CN114385902A

CN114385902A - Content recommendation method and device and storage medium

Info

Publication number: CN114385902A
Application number: CN202011139934.5A
Authority: CN
Inventors: 石磊
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-10-22
Filing date: 2020-10-22
Publication date: 2022-04-22
Anticipated expiration: 2040-10-22
Also published as: CN114385902B

Abstract

The invention discloses a content recommendation method, a content recommendation device and a storage medium; acquiring a content theme set of target content and a plurality of interactive content themes of a user aiming at the target content; extracting content subject characteristic information which corresponds to the content subject set together and interactive content characteristic information of each interactive content subject; determining a first similarity between the content theme set and each interactive content theme based on the content theme characteristic information and the interactive content characteristic information; determining a target interactive content theme from the interactive content themes based on the first similarity; the method comprises the steps of obtaining candidate recommended content subjects corresponding to a plurality of candidate recommended contents, determining a second similarity between the candidate recommended content subjects and a target interactive content subject so as to determine target recommended contents from the candidate recommended contents, recommending the target recommended contents, storing the target recommended contents in a cloud server, and facilitating subsequent information reading. The method and the device can improve the accuracy of content recommendation.

Description

Content recommendation method and device and storage medium

Technical Field

The present application relates to the field of communications technologies, and in particular, to a content recommendation method, apparatus, and storage medium.

Background

With the rapid development of information technology, related content similar to the current reading content may be recommended to the user based on the current reading content of the user, for example, the related content similar to the current reading content may be recommended to the user through a cloud server.

In the research and practice process of the related art, the inventors of the present application found that when determining related content related to current reading content of a user at present, the related content related to the current reading content is determined from a plurality of reading contents based on the content attribute of each reading content itself and the correlation comparison with the current reading content of the user, and then the related content is recommended, and the accuracy of content recommendation is low.

Disclosure of Invention

The embodiment of the application provides a content recommendation method, a content recommendation device and a storage medium, which can improve the accuracy of content recommendation.

The embodiment of the application provides a content recommendation method, which comprises the following steps:

acquiring a content subject set of target content and a plurality of interactive content subjects of a user aiming at the target content, wherein the content subject set comprises a plurality of content subjects of the target content;

extracting content subject characteristic information which corresponds to the content subject set together and interactive content characteristic information of each interactive content subject;

determining a first similarity between the content subject matter set and each interactive content subject matter based on the content subject matter feature information and the interactive content feature information;

determining a target interactive content theme from the interactive content themes based on the first similarity;

obtaining candidate recommended content subjects corresponding to a plurality of candidate recommended contents, and determining a second similarity between the candidate recommended content subjects and the target interactive content subject;

and determining target recommended content from the candidate recommended contents based on the second similarity, and recommending the target recommended content.

Correspondingly, an embodiment of the present application provides a content recommendation device, including:

the device comprises a first obtaining unit, a second obtaining unit and a third obtaining unit, wherein the first obtaining unit is used for obtaining a content subject set of target content and a plurality of interactive content subjects of a user aiming at the target content, and the content subject set comprises a plurality of content subjects of the target content;

the extraction unit is used for extracting content subject characteristic information which corresponds to the content subject sets together and interactive content characteristic information of each interactive content subject;

a similarity determining unit, configured to determine, based on the content topic feature information and the interactive content feature information, a first similarity between the content topic set and each interactive content topic;

the theme determining unit is used for determining a target interactive content theme from the interactive content themes on the basis of the first similarity;

the second obtaining unit is used for obtaining candidate recommended content topics corresponding to a plurality of candidate recommended contents and determining a second similarity between the candidate recommended content topics and the target interactive content topics;

and the content determining unit is used for determining target recommended content from the candidate recommended contents based on the second similarity and recommending the target recommended content.

In an embodiment, the first obtaining unit includes:

the first word segmentation subunit is used for carrying out word segmentation processing on the target content to obtain a target content word group;

a first theme determining subunit, configured to determine, based on distribution information of the target content phrase in the target content, a content theme of the target content from the target content phrase, so as to obtain a content theme set of the target content;

and the obtaining subunit is used for obtaining a plurality of interactive content themes of the user aiming at the target content.

In an embodiment, the obtaining subunit is further configured to obtain a plurality of interactive contents for the target content; performing word segmentation processing on the plurality of interactive contents to obtain interactive content word groups; and determining a plurality of interactive content subjects from the interactive content phrases based on the distribution information of the interactive content phrases in the corresponding interactive content.

In one embodiment, the extraction unit includes:

the first vectorization subunit is configured to vectorize the plurality of content topics in the content topic set to obtain a plurality of content topic word vectors;

a fusion subunit, configured to perform vector fusion on the multiple content topic word vectors to obtain content topic set word vectors corresponding to the content topic sets;

the first feature extraction subunit is used for performing feature extraction on the content subject set word vectors to obtain content subject feature information corresponding to the content subject sets;

the second directional quantization subunit is used for vectorizing each interactive content topic to obtain a plurality of interactive content topic word vectors;

and the second feature extraction subunit is used for performing feature extraction on the multiple interactive content subject word vectors to obtain the interactive content feature information of each interactive content subject.

In one embodiment, the theme determination unit includes:

the first sequencing subunit is used for sequencing the interactive content theme based on the first similarity;

and the theme determining subunit is used for determining a target interactive content theme from the interactive content themes on the basis of the sequencing result and a first preset selection rule.

