CN114707068A

CN114707068A - Method, device, equipment and medium for recommending intelligence base knowledge

Info

Publication number: CN114707068A
Application number: CN202210368817.9A
Authority: CN
Inventors: 孙显; 郎公福; 李树超; 李晓宇; 金力; 马玉辉
Original assignee: Aerospace Information Research Institute of CAS
Current assignee: Aerospace Information Research Institute of CAS
Priority date: 2022-04-08
Filing date: 2022-04-08
Publication date: 2022-07-05

Abstract

The disclosure provides a method for recommending wisdom knowledge, which comprises the following steps: acquiring target user data, wherein the target user data comprises: target user attributes and target user behavior; tracking the wisdom base knowledge according to the target user data to obtain the target wisdom base knowledge, wherein the wisdom base knowledge comprises at least one of the following: text, video, sound; processing the target wisdom knowledge by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom knowledge; and inputting the initial recommendation result into a trained recommendation model to obtain a final recommendation result, wherein the final recommendation result comprises a plurality of finally recommended wisdom knowledge. The present disclosure also provides an intellectual knowledge recommendation apparatus, device, storage medium and program product.

Description

Method, device, equipment and medium for recommending knowledge of intellectual property library

Technical Field

The present disclosure relates to the field of intelligence repository knowledge recommendation, and in particular, to a method, an apparatus, a device, a storage medium, and a program product for intelligence repository knowledge recommendation.

Background

The intelligence base provides a great deal of valuable information for the public, but the knowledge quantity in the intelligence base is huge, the variety is various, and how to quickly track the knowledge in which the user is interested from the massive intelligence base knowledge becomes a problem to be solved urgently.

The traditional wisdom base knowledge tracking technology mostly adopts a single tracking mode mainly based on 'searching', and along with the richness of wisdom base knowledge, users have difficulty in expressing own requirements by using a plurality of search terms. Under the background, how to draw user preferences in multiple dimensions and quickly and accurately track the intellectual property knowledge in which the user is interested from massive intellectual property knowledge becomes a problem to be solved urgently.

Disclosure of Invention

In view of the foregoing, the present disclosure provides a method, apparatus, device, storage medium, and program product for intellectual knowledge recommendation.

According to a first aspect of the present disclosure, there is provided a method of intellectual knowledge recommendation, comprising:

acquiring target user data, wherein the target user data comprises: target user attributes and target user behavior; tracking the wisdom base knowledge according to the target user data to obtain the target wisdom base knowledge, wherein the wisdom base knowledge comprises at least one of the following: text, video, sound; processing the target wisdom knowledge by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom knowledge; and inputting the initial recommendation result into a trained recommendation model to obtain a final recommendation result, wherein the final recommendation result comprises a plurality of finally recommended wisdom knowledge.

According to an embodiment of the present disclosure, the recommendation model is trained as follows:

obtaining sample user data, wherein the sample user data comprises: sample user attributes and sample user behaviors; tracking the intellectual property base knowledge according to the sample user data to obtain an intellectual property base knowledge sample; processing the intellectual property samples by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended intellectual property samples; taking the sample user data and the initial recommendation result as samples in a sample pool, carrying out classification prediction on the samples in the sample pool by using a classifier to obtain marked samples, and adding the marked samples into a training set to construct a training sample set; and training the untrained recommendation model by using the training sample set to obtain the trained recommendation model.

According to the embodiment of the disclosure, the trained recommendation model is evaluated to obtain an evaluation result.

According to the embodiment of the disclosure, the individual recommendation algorithm is utilized to process the wisdom knowledge samples to obtain the initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom knowledge samples including:

processing the intellectual property base knowledge sample based on a heat individual recommendation algorithm to obtain an intellectual property base knowledge sample based on heat initial recommendation; processing the intellectual property base knowledge sample based on a collaborative filtering personalized recommendation algorithm to obtain an intellectual property base knowledge sample based on collaborative filtering initial recommendation; processing the intellectual property sample based on a matrix decomposition individual recommendation algorithm to obtain an intellectual property sample based on matrix decomposition initial recommendation; and processing the intellectual property base knowledge sample based on the content individual recommendation algorithm to obtain the intellectual property base knowledge sample based on the content initial recommendation.

According to an embodiment of the present disclosure, constructing a training sample set includes:

setting a change rate threshold value of sample prediction class targets in a sample pool as mu; calculating the uncertainty of each sample class in the sample pool when the change rate of the classifiers adjacent twice to the prediction class labels of the samples in the sample pool is greater than a threshold value; marking the sample with the maximum uncertainty in the sample pool to obtain a marked sample; removing the marked samples from the sample pool, and adding the samples into a training sample set to construct a training sample set; or

And when the change rate of the classifiers in two adjacent times to the prediction class labels of the samples in the sample pool is smaller than a threshold value, completing the construction of the training sample set.

A second aspect of the present disclosure provides a wisdom knowledge recommendation device, including:

the first obtaining module is used for obtaining target user data, wherein the target user data comprises: target user attributes and target user behavior;

the first tracking module is used for tracking the wisdom base knowledge according to the target user data to obtain the target wisdom base knowledge, wherein the wisdom base knowledge comprises at least one of the following: text, video, sound;

the first recommendation module is used for processing the target wisdom knowledge by utilizing an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom knowledge;

and the second recommendation module is used for inputting the initial recommendation result into the trained recommendation model to obtain a final recommendation result, wherein the final recommendation result comprises a plurality of finally recommended wisdom knowledge.

