second potential degree of interest to determine the comprehensive potential degree of interest of the target user to any official document.

In an optional implementation manner, the apparatus 400 may further include a third attention degree calculation unit, which is used for: for any official document in the plurality of official documents, according to the time between the publication time of the any official document and the current time Determine the first time penalty item, and use the first time penalty item to determine the third potential degree of attention of the target user to any official document. The official document determining unit 402 may be further configured to: combine the first potential interest degree, the second potential interest degree and the third potential interest degree to determine the comprehensive potential interest degree of the target user on any official document.

Preferably, the third attention degree calculation unit may be further configured to: for any official document among the plurality of official documents: when the target user has browsed the any official document at a historical moment, according to the last time the target user browsed the any official document. The duration between the moment of the official document and the current moment determines the second time penalty item; and the first time penalty item and the second time penalty item are used to determine the third potential degree of attention of the target user to any official document.

In this embodiment of the present invention, the apparatus 400 may further include a fourth attention degree calculation unit, which is configured to: for any official document in the plurality of official documents, according to the browsing of the any official document in at least one historical time interval The fourth potential degree of attention of any official document is determined by the number; the official document determination unit 402 may be further configured to: combine the first potential degree of interest, the second potential degree of interest, the third potential degree of interest and the third potential degree of interest Four potential attention degrees Determine the comprehensive potential attention degree of the target user to any official document.

Preferably, the fourth degree of attention calculation unit may be further configured to: when any official document in the plurality of official documents is related to a predetermined hot event, increase the fourth potential degree of interest of the any official document according to a preset strategy. .

As a preferred solution, the apparatus 400 may further include a fifth attention degree calculation unit, which is configured to: when any official document in the plurality of official documents and the target user satisfy at least one preset rule, according to the preset The recommendation weight set in the rule determines the fifth potential degree of attention of any official document; the official document determination unit 402 may be further configured to: combine the first potential degree of interest, the second potential degree of interest, and the third potential degree of interest degree, the fourth potential degree of attention and the fifth potential degree of attention determine the comprehensive potential degree of attention of the target user to any official document.

Preferably, the fifth degree of attention calculation unit may be further configured to: when the preset rule is one, determine the recommendation weight set in the preset rule as the fifth potential degree of attention of any official document; When there are multiple preset rules, the maximum value of the recommendation weight in the multiple preset rules is determined as the fifth potential degree of attention of any official document.

In this embodiment of the present invention, the official document determining unit 402 may be further configured to: determine the first potential attention degree, the second potential attention degree, the third potential attention degree, the fourth potential attention degree and the The weighted sum of the five potential attention degrees is determined as the comprehensive potential attention degree of the target user to any official document.

In one embodiment, the official document determining unit 402 may be further configured to: arrange the plurality of official documents in descending order of comprehensive potential attention;

In practical applications, the characteristic data of the target user includes: official documents that the target user has browsed, collected or subscribed to in a preset historical time period; the official document determining unit 402 can be further used for: An official document among the plurality of official documents that has been collected or subscribed is determined as a target official document; at least one official document having an associated relationship with the target official document is determined as an official document to be recommended.

In addition, in the embodiment of the present invention, the types of the association relationship include: revision relationship, abolition relationship, based on relationship, and mention relationship; the types of historical behavior include: browse, favorite, subscribe, ignore, unfavorite, and cancel subscription.