In an embodiment, the second obtaining unit includes:

the second word segmentation subunit is used for carrying out word segmentation processing on the plurality of candidate recommended contents to obtain target candidate recommended content word groups;

a second theme determining subunit, configured to determine, based on distribution information of the target candidate recommended content word group in corresponding to-be-recommended content, at least one original candidate recommended content theme corresponding to the multiple candidate recommended contents;

a third theme determining subunit, configured to determine a candidate recommended content theme corresponding to the at least one original candidate recommended content;

and the calculating subunit is used for calculating a second similarity between the candidate recommended content subject and the target interactive content subject.

In one embodiment, the computing subunit is further configured to extract word vectors of the candidate recommended content topics and word vectors of the target interactive content topics; and calculating the vector similarity between the word vector of the to-be-recommended content theme and the word vector of the target interactive content theme to obtain a second similarity between the candidate recommended content theme and the target interactive content theme.

In one embodiment, the content determining unit includes:

a second ranking subunit, configured to rank the plurality of candidate recommended contents based on the second similarity;

and the content determining subunit is used for determining target candidate recommended content from the plurality of candidate recommended contents based on the sorting result and a second preset selection rule, and recommending the target recommended content.

Accordingly, embodiments of the present application further provide a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the steps in the content recommendation method provided in any of the embodiments of the present application.

Correspondingly, an embodiment of the present application further provides a storage medium, where the storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to perform steps in any of the content recommendation methods provided in the embodiments of the present application.

The method and the device for obtaining the target content can obtain a content theme set of the target content and a plurality of interactive content themes of a user aiming at the target content, wherein the content theme set comprises the plurality of content themes of the target content; extracting content subject characteristic information which corresponds to the content subject set together and interactive content characteristic information of each interactive content subject; determining a first similarity between the content subject matter set and each interactive content subject matter based on the content subject matter feature information and the interactive content feature information; determining a target interactive content theme from the interactive content themes based on the first similarity; obtaining candidate recommended content subjects corresponding to a plurality of candidate recommended contents, and determining a second similarity between the candidate recommended content subjects and the target interactive content subject; and determining target recommended content from the candidate recommended contents based on the second similarity, and recommending the target recommended content. According to the scheme, the interactive content theme most relevant to the target reading content theme can be determined according to the similarity between the content theme of the known target reading content and the multiple interactive content themes, then the target recommended content of the most relevant interactive theme is recalled from the multiple candidate recommended contents based on the most relevant interactive content theme, because the recalled target recommended content is the recommended content most relevant to the most relevant interactive content theme, the recalled target recommended content and the target reading content are also the most relevant recommended content, the diversity of the interactive content theme brought by the user side aiming at the target content is used, meanwhile, the target recommended content is determined by utilizing a recall mode based on the similarity, the target recommended content is recommended subsequently, and the accuracy of content recommendation can be improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a scene schematic diagram of a content recommendation method provided in an embodiment of the present application;

fig. 2a is a flowchart of a content recommendation method provided in an embodiment of the present application;

fig. 2b is a similarity determination flowchart of a content recommendation method according to an embodiment of the present application;

FIG. 2c is a flowchart of determining target recommended content in the content recommendation method according to an embodiment of the present application;

FIG. 3 is another flowchart of a content recommendation method provided by an embodiment of the present application;

fig. 4a is a device diagram of a content recommendation method provided in an embodiment of the present application;

fig. 4b is another apparatus diagram of a content recommendation method provided in an embodiment of the present application;

fig. 4c is another apparatus diagram of a content recommendation method provided in an embodiment of the present application;

fig. 4d is another apparatus diagram of a content recommendation method provided in an embodiment of the present application;

fig. 4e is another apparatus diagram of a content recommendation method provided in an embodiment of the present application;

fig. 4f is another apparatus diagram of a content recommendation method provided in an embodiment of the present application;

fig. 5 is a schematic structural diagram of a computer device provided in an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The embodiment of the application provides a content recommendation method and device, computer equipment and a storage medium. Specifically, the embodiment of the application provides a content recommendation device suitable for computer equipment. The computer device may be a terminal or a server, the server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a network service, cloud communication, middleware service, a domain name service, a security service, a CDN, a big data and artificial intelligence platform, and the like. The terminal may be, but is not limited to, a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and the like. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the application is not limited herein.

Referring to fig. 1, taking a computer device as a cloud server as an example, the cloud server may obtain a content topic set of target content and a plurality of interactive content topics of a user for the target content, where the content topic set includes a plurality of content topics of the target content; extracting content subject characteristic information which corresponds to the content subject set together and interactive content characteristic information of each interactive content subject; determining a first similarity between the content theme set and each interactive content theme based on the content theme characteristic information and the interactive content characteristic information; determining a target interactive content theme from the interactive content themes based on the first similarity; obtaining candidate recommended content subjects corresponding to the candidate recommended contents, and determining a second similarity between the candidate recommended content subjects and the target interactive content subject; and determining target recommended content from the candidate recommended contents based on the second similarity, and recommending the target recommended content.