According to the embodiment of the present disclosure, the training device of the intellectual knowledge recommendation model includes:

a second obtaining module, configured to obtain sample user data, where the sample user data includes: sample user attributes and sample user behaviors;

the second tracking module is used for tracking the intellectual property base knowledge according to the sample user data to obtain an intellectual property base knowledge sample, wherein the intellectual property base knowledge comprises at least one of the following: text, video, sound;

the third recommendation module is used for processing the wisdom base knowledge samples by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom base knowledge samples;

the training sample building module is used for taking the sample user data and the initial recommendation result as samples in a sample pool, carrying out classification prediction on the samples in the sample pool by using a classifier to obtain marked samples, and adding the marked samples into a training set to build a training sample set;

and the training module is used for training the untrained recommendation model by utilizing the training sample set to obtain the trained recommendation model.

A third aspect of the present disclosure provides an electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of intelligence knowledge recommendation.

A fourth aspect of the present disclosure also provides a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the method of intelligence-base knowledge recommendation described above.

A fifth aspect of the present disclosure also provides a computer program product comprising a computer program which, when executed by a processor, implements the method of intelligence knowledge recommendation described above.

According to the embodiment of the disclosure, the wisdom base knowledge which is interested by the user can be tracked from multiple dimensions according to the user attribute and the user behavior, so that the wisdom base knowledge which is interested by the user in a larger range is obtained, and the method is more comprehensive than the traditional single tracking mode which mainly takes searching. The method has the advantages that initial recommendation is performed on the intellectual database knowledge which is interested by a user in a large range through an individual recommendation algorithm, an initial recommendation result is obtained, the initial recommendation result is combined with active learning deep recommendation, more accurate recommendation can be provided for the user, and the problem of inaccurate recommendation in the prior art is solved.

Drawings

The foregoing and other objects, features and advantages of the disclosure will be apparent from the following description of embodiments of the disclosure, which proceeds with reference to the accompanying drawings, in which:

FIG. 1 schematically illustrates a system architecture diagram of a method, apparatus, device, storage medium and program product for intelligence knowledge recommendation in accordance with embodiments of the present disclosure;

FIG. 2 schematically illustrates a flow diagram of a method of intelligence-base knowledge recommendation in accordance with an embodiment of the disclosure;

FIG. 3 schematically illustrates a framework diagram of a method of intelligence-base knowledge recommendation in an embodiment of the disclosure;

FIG. 4 schematically illustrates a flow diagram of recommendation model training for intellectual knowledge in accordance with an embodiment of the disclosure;

FIG. 5 schematically shows a block diagram of a wisdom knowledge recommendation device according to an embodiment of the present disclosure;

FIG. 6 schematically illustrates a block diagram of a training apparatus for a wisdom-based knowledge recommendation model according to an embodiment of the present disclosure; and

FIG. 7 schematically illustrates a block diagram of an electronic device adapted to implement a method of intelligence-base knowledge recommendation in accordance with an embodiment of the disclosure.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is illustrative only and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It is noted that the terms used herein should be interpreted as having a meaning that is consistent with the context of this specification and should not be interpreted in an idealized or overly formal sense.

Where a convention analogous to "at least one of A, B and C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B and C" would include but not be limited to systems that have a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, necessary security measures are taken, and the customs of the public order is not violated.

In the technical scheme of the embodiment of the disclosure, before the personal information of the user is obtained or collected, the authorization or the consent of the user is obtained.

The wisdom base is a stable social organization, and refers to a consulting and researching organization specialized in developing research. The method gathers expert and scholars of each subject, provides a satisfactory scheme or an optimization scheme for the development of the fields of social economy and the like by applying the intelligence and talent of the expert and scholars, and is an indispensable important component in the modern leadership management system. The main task is to provide consultation, to contribute to the decision maker, to judge operation and to provide various designs; and (4) feeding back information, tracking, investigating and researching the embodiment, and feeding back the operation result to a decision maker so as to facilitate deviation correction. For better understanding, an environment intelligence library is taken as an example for explanation, the environment intelligence library is composed of environment expert scholars, focuses on global environment problems, provides public research institutions of ideas, strategies and the like for the country, the society and the like, and bears social functions of national environment policy service, service knowledge propagation, low-carbon development of the service society and the like.

Wisdom knowledge is the information that provides valuable information to everyone, such as web page text, voice, video, etc., without being limited to the ones listed. For example, an environmental protection favorite can obtain the information related to the environment from the website by browsing the related wisdom library website, and the process of finding the information related to the environment is wisdom library knowledge tracking.

In the following, the target wisdom base knowledge and wisdom base knowledge samples are used as examples of the environmental wisdom base knowledge, but the examples are only used for illustrating the target wisdom base knowledge and wisdom base knowledge samples, and the target wisdom base knowledge and wisdom base knowledge samples provided by the embodiments of the present disclosure are not limited to the environmental wisdom base knowledge.

With the increasing severity of ecological environmental pollution, public and society pay more attention to environmental protection, people's environmental awareness is gradually improved, and environmental intelligence plays an increasingly important role. The environment intelligence library provides a large amount of knowledge such as sustainable development concepts, low-carbon policies and the like for citizens, and makes a great contribution to the scientificity of environment protection. However, the knowledge in the intellectual property library is huge in quantity and various in variety, and how to quickly track the knowledge in which the user is interested from the massive intellectual property library knowledge becomes a problem to be solved urgently.