In the technical solution of the embodiment of the present invention, firstly, the association relationship between multiple official documents such as revision, revocation, basis, and mention is extracted, then the official document to be recommended is determined based on this association relationship and the characteristic data of the target user, and finally the official document to be recommended is determined. The official document to be recommended is recommended to the target user. Based on the above settings, it is possible to fully consider the logical relationship between the official documents to determine the official document matching the target user, thereby improving the recommendation accuracy. When determining the official document to be recommended based on the relationship between the official documents, the embodiment of the present invention may first calculate the feature vector of any official document, and then determine the target user's first feature vector of the official document according to the feature vector of the official document and the feature data of the target user. The potential attention degree, and finally the official document to be recommended is determined based on the first potential attention degree. When calculating the feature vector of any official document, the original feature of the official document, the feature vector of the official document that has an associated relationship with the official document, and the weight value corresponding to the type of the association relationship can be input into the preset function for direct calculation, or the official document can be directly calculated. The correlation data between the original features and official documents is input into the pre-trained feature extraction model for calculation. When calculating the first potential degree of attention, it can be based on the similarity between the feature vector of the official document and the feature data of the target user, or the focus of the target user on the official document obtained by inputting the feature vector of the official document and the feature data of the target user into the pre-trained intent recognition model. probability. Through the above settings, document feature calculation suitable for a variety of specific scenarios and document and user matching are realized. In this embodiment of the present invention, the feature data of the target user may be the current feature vector of the target user, and the current feature vector may dynamically change with the target user's behaviors such as browsing, saving, and subscribing to official documents on the basis of the initial feature vector of the target user. , in this way, after the target user browses, collects or subscribes to a certain document, its current feature vector may change towards the feature vector of the official document, and this change will make a higher relationship between the target user and the document associated with the official document. Matching degree, so that accurate related document recommendation and real-time recommendation can be achieved. In this embodiment of the present invention, in addition to considering the relationship between official documents, collaborative recommendation can also be performed in combination with one or more of the following methods. These methods include: recommendation based on user behavior, recommendation based on time penalty items, Recommendations based on the number of official document views and recommendations based on preset rules can further improve the accuracy of recommendations.

FIG. 5 shows an exemplary system architecture 500 to which the method for recommending official documents or the apparatus for recommending official documents according to embodiments of the present invention can be applied.

As shown in FIG. 5, the system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505 (this architecture is only an example, and the components included in the specific architecture can be adjusted according to the specific application). The network 504 is a medium used to provide a communication link between the terminal devices 501 , 502 , 503 and the server 505 . Network 504 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user can use the terminal devices 501, 502, 503 to interact with the server 505 through the network 504 to receive or send messages and the like. Various client applications may be installed on the terminal devices 501 , 502 , and 503 , such as an official document reading application (just an example).

The terminal devices 501, 502, 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

The server 505 may be a server that provides various services, such as a background server (just an example) that provides support for the document reading application operated by the user using the terminal devices 501 , 502 , and 503 . The background server may process the received official document recommendation request, and feed back the processing result (for example, the official document to be recommended—just an example) to the terminal devices 501 , 502 , and 503 .

It should be noted that the method for recommending official documents provided by the embodiments of the present invention is generally executed by the server 505 , and accordingly, the device for recommending official documents is generally set in the server 505 .

It should be understood that the numbers of terminal devices, networks and servers in FIG. 5 are only illustrative. There can be any number of terminal devices, networks and servers according to implementation needs.

The present invention also provides an electronic device. An electronic device according to an embodiment of the present invention includes: one or more processors; and a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more programs cause the One or more processors implement the document recommendation method provided by the present invention.

Referring next to FIG. 6 , it shows a schematic structural diagram of a computer system 600 suitable for implementing an electronic device according to an embodiment of the present invention. The electronic device shown in FIG. 6 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present invention.

As shown in FIG. 6, a computer system 600 includes a central processing unit (CPU) 601, which can be loaded into a random access memory (RAM) 603 according to a program stored in a read only memory (ROM) 602 or a program from a storage section 608 Instead, various appropriate actions and processes are performed. In the RAM 603, various programs and data necessary for the operation of the computer system 600 are also stored. The CPU 601 , the ROM 602 , and the RAM 603 are connected to each other through a bus 604 . An input/output (I/O) interface 605 is also connected to bus 604 .

The following components are connected to the I/O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc. ; and a communication section 609 including a network interface card such as a LAN card, a modem, and the like. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I/O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

In particular, according to the disclosed embodiments of the present invention, the processes described in the main step diagrams above may be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for executing the method shown in the main step diagram. In the above-described embodiment, the computer program can be downloaded and installed from the network through the communication section 609 and/or installed from the removable medium 611 . When the computer program is executed by the central processing unit 601, the above-described functions defined in the system of the present invention are performed.