The content theme set of the target content and a plurality of interactive content themes of the user aiming at the target content can be obtained based on a cloud platform technology, and the cloud platform is also called a cloud computing platform, is a service based on hardware resources and software resources and provides computing, network and storage capabilities. The cloud platform in this embodiment may be understood as a platform formed based on the cloud server, where the platform may provide various required services for a user, for example, provide company website building and operation services for the user, and the like, the user may purchase cloud server resources, for example, storage resources, computing resources, and the like, on the cloud platform, and the user may complete various tasks based on the purchased resources and various components provided by the cloud server.

Cloud computing (cloud computing) is a computing model that distributes computing tasks over a pool of resources formed by a large number of computers, enabling various application systems to obtain computing power, storage space, and information services as needed. The network that provides the resources is referred to as the "cloud". Resources in the "cloud" appear to the user as being infinitely expandable and available at any time, available on demand, expandable at any time, and paid for on-demand.

As can be seen from the above, in the embodiment of the application, the interactive content topic most relevant to the target reading content topic can be determined through the similarity between the content topic of the known target reading content and the multiple interactive content topics, and then the target recommended content most relevant to the interactive topic is recalled from the multiple candidate recommended contents based on the most relevant interactive content topic.

The present embodiment can be described in detail below, and it should be noted that the following description of the embodiment is not intended to limit the preferred order of the embodiment.

The embodiment of the application provides a content recommendation method, which can be executed by a terminal or a server, or can be executed by the terminal and the server together; the embodiment of the present application is described by taking an example in which the content recommendation method is executed by a server, and specifically, is executed by a content recommendation apparatus integrated in the server. As shown in fig. 2a, a specific flow of the content recommendation method may be as follows:

201. the method comprises the steps of obtaining a content subject set of target content and a plurality of interactive content subjects of a user aiming at the target content, wherein the content subject set comprises a plurality of content subjects of the target content.

The content theme refers to a theme of the target content, the content theme can represent content semantics of the target content, a plurality of interactive content themes of the target content can be obtained by theme extraction of a plurality of interactive contents of the target interactive content, and similarly, the interactive content theme can represent content semantics of the interactive content.

In an example, a target content is taken as an article for explanation, an obtained content topic set is a topic set of the article, the topic set may include a plurality of article topics of the article, and a plurality of interactive content topics of a user for the target content may be comment topics corresponding to a plurality of comments of the user for the article.

In one embodiment, the step of obtaining a content topic set of the target content and a plurality of interactive content topics of the target content by the user may include:

performing word segmentation processing on the target content to obtain a target content word group;

determining a content subject of the target content from the target content phrase based on the distribution information of the target content phrase in the target content to obtain a content subject set of the target content;

and acquiring a plurality of interactive content themes of the user aiming at the target content.

The distribution information may include distribution rule information, for example, a distribution rule of the target content phrase in the target content, such as information about a distribution probability of the target content phrase in the target content, a distribution rule of the interactive content phrase in the corresponding interactive content, such as information about a distribution probability of the interactive content phrase in the corresponding interactive content, and the distribution information may further include other information.

In one embodiment, the step of "obtaining a plurality of interactive content topics of the user for the target content" may include:

acquiring a plurality of interactive contents of a user aiming at target contents;

performing word segmentation processing on the plurality of interactive contents to obtain interactive content phrases;

and determining a plurality of interactive content topics from the interactive content phrases based on the distribution information of the interactive content phrases in the corresponding interactive content.

202. And extracting content subject characteristic information which corresponds to the content subject set together and interactive content characteristic information of each interactive content subject.

The content subject characteristic information is characteristic information representing semantic characteristics of the target content, and the interactive content characteristic information is characteristic information representing semantic characteristics of the corresponding interactive content.

In an embodiment, the step of "extracting content topic feature information corresponding to the content topic sets in common, and interactive content feature information of each interactive content topic" may include:

vectorizing a plurality of content topics in a content topic set to obtain a plurality of content topic word vectors;

vector fusion is carried out on the content subject word vectors to obtain content subject set word vectors which correspond to the content subject sets;

extracting the characteristics of the word vectors of the content theme sets to obtain content theme characteristic information corresponding to the content theme sets;

vectorizing each interactive content theme to obtain a plurality of interactive content theme word vectors;

and performing feature extraction on the multiple interactive content subject word vectors to obtain the interactive content feature information of each interactive content subject.

In an example, the content topic set may include a content topic1, a content topic 2, a content topic 3, and a content topic4, respectively extract word vectors corresponding to the content topics, may obtain a content topic word vector 1, a content topic word vector 2, a content topic word vector 3, and a content topic word vector 4, perform vector fusion on each content topic word vector, may perform normalization by accumulating the content topic word vector 1, the content topic word vector 2, the content topic word vector 3, and the content topic word vector 4, and may then obtain topic exemplar (word vector) of the target content, that is, may obtain a content topic set word vector corresponding to the content topic set.

203. And determining a first similarity between the content theme set and each interactive content theme based on the content theme characteristic information and the interactive content characteristic information.

The similarity between the content subject set and each interactive content subject is characterized by the first similarity, and can be determined through content subject characteristic information and interactive content characteristic information, or can be determined according to cosine similarity between content subject word vectors and each interactive content word vector which are commonly corresponding to the content subject set, because the word vectors are vectors which convert words in natural language into dense vectors, and words with similar semantics can be represented by similar vectors, the similarity between the content subject set and each interactive content subject can be determined through calculating the cosine similarity between the word vectors and each interactive content word vector which are commonly corresponding to the content subject set, and can also be determined through other modes.