In the process of implementing the present disclosure, it is found that the traditional wisdom base knowledge tracking technology mostly adopts a single tracking mode mainly based on "search", but with the richness of wisdom base knowledge, users have difficulty in expressing own requirements by using several search terms. The method and the system realize the tracking of the knowledge of the environment intelligence base based on multiple dimensions such as intelligent search, historical browsing, user attention, individual recommendation and the like, and realize accurate depiction of user preference.

However, a single personality recommendation technology is mostly adopted in the traditional wisdom base knowledge recommendation, but the personality recommendation technology has the condition of insufficient standard data. Under the condition, the active learning model is proposed, the label is obtained by selecting the most needed data of the model at present for labeling, and then the sample and the label are added into the training set, so that the aim of constructing a proper training sample set at lower cost is fulfilled. The untrained recommendation model is trained by the well-constructed training set, so that the trained recommendation model can be obtained, and the problem of unreliable recommendation caused by sparse data and uneven distribution in the personalized recommendation algorithm is solved. Therefore, on the basis of the personalized recommendation algorithm, the active learning recommendation strategy is introduced to help construct a proper training sample set, so that a more reliable recommendation result is obtained.

The embodiment of the disclosure provides a method for recommending wisdom knowledge, which comprises the following steps:

acquiring target user data, wherein the target user data comprises: target user attributes and target user behavior; tracking the wisdom base knowledge according to the target user data to obtain the target wisdom base knowledge, wherein the wisdom base knowledge comprises at least one of the following: text, video, sound; processing the target wisdom base knowledge by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom base knowledge; and inputting the initial recommendation result into a trained recommendation model to obtain a final recommendation result, wherein the final recommendation result comprises a plurality of finally recommended wisdom knowledge.

According to the embodiment of the disclosure, the wisdom base knowledge which is interested by the user can be tracked from multiple dimensions according to the user attribute and the user behavior, so that the wisdom base knowledge which is interested by the user in a larger range is obtained, and the method is more comprehensive than the traditional single tracking mode which mainly takes searching. And then, performing initial recommendation on the intellectual database knowledge which is interested by the user in a larger range through an individual recommendation algorithm to obtain an initial recommendation result, and combining the initial recommendation result with the active learning deep recommendation to provide more accurate recommendation for the user, thereby solving the problem of inaccurate recommendation in the prior art.

FIG. 1 schematically illustrates a system architecture diagram of a method, apparatus, device, storage medium and program product for intelligence-based knowledge recommendation, in accordance with embodiments of the present disclosure.

As shown in fig. 1, the application scenario 100 according to this embodiment may include

terminal devices

101, 102, a network device 104, and a server 103. The network 104 serves as a medium for providing communication links between the

terminal devices

101, 102 and the server 103. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user may use the

terminal devices

101, 102 to interact with the server 103 via the network 104 to receive or send messages or the like. The

terminal devices

101, 102 may have installed thereon various communication client applications and programs, such as shopping-like applications, web browser applications, search-like applications, instant messaging tools, mailbox clients, social platform software, etc. (by way of example only) and programs.

The

terminal devices

101, 102 may be various electronic devices having a display screen and supporting web browsing, including but not limited to tablet computers, laptop portable computers, desktop computers, and the like.

The server 103 may be a server providing various services, such as a background management server (for example only) providing support for websites browsed by users using the

terminal devices

101, 102. The background management server may analyze and perform other processing on the received data such as the user request, and feed back a processing result (e.g., a webpage, information, or data obtained or generated according to the user request) to the terminal device.

It should be noted that the method for recommending the wisdom base knowledge provided by the embodiment of the present disclosure may be generally executed by the server 103. Accordingly, the apparatus for intelligence knowledge recommendation provided by the embodiments of the present disclosure may be generally disposed in the server 103. The method of wisdom base knowledge recommendation provided by the embodiments of the present disclosure may also be performed by a server or a cluster of servers different from the server 103 and capable of communicating with the

terminal devices

101, 102 and/or the server 103. Accordingly, the apparatus for wisdom-based knowledge recommendation provided by the embodiment of the present disclosure may also be disposed in a server or a server cluster different from the server 103 and capable of communicating with the

terminal devices

101 and 102 and/or the server 103.

The method for recommending the wisdom base knowledge provided by the embodiment of the disclosure may also be executed by the

terminal devices

101 and 102, and the apparatus for recommending the wisdom base knowledge provided by the embodiment of the disclosure may also be generally disposed in the

terminal devices

101 and 102. The method for wisdom-based knowledge recommendation provided by the embodiments of the present disclosure may also be performed by other terminals different from the

terminal devices

101 and 102. Accordingly, the apparatus for wisdom-based knowledge recommendation provided by the embodiment of the present disclosure may also be disposed in other terminals of different

terminal devices

101 and 102.

It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

The method of wisdom base knowledge recommendation of the disclosed embodiments will be described in detail below with fig. 2-4 based on the scenario described in fig. 1.

FIG. 2 schematically shows a flow diagram of a method of intelligence-base knowledge recommendation, in accordance with an embodiment of the disclosure.

As shown in fig. 2, the method 200 of intelligence knowledge recommendation of this embodiment includes operations S201 to S204.