It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples of computer readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable Programmable read only memory (EPROM or flash memory), fiber optics, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. In the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer readable program code therein. Such propagated data signals may take a variety of forms including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium, other than a computer-readable storage medium, that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any suitable medium including, but not limited to, wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more logical functions for implementing the specified functions executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams or flowchart illustrations, and combinations of blocks in the block diagrams or flowchart illustrations, can be implemented in special purpose hardware-based systems that perform the specified functions or operations, or can be implemented using A combination of dedicated hardware and computer instructions is implemented.

The units involved in the embodiments of the present invention may be implemented in a software manner, and may also be implemented in a hardware manner. The described unit can also be set in the processor, for example, it can be described as: a processor includes an information extraction unit, an official document determination unit and a recommendation unit. Wherein, the names of these units do not constitute a limitation on the unit itself under certain circumstances. For example, the information extraction unit can also be described as "a unit that provides the official document determination unit with the relationship between official documents".

As another aspect, the present invention also provides a computer-readable medium. The computer-readable medium may be included in the device described in the above embodiments; it may also exist independently without being assembled into the device. . The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the device, the steps performed by the device include: acquiring the association relationship between multiple official documents; according to the association relationship and The feature data of the target user determines the official documents to be recommended among the plurality of official documents; the official documents to be recommended are recommended to the target user.

In the technical solution of the embodiment of the present invention, firstly, the association relationship between multiple official documents such as revision, revocation, basis, and mention is extracted, then the official document to be recommended is determined based on this association relationship and the characteristic data of the target user, and finally the official document to be recommended is determined. The official document to be recommended is recommended to the target user. Based on the above settings, it is possible to fully consider the logical relationship between the official documents to determine the official document matching the target user, thereby improving the recommendation accuracy. When determining the official document to be recommended based on the relationship between the official documents, the embodiment of the present invention may first calculate the feature vector of any official document, and then determine the target user's first feature vector of the official document according to the feature vector of the official document and the feature data of the target user. The potential attention degree, and finally the official document to be recommended is determined based on the first potential attention degree. When calculating the feature vector of any official document, the original feature of the official document, the feature vector of the official document that has an associated relationship with the official document, and the weight value corresponding to the type of the association relationship can be input into the preset function for direct calculation, or the official document can be directly calculated. The correlation data between the original features and official documents is input into the pre-trained feature extraction model for calculation. When calculating the first potential degree of attention, it can be based on the similarity between the feature vector of the official document and the feature data of the target user, or the focus of the target user on the official document obtained by inputting the feature vector of the official document and the feature data of the target user into the pre-trained intent recognition model. probability. Through the above settings, document feature calculation suitable for a variety of specific scenarios and document and user matching are realized. In this embodiment of the present invention, the feature data of the target user may be the current feature vector of the target user, and the current feature vector may dynamically change with the target user's behaviors such as browsing, saving, and subscribing to official documents on the basis of the initial feature vector of the target user. , in this way, after the target user browses, collects or subscribes to a certain document, its current feature vector may change towards the feature vector of the official document, and this change will make a higher relationship between the target user and the document associated with the official document. Matching degree, so that accurate related document recommendation and real-time recommendation can be achieved. In this embodiment of the present invention, in addition to considering the relationship between official documents, collaborative recommendation can also be performed in combination with one or more of the following methods. These methods include: recommendation based on user behavior, recommendation based on time penalty items, Recommendations based on the number of official document views and recommendations based on preset rules can further improve the accuracy of recommendations.

The above-mentioned specific embodiments do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (25)

1. An official document recommendation method, comprising:

acquiring an incidence relation among a plurality of documents;

determining the official documents to be recommended in the multiple official documents according to the incidence relation and the characteristic data of the target user;

and recommending the document to be recommended to the target user.