In an embodiment, as shown in fig. 2b, the first similarity between the content topic set and each interactive content topic may also be determined by extracting the content topic embedding corresponding to the content topic set, the interactive content topic embedding corresponding to each interactive content topic, and then calculating the cosine similarity therebetween. Assume that the topic set of the target content is aTopic [ content topic1, content topic 2, content topic 3, … ], and the corresponding interactive content topic takes a top1 topic, i.e., comment1 (interactive content 1): interactive content topic1, comment2 (interactive content 2): interactive content topic 2, etc., where the set of interactive content topics [ interactive content topic1, interactive content topic 2, interactive content topic 3, … ], is taken, each element inside represents a topic corresponding to the interactive content. The way of extracting the theme may be through LDA algorithm. Then, using word2vec to carry out embedding on a topic (topic) set (namely [ content topic1, content topic 2, content topic 3, … ]) corresponding to the target content, so as to form an embedding vector corresponding to the topic of the article; the specific mode is that word2vec corresponding to each topic is accumulated and normalized to form an embedding vector of 100 dimensions. Then, finding an embedding vector corresponding to each interactive content theme in the same way, and then calculating a first similarity between the content theme set and each interactive content theme by using a cos (cosine similarity) similarity calculation way.

In one embodiment, at least 100w sets of historical content in the content library can be utilized to train a 100-dimensional version of word2vec word set as a dependency of embedding; the content subject set is composed of four content subject keywords, namely topic1 to topic4, each content subject keyword can find a corresponding topic vector (word vector) in the word2vec as embedding of the topic, then the embedding corresponding to the four topic vectors is accumulated and normalized, and the topic embedding of the content, namely the attribute embedding, is obtained. Similarly, the interactive content subject of each piece of interactive content can also find the corresponding interactive content subject embedding in the word2 vec. The similarity can then be calculated by cos: cos < angle Embedding > and interactive content theme Embedding >, and then arranging backwards, the most relevant interactive content theme Embedding can be obtained.

204. And determining a target interactive content theme from the interactive content themes based on the first similarity.

In an embodiment, after extracting the embedding vector of the content theme with the theme set aTopic corresponding to [ content theme 1, content theme 2, content theme 3, … ] together, and extracting the embedding vector corresponding to each interactive content theme, the most similar interactive content theme corresponding to the target content may be calculated in a cos (cosine similarity) similarity calculation manner, and the top 3 may be taken out to obtain the embedding vector of the interactive content of the top 3, that is, the target interactive content theme may be obtained, and the embedding vector of the interactive content of the top4 may also be taken out to serve as the target interactive content theme, and the range of values may be configured according to actual conditions, and will not be described herein again.

In one embodiment, the step "determining a target interactive content topic from the interactive content topics based on the first similarity" may include:

based on the first similarity, sequencing the interactive content topics;

and determining a target interactive content theme from the interactive content themes based on the sequencing result and the first preset selection rule.

The first predetermined selection rule is a rule for determining a target interactive content theme, for example, the first predetermined selection rule may include a preset number of themes of the target interactive content theme, and based on the sorting result and the preset number of themes, the target interactive content theme may be determined from the interactive content themes, and if the preset number of themes of the target interactive content theme is 3, sorting according to the first similarity may be selected from the interactive content themes, and the 3 interactive content themes arranged in front are the target interactive content theme.

In one example, on the information flow recommendation side, a high-exposure high-click article, or an article with a C-side posteriori comment, we would like to get more related articles with similar topics. The corresponding comment content styles of the articles are different, and the themes are various; for example: an article describing the european football, corresponding to the review board, may have these several contents: "our bayer munich is champion"; "how often Chinese football can have this level …"; "the person is on the building and has a tear on the floor"; "how the world cup is never to feel bored". When all the comment data in the article are utilized, the topic mining of the article is more diversified and three-dimensional. Through a comment of an article teaching a european football, subjects like "beneyeMunich", "chinese football" can be mined and no additional dependency is required; similar to comment 3, the comments are invalid, no available topics are mined to help recall the relevant articles, similarity calculation can be performed between the article topics and the comment topics, and then the invalid comments are filtered based on the similarity.

205. And obtaining candidate recommended content subjects corresponding to the candidate recommended contents, and determining a second similarity between the candidate recommended content subjects and the target interactive content subject.

The second similarity between the candidate recommended content theme and the target interactive content theme can be determined through candidate content theme characteristic information corresponding to each candidate content theme and target interactive content characteristic information of the target interactive content theme, can also be determined according to vector similarity between word vectors corresponding to each content theme to be recommended and word vectors of the target interactive content theme, and can also be determined in other manners.

In an embodiment, the step of "obtaining candidate recommended content topics corresponding to a plurality of candidate recommended content and determining a second similarity between the candidate recommended content topics and the target interactive content topic" may include:

performing word segmentation processing on the candidate recommended contents to obtain target candidate recommended content word groups;

determining at least one original candidate recommended content subject corresponding to a plurality of candidate recommended contents based on the distribution information of the target candidate recommended content word group in the corresponding to-be-recommended content;

determining a candidate recommended content subject corresponding to at least one original candidate recommended content;

and calculating a second similarity between the candidate recommended content subject and the target interactive content subject.