In operation S201, target user data is acquired, wherein the target user data includes: target user attributes and target user behavior.

According to an embodiment of the present disclosure, the target user attributes include: gender, age, cultural degree, hobbies and interests of the target user, and the like. Target user behavior data can be obtained through intelligent search, historical browsing, attention and individual recommendation of a target user, wherein the target user behavior comprises the following steps: browse, focus, share, download, etc.

According to the embodiment of the disclosure, the intelligent search technology satisfies the functions of fuzzy search, accurate matching and the like. For example, based on an elastic search data analysis engine, the method supports fuzzy search using 'keywords' as search conditions, and supports multi-attribute and multi-condition precise search.

In operation S202, tracking the wisdom-based knowledge according to the target user data to obtain a target wisdom-based knowledge, wherein the wisdom-based knowledge includes at least one of: text, video, sound.

According to the embodiment of the disclosure, a table structure of a database is designed according to target user attributes and target user behavior data, such as: designing an intelligent search record table of a target user, a browsing record table of the target user, an attention record table of the target user and a downloading record table of the target user. And tracking the target wisdom knowledge by taking the record tables as a reference through extracting the attribute characteristics and the behavior characteristics of the target user, and acquiring the wisdom knowledge which is probably liked by the user.

In operation S203, the target wisdom knowledge is processed by using a personalized recommendation algorithm to obtain an initial recommendation result, where the initial recommendation result includes a plurality of initially recommended wisdom knowledge.

According to the embodiment of the disclosure, the target wisdom knowledge is processed by using the individual recommendation algorithm, so that an initial recommendation result can be obtained, namely the wisdom knowledge which a user may like can be obtained. The personalized recommendation algorithm comprises the following steps: and (4) based on individual recommendation algorithms such as heat, collaborative filtering, matrix decomposition and content.

According to the embodiment of the disclosure, the target wisdom knowledge is processed by utilizing the individual recommendation algorithm based on the heat degree, and the wisdom knowledge based on the initial recommendation of the heat degree is obtained.

Firstly, five types of environment intelligence knowledge of 'environment management', 'low-carbon economy', 'low-carbon city', 'sustainable development countermeasure' and 'complex environment system' are selected and given initial heat value S_oriAssuming an initial heat value S_ori＝20。

The heat value S of certain intellectual property base knowledge is increased along with the interactive action of the target user on the certain intellectual property base knowledge_user. For example, for a wisdom knowledge, the score increases by S every time it is searched_seaEach time it is downloaded, the score is increased by S_dowEach time the browser is browsed, the score is increased by S_broScore, each time it is focused on, increases score by S_folThe score can be obtained, and the target user behavior score is S_user＝S_sea+S_dow+S_bro+S_fol。

Because the intelligence knowledge has stronger timeliness, the intelligence knowledge shows a decay trend along with time, and then S is generated_timeThe attenuation value of (2). Therefore, the above five may be obtained according to the initial heat value of the intellectual property knowledge, the user behavior score, and the attenuation value of the intellectual property knowledgeThe heat value of the type of environment intelligence base knowledge is S ═ S_ori+S_user-S_time. Finally, the intelligence knowledge is sorted in a descending order according to the heat value of the intelligence knowledge, so that ten environment-related intelligence knowledge recommended based on the highest heat can be provided for the target user, and it should be noted that the recommended number may not be limited to ten.

According to the embodiment of the disclosure, the target wisdom knowledge is processed by utilizing a collaborative filtering-based personalized recommendation algorithm, so that the wisdom knowledge based on collaborative filtering initial recommendation is obtained.

And processing the environmental wisdom knowledge by adopting a collaborative filtering algorithm of the user, firstly, searching for users similar to the target user, and then selecting the environmental wisdom knowledge which is interested by the similar users and is not browsed by the target user for recommendation aiming at each similar user.

Calculating the similarity of the users by adopting a Pearson correlation coefficient method, and finding out 5 users most similar to the target user, wherein the similarity of the 5 users is as follows:

user U_aAnd similarity: 0.8742, respectively; user U_bAnd similarity: 0.8611, respectively; user U_cSimilarity 0.7823; user U_dSimilarity 0.7644; user U_eSimilarity 0.7422.

Then, the knowledge of the environmental wisdom that they are interested in and that the target user has not browsed is sorted out for these 5 similar users and their scores are calculated.

For example, user U_aThe most interesting environmental wisdom knowledge is wisdom knowledge I_a、I_b、I_c… …, scores of 0.987, 0.982 and 0.974 … ….

Finally, weighting is carried out according to the similarity of the users and the scores of the environmental wisdom knowledge interested by the similar users, ten highest environmental wisdom knowledge recommended based on collaborative filtering can be provided for the target user, and the number of recommendations can be not limited to ten.

According to the embodiment of the disclosure, the target intellectual property knowledge is processed by utilizing a matrix decomposition-based individual recommendation algorithm, and the intellectual property knowledge initially recommended based on matrix decomposition is obtained.

Adopting Baseline SVD method to count two-dimensional scoring matrix of user and environment wisdom knowledge, such as user U_a、U_b、U_c… … intellectual property knowledge I_a、I_b、I_c… …, score S by stochastic gradient descent method_aa、S_ab、S_ac… …, the unknown scoring symbol "? "means. The two-dimensional scoring matrix is shown in table 1:

TABLE 1 two-dimensional scoring matrix

Score S calculated by random gradient descent method_ab、S_ba、S_ccThe scores of (A) are respectively: 0.869, 0.756 and 0.789, sorting the recommendations according to the scores in a descending order, and providing the user with the ten environment-related wisdom knowledge recommended at the highest based on matrix decomposition.