2. The method of claim 1, wherein the association between the plurality of documents comprises: at least one official document having an association relation with any one of the plurality of official documents and a type of the association relation; and determining the official documents to be recommended in the plurality of official documents according to the incidence relation and the characteristic data of the target user, wherein the method comprises the following steps:

for any of the plurality of documents:

determining a feature vector of any document according to the document having an association relation with the document and the type of the association relation;

determining a first potential attention degree of the target user to any official document according to the feature vector of the official document and the feature data of the target user;

acquiring the comprehensive potential attention degree of the target user to any official document based on the first potential attention degree;

and when the comprehensive potential attention degree meets a preset recommendation condition, determining any official document as an official document to be recommended.

3. The method of claim 2, wherein determining the feature vector of the document according to the document having an association relationship with the document and the type of the association relationship comprises:

for any of the plurality of documents:

inputting the original features of any document, the feature vector of each document having an association relation with the document and the preset weight value corresponding to the type of the association relation into a preset function for calculation to obtain the feature vector of any document.

4. The method of claim 2, wherein determining the feature vector of the document according to the document having an association relationship with the document and the type of the association relationship comprises:

inputting the original features of the multiple documents and the incidence relation data between the multiple documents into a pre-trained feature extraction model to obtain the feature vector of any document in the multiple documents.

5. The method of claim 2, wherein the feature data of the target user comprises a current feature vector of the target user; and, the method further comprises:

acquiring an initial feature vector of a target user;

and determining the current characteristic vector of the target user according to the initial characteristic vector of the target user and the historical behavior data of the target user for the official document.

6. The method of claim 5, wherein the obtaining the initial feature vector of the target user comprises:

and determining the initial characteristic vector of the target user according to the preference document category provided by the target user.

7. The method of claim 5, wherein determining the current feature vector of the target user according to the initial feature vector of the target user and historical behavior data of the target user for the official document comprises:

acquiring a feature vector of each official document corresponding to the historical behavior data;

multiplying the feature vector of any document with the influence fraction of the document to obtain the influence vector of the document; the influence score is determined according to the type of the historical behaviors of the target user aiming at the official document;

and determining the current feature vector of the target user according to the influence vector of each official document and the initial feature vector of the target user.

8. The method of claim 5, wherein determining the first potential degree of interest of the target user for the any document according to the feature vector of the any document and the feature data of the target user comprises:

for any document in the plurality of documents, calculating the similarity between the feature vector of the document and the current feature vector of the target user;

and determining a first potential attention degree of the target user to any document according to the similarity.

9. The method of claim 5, wherein determining the first potential degree of interest of the target user for the any document according to the feature vector of the any document and the feature data of the target user comprises:

inputting the feature vector of any one of the multiple documents and the current feature vector of the target user into a pre-trained intention recognition model to obtain the attention probability of the target user to the any document;

and determining a first potential attention degree of the target user to the any official document according to the attention probability.

10. The method of claim 2, wherein the obtaining of the comprehensive potential attention of the target user to the arbitrary document based on the first potential attention comprises:

and determining the first potential attention degree of the target user to any one official document in the plurality of official documents as the comprehensive potential attention degree of the target user to the any one official document.

11. The method of claim 2,

the method further comprises: for any official document in the multiple official documents, calculating the similarity between the original characteristic of the official document and the characteristic data of the target user; determining a second potential attention degree of the target user to any official document according to the similarity degree; and the number of the first and second groups,

the acquiring of the comprehensive potential attention degree of the target user to any official document based on the first potential attention degree comprises: and determining the comprehensive potential attention degree of the target user to any document by combining the first potential attention degree and the second potential attention degree.

12. The method of claim 2,

the method further comprises: inputting the original features of any one of the multiple documents and the comprehensive features of the target user into a pre-trained intention judging model to obtain the attention probability of the target user to the any one document, and determining a second potential attention degree of the target user to the any one document according to the attention probability; the comprehensive characteristics of the target user are determined by the historical behavior data of the target user aiming at the official document and the attribute characteristics of the target user; and the number of the first and second groups,

the acquiring of the comprehensive potential attention degree of the target user to any official document based on the first potential attention degree comprises: and determining the comprehensive potential attention degree of the target user to any document by combining the first potential attention degree and the second potential attention degree.