In an example, a candidate recommended content subject of each candidate recommended content may be obtained, then candidate content subject feature information of the candidate recommended content subject and target interactive content feature information of the target interactive content subject are extracted, and finally, a second similarity between the candidate recommended content subject and the target interactive content subject is determined based on the candidate content subject feature information and the target interactive content feature information.

In one embodiment, the step of "calculating a second similarity between the candidate recommended content subject and the target interactive content subject" may include:

extracting word vectors of candidate recommended content subjects and word vectors of target interactive content subjects;

and calculating the vector similarity between the word vector of the subject of the content to be recommended and the word vector of the subject of the target interactive content to obtain a second similarity between the candidate recommended content subject and the target interactive content subject.

In an embodiment, as shown in fig. 2c, the target interactive content theme which is obtained by the above process and is most similar to the content theme of the target content may be used to perform correlation calculation on the interactive content feature information (for example, the interactive content feature information 1, the interactive content feature information 2, the interactive content feature information 3, and the like, the number of the interactive content feature information may be specifically determined according to the number of the target interactive content theme) of the target interactive content theme and the candidate content theme feature information of each candidate recommended content in the content library, where the specific way is to perform the same operation mode as the target content on the candidate recommended content respectively: and mining the content subject of the candidate content and embedding the candidate recommended content subject to obtain the candidate content subject embedding. And then cos similarity calculation can be carried out on the imbedding and the candidate content theme imbedding of the target interactive content theme, top N candidate recommended content is obtained to serve as a final related article, namely the obtained top N candidate recommended content is determined to be target recommended content and serves as a recommendation strategy of a related content plate. And finally, recommending the target recommended content, for example, displaying summary information of the target recommended content on a page where the user reads the target content to recommend the target recommended content.

206. And determining target recommended content from the candidate recommended contents based on the second similarity, and recommending the target recommended content.

The target recommended content is a content which is most relevant to the target content, such as content semantics being most relevant, among the plurality of candidate recommended contents.

In an embodiment, the step of determining a target recommended content from a plurality of candidate recommended contents based on the second similarity, and recommending the target recommended content, may include:

ranking the plurality of candidate recommended contents based on the second similarity;

and determining target candidate recommended content from the plurality of candidate recommended contents based on the sorting result and a second preset selection rule, and recommending the target recommended content.

The second predetermined selection rule is a rule for determining the target recommended content, for example, the second predetermined selection rule may include a preset number of contents of the target recommended content, the target recommended content may be determined from a plurality of candidate recommended contents based on the sorting result and the preset number of contents, if the preset number of contents is 4, the target recommended content may be sorted according to the second similarity from the plurality of candidate recommended contents, and the top4 candidate recommended contents are determined as the target recommended content.

In an embodiment, the scheme of the application can provide an information flow recommendation side related article recall strategy based on posterior comment data topic mining, and the embodiment of active feedback information of a user in comments is used as expansion of related article recall plates on a business level, and then the recalled related articles are recommended. The main diversity of the related recommended plates is increased, and the user experience is further improved.

According to the method and the device, under the scene of information flow graph text recommendation, the interaction content of the user aiming at the target content, such as comment data of the user aiming at the target content, is utilized to assist in recalling and recommending the related candidate recommended content. Different from the traditional article inherent attribute feature-based related recall, the method and the device for recalling other related candidate recommended contents are characterized in that the theme characteristics of the interactive contents corresponding to the target contents, the correlation between the interactive contents theme and the target contents theme and the correlation between the interactive contents theme and the other candidate recommended contents theme are utilized to recall other related candidate recommended contents, and the content recommendation efficiency and the posterior index of the information stream recommendation side are improved.

Based on the above description, the content recommendation method of the present application will be further described below by way of example. Referring to fig. 3, a content recommendation method may include the following specific processes:

301. the server obtains a plurality of content topics of the target content and a plurality of interactive content topics of the user aiming at the target content.

In an example, the topic mining model may be used to mine a plurality of content topics of the target content and a plurality of interactive content topics of a plurality of interactive contents for the target content, for example, the topic mining model may be used to mine a topic of the target content and a plurality of interactive contents through LDA (document topic Allocation), and the topic mining model may also be used to mine a topic of the target content and a plurality of interactive contents through other manners.

The LDA is also called a three-layer Bayes probability model and comprises three layers of structures of words, topics and documents, the generative model is obtained by a process that every word of an article selects a certain topic with a certain probability, and selects a certain word from the topic with a certain probability, the document topics are distributed according to a polynomial, and the topics are distributed according to the polynomial. LDA is an unsupervised machine learning technique that can be used to identify underlying topic information in large-scale document collections (document collections) or corpora (corpus).

302. The server vectorizes the plurality of content topics and the plurality of interactive content topics to obtain a plurality of content subject term vectors and a plurality of interactive content subject term vectors.

In an embodiment, a plurality of content topics and a plurality of interactive content topics corresponding to a target content may be subjected to embedding by the word2vec, so as to form embedding vectors corresponding to the plurality of content topics, that is, a plurality of content subject word vectors, and embedding vectors corresponding to the plurality of content topics, that is, a plurality of interactive content subject word vectors, respectively, and the plurality of content topics and the plurality of interactive content topics may also be subjected to vectorization in other manners, for example, the plurality of content topics and the plurality of interactive content topics corresponding to the target content may also be subjected to vectorization by ELMo (embedded from Language Models), BERT (Bidirectional encoding retrieval from transformations, word vector Models), and the like.