According to the embodiment of the disclosure, the target wisdom knowledge is processed by using a content-based individual recommendation algorithm to obtain the wisdom knowledge based on content initial recommendation.

Firstly, extracting the content, title, type and other characteristic information in the environment intelligence knowledge, coding the characteristic information extracted from the environment intelligence knowledge by using a doc2vec model, and converting the characteristic information into a word vector I_aImplementing a feature representation of the wisdom-base knowledge, wherein the feature is represented by x_iAnd representing the characteristic dimension as n.

Then, describing the characteristic preference of the target user by using the historical behavior of the target user, and obtaining a preference vector U of the target user by performing characteristic learning on the historical behavior of the target user_bWherein the feature is represented by y_iAnd representing the characteristic dimension as n.

Finally, by calculating I_aAnd U_bThe cosine similarity of (2) is obtained by carrying out the cosine similarity on the obtained cosine valuesThe descending order can provide the target user with the ten environmental-related wisdom knowledge based on the highest content recommendation, and it should be noted that the number of recommendations may not be limited to ten.

In operation S204, the initial recommendation result is input into the trained recommendation model to obtain a final recommendation result, where the final recommendation result includes a plurality of finally recommended wisdom knowledge.

According to the embodiment of the disclosure, the intelligence knowledge is processed by using the individual recommendation algorithm, the obtained plurality of initially recommended intelligence knowledge are summarized to obtain an initial recommendation result, and the characteristics of the initial recommendation result are vectorized and used as the input of the trained recommendation model.

For example, the total features of 40 Chile knowledge obtained from the initial recommendation are input into the trained recommendation model, wherein the total features are obtained by the target user U_aAnd intellectual knowledge base I_aAnd (4) forming.

Obtaining the top 10 wisdom base knowledge with the highest score as the final recommendation result, such as giving the target user U_aThe final recommended wisdom base knowledge is I_a、I_b、I_c… …, scores of 0.891, 0.895, 0.876 … ….

Fig. 3 schematically shows a framework diagram of a method for intelligence-base knowledge recommendation according to an embodiment of the present disclosure.

According to the embodiment of the present disclosure, as shown in fig. 3, in order to quickly track the wisdom knowledge, user data is obtained from multiple dimensions of user attention, smart search, history browsing, and personality recommendation, and the wisdom knowledge in which the user is interested is tracked according to the user data. And processing the intellectual database knowledge by using a personalized recommendation algorithm based on heat, content, collaborative filtering and matrix decomposition in the personalized recommendation algorithm to obtain an initial recommendation result. And taking the user data and the initial recommendation result as samples of a sample pool in the recommendation model, classifying the samples in the sample pool by adopting an uncertainty reduced active learning algorithm strategy, and labeling the samples which do not meet preset conditions to obtain the labeled samples. And then, adding the marked samples into a training set to serve as training samples, and continuously classifying and marking the unlabeled samples in the sample pool by using an uncertainty reduced active learning algorithm strategy to complete the construction of the training sample set. And then, training the untrained recommendation model by using the training sample to obtain the trained recommendation model. . In the application process, the result of the initial recommendation is input into the trained recommendation model, so that a more accurate recommendation result can be obtained.

Through the embodiment of the disclosure, according to the user attribute and the user behavior, the intellectual property knowledge which is interested by the user is tracked from multiple dimensions, the intellectual property knowledge which is interested by the user in a larger range is obtained, and the method is more comprehensive than the traditional single tracking mode which mainly takes searching. The method has the advantages that initial recommendation is performed on the intellectual database knowledge which is interested by a user in a large range through the individual recommendation algorithm, an initial recommendation result is obtained, the initial recommendation result is combined with deep recommendation of the uncertainty-reduced active learning algorithm, more accurate recommendation can be provided for the user, and the problem of inaccurate recommendation in the prior art is solved.

FIG. 4 schematically shows a flowchart of recommendation model training for intellectual knowledge in accordance with an embodiment of the present disclosure.

As shown in fig. 4, the training method 400 of the intellectual property recommendation model of this embodiment includes operations S401 to S405.

In operation S401, sample user data is acquired, wherein the sample user data includes: sample user attributes and sample user behavior.

According to an embodiment of the present disclosure, the sample user attributes include: the gender, age, cultural degree, hobbies and interests and the like of the sample user. Sample user behavior data can be obtained through intelligent search, historical browsing, attention and individual recommendation of sample users, wherein the sample user behavior comprises the following steps: browse, focus, share, download, etc.

In operation S402, the wisdom knowledge is tracked according to the sample user data, and a wisdom knowledge sample is obtained.

According to the embodiment of the disclosure, a table structure of the database is designed according to the sample user attribute and the sample user behavior data, such as: designing a sample user intelligent search record table, a sample user browsing record table, a sample user attention record table and a sample user downloading record table. And tracking the wisdom knowledge sample by taking the record tables of the sample users as a reference and extracting the attribute characteristics of the sample users and the behavior characteristics of the sample users to obtain the wisdom knowledge sample which is possibly liked by the user.