13. The method according to claim 11 or 12, characterized in that the method further comprises:

for any official document in the multiple official documents, determining a first time penalty item according to the time length between the issuing time of the any official document and the current time, and determining a third potential attention degree of a target user to the any official document by using the first time penalty item; and

the acquiring of the comprehensive potential attention degree of the target user to any official document based on the first potential attention degree comprises:

and determining the comprehensive potential attention degree of the target user to any document by combining the first potential attention degree, the second potential attention degree and the third potential attention degree.

14. The method of claim 13,

the method further comprises: for any of the plurality of documents: when the target user browses any official document at the historical moment, determining a second time penalty item according to the time length between the moment when the target user browses any official document at the last time and the current moment; and

the determining a third potential attention degree of the target user to any document by using the first time penalty item comprises the following steps: and determining a third potential attention degree of the target user to any document by using the first time penalty item and the second time penalty item.

15. The method of claim 13,

the method further comprises: for any official document in the multiple official documents, determining a fourth potential attention degree of the any official document according to the browsing number of the any official document in at least one historical time interval; and

the acquiring of the comprehensive potential attention degree of the target user to any official document based on the first potential attention degree comprises: and determining the comprehensive potential attention degree of the target user to any document by combining the first potential attention degree, the second potential attention degree, the third potential attention degree and the fourth potential attention degree.

16. The method of claim 15, further comprising:

and when any document in the plurality of documents is related to a predetermined hot spot event, increasing the fourth potential attention of the document according to a preset strategy.

17. The method of claim 15,

the method further comprises: when any official document in the multiple official documents and a target user meet at least one preset rule, determining a fifth potential attention degree of the official document according to a recommendation weight set in the preset rule;

the acquiring of the comprehensive potential attention degree of the target user to any official document based on the first potential attention degree comprises: and determining the comprehensive potential attention degree of the target user to any document by combining the first potential attention degree, the second potential attention degree, the third potential attention degree, the fourth potential attention degree and the fifth potential attention degree.

18. The method according to claim 17, wherein the determining a fifth potential attention degree of any document according to the recommendation weight set in the preset rule comprises:

when the preset rule is one, determining the recommendation weight set in the preset rule as a fifth potential attention degree of any official document;

and when the preset rules are multiple, determining the maximum value of the recommendation weights in the multiple preset rules as the fifth potential attention of any document.

19. The method of claim 17, wherein the obtaining of the comprehensive potential attention of the target user to the arbitrary document based on the first potential attention comprises:

and determining the weighted sum of the first potential attention degree, the second potential attention degree, the third potential attention degree, the fourth potential attention degree and the fifth potential attention degree as the comprehensive potential attention degree of the target user to any official document.

20. The method according to claim 2, wherein when the comprehensive potential attention degree meets a preset recommendation condition, determining any document as a document to be recommended comprises:

arranging the plurality of documents according to the sequence of the comprehensive potential attention degrees from large to small;

and determining the documents in the preset number arranged in the front as documents to be recommended.

21. The method of claim 1, wherein the target user profile comprises: browsing, collecting or subscribing the official documents by the target user in a preset historical time period; and determining the official documents to be recommended in the plurality of official documents according to the incidence relation and the characteristic data of the target user, wherein the method comprises the following steps:

determining the official documents in the plurality of official documents browsed, collected or subscribed by the target user in a preset historical time period as target official documents;

and determining at least one official document having an association relation with the target official document as an official document to be recommended.

22. The method of claim 7,

the type of the incidence relation comprises: revising relationships, revoking relationships, basing relationships and mentioning relationships;

the types of historical behaviors include: browse, collect, subscribe, ignore, cancel collect, and cancel subscribe.

23. An official document recommendation device, comprising:

the information extraction unit is used for acquiring the incidence relation among a plurality of documents;

the official document determining unit is used for determining the official documents to be recommended in the multiple official documents according to the incidence relation and the characteristic data of the target user;

and the recommending unit is used for recommending the official document to be recommended to the target user.

24. An electronic device, comprising:

one or more processors;

a storage device for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-22.

25. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-22.

Images