303. The server fuses the content subject word vectors to obtain a content subject word set vector corresponding to the content subjects.

In an example, the merging may be understood as accumulating and normalizing a plurality of content subject word vectors to obtain a content subject set word vector corresponding to a plurality of content subjects in common, and the vector accumulation is to accumulate a plurality of word vectors into one word vector, for example, a first digit of a first word vector may be added to a first digit of a second word vector, and then a first digit of a third word vector may be added to the first digit, and so on, and then a content subject set word vector corresponding to a plurality of content subjects in common may be obtained.

304. The server extracts content subject characteristic information of the content subject set word vectors and interactive content characteristic information of each interactive content subject word vector.

The content topic feature information and the interactive content feature information may be used to determine a first similarity between the content topic set and each interactive content topic, that is, the content topic feature information and the interactive content feature information may be used to determine a correlation between the interactive content and the target content, that is, the similarity between the content topic set and each interactive content may be calculated based on the content topic feature information and the interactive content feature information, and the correlation between each interactive content and the target content may be determined by the size of the similarity, for example, the similarity is large, the correlation between the corresponding interactive content and the target content is large, and the interactive content is more related to the target content.

305. The server determines a first similarity between a content subject set including a plurality of content subjects and each interactive content subject based on the content subject characteristic information and the interactive content characteristic information.

In an embodiment, the similarity between the content topic set and each interactive content topic may be determined through the content topic word vector and the interactive content feature information, and the first similarity between the content topic set including a plurality of content topics and each interactive content topic may also be determined by calculating the vector similarity between the content topic set word vector and the interactive content word vector corresponding to each interactive content topic.

306. The server determines a target interactive content theme from the interactive content themes based on the first similarity.

In an example, the plurality of interactive content topics may be sorted based on the size of the first similarity between the content topic set and each interactive content topic, and then the top several interactive content topics are determined as the target interactive content topics based on actual requirements.

The relevance between the interactive content corresponding to the target interactive content theme and the target content is the highest, and the interactive content corresponding to the target interactive content theme is semantically more similar to the target content.

307. The server obtains candidate recommended content subjects corresponding to the candidate recommended contents, and determines a second similarity between the candidate recommended content subjects and the target interactive content subject.

The second similarity between the candidate recommended-content subject and the target interactive-content subject is calculated to determine the recommended content most relevant to the target content from the plurality of candidate recommended contents, i.e. the target recommended content in step 308.

In an embodiment, after the recommended content most relevant to the target content is determined from the plurality of candidate recommended contents, the recommended content can be recommended based on the target content, and the method can also be used for recalling relevant articles from a content library and a large number of articles.

308. And the server determines target recommended content from the candidate recommended contents based on the second similarity and recommends the target recommended content.

In an embodiment, a combined recall mode based on interactive content theme mining, target content theme mining and similarity calculation strategies uses the theme diversity actively brought by a user at a user side aiming at target content, and simultaneously uses an unsupervised recall mode based on cos similarity to increase the recall fuzziness, further enhance the diversity of related content plates at a recommendation side, ensure the consistency and consistency of the related content recommendation plates, and improve the user experience at the recommendation side.

In the content recommendation process, the diversity of theme judgment of the target content can be increased by increasing the characteristics of the posterior interactive content, such as the characteristics of the posterior comment data, so that the themes contained in the related content recommendation block are richer and more varied in the information flow recommendation.

In order to better implement the above method, correspondingly, the embodiment of the present application further provides a content recommendation device, where the content recommendation device may be specifically integrated in a server, and referring to fig. 4a, the content recommendation device may include a first obtaining unit 401, an extracting unit 402, a similarity determining unit 403, a theme determining unit 404, a second obtaining unit 405, and a content determining unit 406, as follows:

(1) a first acquisition unit 401;

the first obtaining unit 401 is configured to obtain a content topic set of the target content and a plurality of interactive content topics of the target content, where the content topic set includes a plurality of content topics of the target content.

In an embodiment, as shown in fig. 4b, the first obtaining unit 401 includes:

the first word segmentation sub-unit 4011 is configured to perform word segmentation processing on the target content to obtain a target content word group;

the first topic determining sub-unit 4012 is configured to determine, based on distribution information of the target content phrases in the target content, content topics of the target content from the target content phrases, and obtain a content topic set of the target content;

the obtaining sub-unit 4013 is configured to obtain a plurality of interactive content topics for the target content.

In an embodiment, the obtaining sub-unit 4013 is further configured to obtain a plurality of interactive contents for the target content; performing word segmentation processing on the plurality of interactive contents to obtain interactive content phrases; and determining a plurality of interactive content topics from the interactive content phrases based on the distribution information of the interactive content phrases in the corresponding interactive content.

(2) An extraction unit 402;

the extracting unit 402 is configured to extract content theme feature information corresponding to the content theme sets and interactive content feature information of each interactive content theme.