In operation S403, the wisdom knowledge samples are processed by using a personalized recommendation algorithm to obtain an initial recommendation result, where the initial recommendation result includes a plurality of initially recommended wisdom knowledge samples.

According to the embodiment of the disclosure, the individual recommendation algorithm is utilized to process the wisdom base knowledge sample, so that an initial recommendation result can be obtained, and the wisdom base knowledge sample liked by the user can be obtained. The personality recommendation algorithm comprises the following steps: and (4) based on individual recommendation algorithms such as heat, collaborative filtering, matrix decomposition and content.

According to the embodiment of the disclosure, the intelligence base knowledge sample is processed by utilizing the individual recommendation algorithm based on the heat degree, and the intelligence base knowledge sample based on the initial recommendation of the heat degree is obtained.

Firstly, according to different label types of environment wisdom base knowledge, different initial heat values S are given to each type of wisdom base knowledge_oriThe tag type can be environment, economy, green travel, environmental management, low-carbon economy, low-carbon cities, sustainable development strategies, complex environmental systems, culture, society, health, safety, civilization and the like.

The heat value S of a certain intellectual property library knowledge sample is increased along with the interactive behavior of a sample user on the certain environmental intellectual property library knowledge sample_user. For example, for a wisdom knowledge sample, the score increases by S once it is searched_seaEach time it is downloaded, the score is increased by S_dowIs divided into two parts, each part is broiledOnce visited, the score increased by S_broScore, each time it is focused on, increases score by S_folThe score can be obtained, and the sample user behavior score is S_user＝S_sea+S_dow+S_bro+S_fol。

Because the timeliness of the intellectual property knowledge sample is strong, the intellectual property knowledge sample shows a decay trend along with the time, and S is generated_timeThe attenuation value of (2). Therefore, according to the initial heat value of the wisdom knowledge sample, the user behavior score and the attenuation value of the wisdom knowledge sample, the heat value S of the wisdom knowledge sample can be obtained, wherein S is S_ori+S_user-S_time. And finally, performing descending order according to the heat value of the wisdom base knowledge samples, and providing environment wisdom base knowledge samples based on initial recommendation of heat for sample users.

According to the embodiment of the disclosure, the intellectually-knowledgebase sample is processed by utilizing the collaborative filtering-based personalized recommendation algorithm, so that the intellectually-knowledgebase sample based on collaborative filtering initial recommendation is obtained.

And processing the environmental intelligence knowledge samples by adopting a collaborative filtering algorithm of the user, firstly, finding out users similar to the sample user, and then selecting the environmental intelligence knowledge samples which are interested by the similar users and are not browsed by the sample user for recommendation aiming at each similar user.

Calculating the similarity of users by adopting a Pearson correlation coefficient method, searching k users most similar to the sample user u, and calculating the similarity of the similar user v and the sample user u by adopting the following formula (1):

where E denotes the mathematical expectation and cov denotes the covariance.

After the users similar to the sample user u are calculated through the formula (1), k most similar users are selected and a set of environment wisdom base knowledge samples interested by the similar users is selected for the sample user u, the set S (u, k) is used for representing, all the environment wisdom base knowledge samples interested by the similar users v in the S are extracted, the environment wisdom base knowledge interested by the sample user u is removed, and the environment wisbase knowledge interested by the similar users and not browsed by the sample user is obtained and scored. Then, the scores and the similarity are weighted and arranged in a descending order, and then the environment intelligence base knowledge sample based on the collaborative filtering initial recommendation can be obtained. Wherein, for each possible recommended environmental intelligence knowledge i, the degree of interest of the similar user v is shown in formula (2):

p(u,i)＝∑_{v∈S(u,k)∩N(i)}w_uv×r_vi (2)

wherein r is_viIndicates the like degree of similar users v to i, w_uvRepresenting the similarity between sample user u and similar user v.

According to the embodiment of the disclosure, the intellectually-knowledgebase sample is processed by utilizing a matrix decomposition-based individual recommendation algorithm, so that the intellectually-knowledgebase sample initially recommended based on matrix decomposition is obtained.

And discovering potential factors of the environmental wisdom knowledge by adopting a Baseline SVD method through a matrix decomposition mode, and describing the characteristics of the sample user and the environmental wisdom knowledge sample by using the factor vectors.

Wherein, the calculation formula of the variable value of the Baseline SVD method is shown as the following (3):

wherein r is_u,iRepresents the score of a sample user u on i-Chile knowledge, u represents a global mean, b_uIndicating the deviation value of the user, b_iDeviation value, p, representing intellectual knowledge_u,kRepresenting the relationship of the sample user u and the kth potential class, q_i,kAnd expressing the relation of the Chile knowledge i to the kth potential class, wherein the size of K is determined by a specific number, and lambda is a regularization factor.

The equation is solved using a random gradient descent method and a least squares method. And finally, performing descending order according to the score S to obtain the initially recommended environment intelligence base knowledge sample based on matrix decomposition.

According to the embodiment of the disclosure, the intellectually-knowledgeable sample is processed by using a content-based individual recommendation algorithm, so that the intellectually-knowledgeable sample based on content initial recommendation is obtained.

Firstly, extracting the characteristic information of content, title, type and the like in the knowledge of the environmental intelligence base, coding the characteristic information extracted from the knowledge of the environmental intelligence base by using a doc2vec model, converting the extracted characteristic information into word vectors, and realizing the characteristic representation of the knowledge of the intelligence base, wherein the characteristic is represented by x_iAnd representing the characteristic dimension as n.