In one embodiment, as shown in fig. 4c, the extracting unit 402 includes:

the first vectorization subunit 4021 is configured to vector a plurality of content topics in a content topic set to obtain a plurality of content topic word vectors;

the fusion subunit 4022 is configured to perform vector fusion on the multiple content topic word vectors to obtain content topic set word vectors corresponding to the content topic sets;

the first feature extraction subunit 4023 is configured to perform feature extraction on the content topic set word vectors to obtain content topic feature information corresponding to the content topic sets;

the second directional quantization subunit 4024 is configured to perform vectorization on each interactive content topic to obtain a plurality of interactive content topic word vectors;

the second feature extraction subunit 4025 is configured to perform feature extraction on the multiple interactive content topic word vectors to obtain interactive content feature information of each interactive content topic.

(3) A similarity determination unit 403;

a similarity determining unit 403, configured to determine a first similarity between the content topic set and each interactive content topic based on the content topic feature information and the interactive content feature information.

(4) A topic determination unit 404;

a topic determining unit 404, configured to determine a target interactive content topic from the interactive content topics based on the first similarity.

In one embodiment, as shown in fig. 4d, the topic determination unit 404 includes:

a first ordering subunit 4041, configured to order the interactive content topics based on the first similarity;

the topic determination subunit 4042 is configured to determine a target interactive content topic from the interactive content topics based on the sorting result and the first predetermined selection rule.

(5) A second acquisition unit 405;

the second obtaining unit 405 is configured to obtain candidate recommended content topics corresponding to the multiple candidate recommended contents, and determine a second similarity between the candidate recommended content topics and the target interactive content topic.

In an embodiment, as shown in fig. 4e, the second obtaining unit 405 includes:

a second word segmentation subunit 4051, configured to perform word segmentation processing on the multiple candidate recommended contents to obtain a target candidate recommended content word group;

a second topic determination subunit 4052, configured to determine, based on distribution information of the target candidate recommended content word group in the corresponding content to be recommended, at least one original candidate recommended content topic corresponding to the multiple candidate recommended contents;

a third topic determination subunit 4053, configured to determine a candidate recommended content topic that at least one original candidate recommended content corresponds to in common;

the computing subunit 4054 is configured to calculate a second similarity between the candidate recommended content topic and the target interactive content topic.

In an embodiment, the calculating subunit 4054 is further configured to extract word vectors of the candidate recommended content topics and word vectors of the target interactive content topics; and calculating the vector similarity between the word vector of the subject of the content to be recommended and the word vector of the subject of the target interactive content to obtain a second similarity between the candidate recommended content subject and the target interactive content subject.

(6) A content determination unit 406;

a content determining unit 406, configured to determine a target recommended content from the multiple candidate recommended contents based on the second similarity, and recommend the target recommended content.

In one embodiment, as shown in fig. 4f, the content determining unit 406 includes:

a second sorting subunit 4061, configured to sort, based on the second similarity, the plurality of candidate recommended contents;

the content determining sub-unit 4062 is configured to determine a target recommended content candidate from the plurality of recommended content candidates based on the sorting result and a second predetermined selection rule, and recommend the target recommended content.

As can be seen from the above, the first obtaining unit 401 of the content recommendation device in the embodiment of the present application obtains the content topic set of the target content and the plurality of interactive content topics of the user for the target content, where the content topic set includes the plurality of content topics of the target content; then, the extracting unit 402 extracts the content subject feature information corresponding to the content subject sets together and the interactive content feature information of each interactive content subject; determining, by the similarity determining unit 403, a first similarity between the content topic set and each interactive content topic based on the content topic feature information and the interactive content feature information; determining, by the topic determination unit 404, a target interactive content topic from the interactive content topics based on the first similarity; a second obtaining unit 405 obtains candidate recommended content topics corresponding to a plurality of candidate recommended contents, and determines a second similarity between the candidate recommended content topics and a target interactive content topic; the content determining unit 406 determines a target recommended content from the plurality of candidate recommended contents based on the second similarity, and recommends the target recommended content. According to the scheme, the interactive content theme most relevant to the target reading content theme can be determined through the similarity between the content theme of the known target reading content and the multiple interactive content themes, then the target recommended content of the most relevant interactive theme is recalled from the multiple candidate recommended contents based on the most relevant interactive content theme, because the recalled target recommended content is the recommended content most relevant to the most relevant interactive content theme, the recalled target recommended content and the target reading content are also the most relevant recommended content, the diversity of the interactive content theme brought by a user side aiming at the target content is used, meanwhile, the target recommended content is determined by utilizing a recall mode based on the similarity, the target recommended content is recommended subsequently, and the accuracy of content recommendation can be improved.

In addition, an embodiment of the present application further provides a computer device, where the computer device may be a device such as a terminal or a server, and as shown in fig. 5, a schematic structural diagram of the computer device according to the embodiment of the present application is shown, specifically:

the computer device may include components such as a processor 501 of one or more processing cores, memory 502 of one or more storage media, a power supply 503, and an input unit 504. Those skilled in the art will appreciate that the computer device configuration illustrated in FIG. 5 does not constitute a limitation of computer devices, and may include more or fewer components than those illustrated, or some components may be combined, or a different arrangement of components. Wherein:

the processor 501 is a control center of the computer device, connects various parts of the entire computer device by using various interfaces and lines, and performs various functions of the computer device and processes data by running or executing software programs and/or modules stored in the memory 502 and calling data stored in the memory 502, thereby monitoring the computer device as a whole. Optionally, processor 501 may include one or more processing cores; preferably, the processor 501 may integrate an application processor, which mainly handles operating systems, user interfaces, application programs, etc., and a modem processor, which mainly handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 501.