Then, describing the characteristic preference of the sample user by the historical behavior of the sample user, and obtaining the preference vector U of the sample user by performing characteristic learning on the historical behavior of the sample user_bThe feature is denoted by yi, and the feature dimension is n.

And comparing by using a cosine similarity formula to obtain a wisdom base knowledge sample based on content initial recommendation. Wherein, the cosine similarity calculation formula is shown as the following (4):

finally, by calculating I_aAnd U_bAnd (4) the cosine similarity, and the obtained cosine values are arranged in a descending order, so that the environmental intelligence base knowledge sample based on the content initial recommendation can be obtained.

In operation S404, the sample user data and the initial recommendation result are used as samples in a sample pool, a classifier is used to perform classification prediction on the samples in the sample pool to obtain labeled samples, and the labeled samples are added to a training set to construct a training sample set.

According to the embodiment of the disclosure, the processing intelligence knowledge samples based on the heat, content, collaborative filtering and matrix decomposition personalized recommendation algorithm are summarized to obtain the initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended environment intelligence knowledge samples. The user data and the total data of the initial recommendation are then used as inputs to the recommendation model.

According to the embodiment of the disclosure, an active learning algorithm strategy based on uncertainty reduction is adopted to carry out deep environment intelligence base knowledge recommendation. Firstly, the change rate of sample class labels is used as a weighing factor in a sample pool, a classifier is used for classifying and predicting samples in the sample pool, the sample with the largest uncertainty is continuously screened out and labeled, the labeled sample is added into a training set, and a final training sample set is constructed.

The samples with class targets are used as a training set, the samples without class targets are used as a sample pool, and the threshold value of the change rate of the sample prediction class targets in the sample pool is set to be mu. Training a classifier C by taking the training set as a reference, and performing classification prediction on the samples in the sample pool by using the classifier C.

And under the condition that the change rate of the classifiers for predicting class labels of the samples in the sample pool in two adjacent times is greater than a threshold value, calculating the uncertainty of each sample class in the sample pool, labeling the sample with the maximum uncertainty in the sample pool to obtain a labeled sample, removing the labeled sample from the sample pool, and adding the labeled sample into a training set to construct a training sample set. And continuing to train the classifier C on the newly formed training sample set, and then classifying and predicting the unlabeled samples in the sample pool by using the classifier C until the change rate of the classifiers which are adjacent twice on the prediction class labels of the samples in the sample pool is smaller than a threshold value, stopping classifying and predicting the samples in the sample pool, and completing the construction of the training sample set.

It should be noted that the manner of labeling the samples in the sample pool is not limited to manual labeling, and may be machine labeling.

Through the embodiment of the disclosure, a reasonable training sample set is screened out for the recommendation model by using an active learning algorithm based on uncertainty reduction and adopting a labeling mode, so that the technical problems of data sparsity and uneven distribution are solved.

According to the embodiment of the disclosure, XGboost is used as a classifier of an active learning algorithm, and the algorithm idea is to continuously perform feature splitting to add a tree, wherein the tree is added by learning a new function to fit the residual error of the last prediction. And obtaining K trees after training, knowing the leaf nodes corresponding to the sample falling into each tree according to the characteristics of the sample, and adding the scores of the leaf nodes corresponding to each tree to obtain the predicted value of the sample. The XGBoost algorithm objective function is divided into two parts: an error function L (θ) and a regularization term Ω (θ). The formula is shown in (5) below:

obj(θ)＝L(θ)+Ω(θ) (5)

the training loss function is expressed as follows (6):

wherein, y_iIn order to be the true value of the value,

is a predicted value.

The newly generated tree needs to be fitted with the residual value predicted last time, and after t trees are generated, the prediction score can be changed into:

meanwhile, the objective function can be rewritten as:

wherein g is_iIs the first derivative, h_iIs the second derivative.

The residual error between the prediction score of the first t-1 tree and y does not influence the optimization of the objective function, and can be directly removed. The simplified objective function is:

the regularization term for the XGboost algorithm:

the complexity of the XGboost model to the model is mainly punished on the number T of leaf nodes and the predicted value w of the leaf nodes, and the punishment on the number of the leaf nodes is equivalent to pruning operation on the tree. Wherein w_jA predicted value representing the jth leaf node in a tree; t represents a tree with T leaf nodes; gamma and lambda are self-defined values, and can be set when using a model, if gamma is large, the penalty is larger when the number of leaf nodes of the tree is larger, and lambda can penalize the total predicted value of the leaf nodes.

Let the prediction result of the t-th classifier be w_q(x)：

f_t(x)＝w_q(x) (13)

Where q (x) is a function of the leaf node, the objective function may be changed to:

order:

the final objective function can be written as:

thus obtaining the best

And obj^*：

The XGboost objective function value is used as an evaluation function, and all characteristic division points are traversed by a greedy algorithm. The specific method is that the value of the objective function after splitting is larger than the value of the gain of the objective function of a single leaf node, meanwhile, in order to limit the tree from growing too deeply, a threshold value is set, splitting is carried out only when the gain is larger than the threshold value, meanwhile, the maximum depth of the tree is set, and the growth is stopped to prevent overfitting when the weight of a sample is smaller than the set threshold value. Left and right leaf nodes are distinguished by subscripts L and R respectively, the Gain fraction can be calculated as Gain, and the formula is as follows:

according to an embodiment of the present disclosure, sample user data and total data of the initial recommendation result are divided into 3 parts, wherein 25% is used as a training set, 25% is used as a test set, and 50% is used as samples in a sample pool for selection by a classifier. In the embodiment of the present disclosure, the threshold of the change rate of the sample class target is set to 0.02, and when the change rate of the sample class in the sample pool by the two adjacent classifiers is less than 0.02, the construction of the training sample set is completed.