The memory 502 may be used to store software programs and modules, and the processor 501 executes various functional applications and data processing by operating the software programs and modules stored in the memory 502. The memory 502 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to use of the computer device, and the like. Further, the memory 502 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 502 may also include a memory controller to provide the processor 501 with access to the memory 502.

The computer device further comprises a power supply 503 for supplying power to the various components, and preferably, the power supply 503 may be logically connected to the processor 501 through a power management system, so that functions of managing charging, discharging, power consumption, and the like are realized through the power management system. The power supply 503 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.

The computer device may also include an input unit 504, and the input unit 504 may be used to receive input numeric or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

Although not shown, the computer device may further include a display unit and the like, which are not described in detail herein. Specifically, in this embodiment, the processor 501 in the computer device loads the executable file corresponding to the process of one or more application programs into the memory 502 according to the following instructions, and the processor 501 runs the application programs stored in the memory 502, so as to implement various functions as follows:

acquiring a content subject set of target content and a plurality of interactive content subjects of a user aiming at the target content, wherein the content subject set comprises a plurality of content subjects of the target content; extracting content subject characteristic information which corresponds to the content subject set together and interactive content characteristic information of each interactive content subject; determining a first similarity between the content theme set and each interactive content theme based on the content theme characteristic information and the interactive content characteristic information; determining a target interactive content theme from the interactive content themes based on the first similarity; obtaining candidate recommended content subjects corresponding to the candidate recommended contents, and determining a second similarity between the candidate recommended content subjects and the target interactive content subject; and determining target recommended content from the candidate recommended contents based on the second similarity, and recommending the target recommended content.

It will be understood by those skilled in the art that all or part of the steps in the methods of the above embodiments may be performed by instructions or by instructions controlling associated hardware, and the instructions may be stored in a storage medium and loaded and executed by a processor.

To this end, the present application provides a storage medium, in which a plurality of instructions are stored, where the instructions can be loaded by a processor to execute the steps in any one of the content recommendation methods provided in the present application. For example, the instructions may perform the steps of:

Wherein the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.

Since the instructions stored in the storage medium may execute the steps in any content recommendation method provided in the embodiments of the present application, beneficial effects that can be achieved by any content recommendation method provided in the embodiments of the present application may be achieved, which are detailed in the foregoing embodiments and will not be described herein again.

According to an aspect of the application, there is provided, among other things, a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the content recommendation method provided in the above-described contents and embodiments of the invention.

The content recommendation method, device, computer device and storage medium provided by the embodiments of the present application are described in detail above, and a specific example is applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiments is only used to help understand the method and the core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A content recommendation method, comprising:

2. The method of claim 1, wherein the obtaining of the content subject set of the target content and the plurality of interactive content subjects of the user for the target content comprises:

3. The method of claim 2, wherein the obtaining of the plurality of interactive content topics of the user for the target content comprises:

acquiring a plurality of interactive contents of a user aiming at the target content;

performing word segmentation processing on the plurality of interactive contents to obtain interactive content word groups;

and determining a plurality of interactive content subjects from the interactive content phrases based on the distribution information of the interactive content phrases in the corresponding interactive content.

4. The method according to claim 1, wherein the extracting of the content subject feature information corresponding to the content subject set and the interactive content feature information of each interactive content subject includes:

vectorizing a plurality of content topics in the content topic set to obtain a plurality of content topic word vectors;

vector fusion is carried out on the content subject term vectors to obtain content subject term vectors which correspond to the content subject sets;

performing feature extraction on the content subject set word vectors to obtain content subject feature information corresponding to the content subject sets;

and performing feature extraction on the plurality of interactive content subject word vectors to obtain interactive content feature information of each interactive content subject.

5. The method of claim 1, wherein the determining a target interactive content topic from the interactive content topics based on the first similarity comprises:

based on the first similarity, sequencing the interactive content theme;

and determining a target interactive content theme from the interactive content themes based on the sequencing result and a first preset selection rule.

6. The method of claim 1, wherein the obtaining candidate recommended-content topics corresponding to a plurality of candidate recommended-content and determining a second similarity between the candidate recommended-content topics and the target interactive-content topic comprises:

determining at least one original candidate recommended content subject corresponding to the candidate recommended contents based on the distribution information of the target candidate recommended content word group in the corresponding to-be-recommended content;

determining a candidate recommended content subject corresponding to the at least one original candidate recommended content;

calculating a second similarity between the candidate recommended content subject and the target interactive content subject.

7. The method of claim 6, wherein the calculating the second similarity between the candidate recommended content subject and the target interactive content subject comprises:

extracting word vectors of the candidate recommended content subjects and word vectors of the target interactive content subjects;

and calculating the vector similarity between the word vector of the to-be-recommended content theme and the word vector of the target interactive content theme to obtain a second similarity between the candidate recommended content theme and the target interactive content theme.

8. The method of claim 1, wherein the determining the target recommended content from the plurality of candidate recommended contents based on the second similarity includes:

9. A content recommendation apparatus characterized by comprising:

10. A storage medium storing instructions adapted to be loaded by a processor to perform the steps of the content recommendation method according to any one of claims 1 to 8.