In operation S405, the untrained recommendation model is trained by using the training sample set, so as to obtain a trained recommendation model.

According to the embodiment of the present disclosure, when the selected number of samples in the sample pool is 30, the rate of change of the sample class is 0.018, which is smaller than the set threshold value of 0.02, and the test accuracy of XGboost at this time is 0.8321. When all samples in the sample pool are annotated, the testing accuracy of XGBoost is only 0.8123. The active learning algorithm-based strategy provided by the disclosure can achieve low-cost labeling and realize a high-precision training effect.

According to the embodiment of the disclosure, the initial recommended wisdom knowledge is ranked based on the XGboost algorithm, and the first N pieces of environment wisdom knowledge with the highest scores are obtained as the final recommended result.

In order to evaluate the recommendation effect, AUC and CVR are used as evaluation indexes to measure the prediction error and the adoption degree of the recommendation result of the recommendation model by the user.

The definition of AUC refers to randomly extracting a positive sample and a negative sample from a positive and negative sample set, where the predicted value of the positive sample is greater than the probability of the negative sample, and the calculation formula of AUC is as follows (18):

the denominator represents the total combination number of the positive and negative samples, and the numerator means that the prediction probability of the positive sample is greater than the combination number of the negative sample.

CVR refers to conversion rate, and is an index for measuring advertisement effectiveness. In the disclosed embodiment, when the user clicks on the recommended environmental wisdom knowledge, such as text in the environmental wisdom knowledge, and browses for more than 20s, the recommendation is considered to be valid. When the browsing time is too short, the user's interest in the text is not great, wherein the calculation formula of CVR is as follows (19):

among them, effective_numMeans effective number of clicks, total_numRefers to the total number of clicks.

For example, AUC and CVR are used as evaluation indexes to evaluate the final recommendation effect. A single personalized recommendation algorithm such as an active learning algorithm based on heat, collaborative filtering, matrix decomposition and content and the technology is compared in an experiment, and specific comparison results are shown in a table 2.

TABLE 2 result of recommendation by single personality recommendation algorithm and active learning algorithm

As can be seen from table 2, the recommendation effect based on the active learning algorithm is better than that of the single personalized recommendation algorithm.

Based on the method for recommending the wisdom base knowledge, the disclosure also provides a device for recommending the wisdom base knowledge. The apparatus will be described in detail below with reference to fig. 5.

Fig. 5 schematically shows a block diagram of a wisdom-knowledge recommendation apparatus according to an embodiment of the present disclosure.

As shown in fig. 5, the wisdom-knowledge recommending apparatus 500 of this embodiment includes a first obtaining module 501, a first tracking module 502, a first recommending module 503, and a second recommending module 504.

The first obtaining module 501 is configured to obtain target user data, where the target user data includes: target user attributes and target user behavior. In an embodiment, the first obtaining module 501 may be configured to perform the operation S201 described above, which is not described herein again.

The first tracking module 502 is configured to track the wisdom-based knowledge according to the target user data to obtain the target wisdom-based knowledge, wherein the wisdom-based knowledge includes at least one of: text, video, sound. In an embodiment, the first tracking module 502 may be configured to perform the operation S202 described above, and is not described herein again.

The first recommending module 503 is configured to process the target wisdom knowledge by using a personalized recommendation algorithm to obtain an initial recommendation result, where the initial recommendation result includes a plurality of initially recommended wisdom knowledge. In an embodiment, the first recommending module 503 may be configured to perform the operation S203 described above, which is not described herein again.

The second recommending module 504 is configured to input the initial recommending result into the trained recommending model to obtain a final recommending result, where the final recommending result includes a plurality of finally recommended wisdom knowledge. In an embodiment, the second recommending module 504 may be configured to perform the operation S204 described above, which is not described herein again.

According to the embodiment of the present disclosure, any plurality of the first obtaining module 501, the first tracking module 502, the first recommending module 503 and the second recommending module 504 may be combined and implemented in one module, or any one of them may be split into a plurality of modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of the other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the first obtaining module 501, the first tracking module 502, the first recommending module 503 and the second recommending module 504 may be implemented at least partially as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging a circuit, or implemented by any one of three implementation manners of software, hardware and firmware, or any suitable combination of any of them. Alternatively, at least one of the first obtaining module 501, the first tracking module 502, the first recommending module 503 and the second recommending module 504 may be at least partially implemented as a computer program module, which when executed, may perform a corresponding function.

Fig. 6 schematically shows a block diagram of a training apparatus of a wisdom-based knowledge recommendation model according to an embodiment of the present disclosure.

As shown in fig. 6, the training apparatus 600 of the wisdom-based knowledge recommendation model of this embodiment includes a second obtaining module 601, a second tracking module 602, a third recommending module 603, a building training sample module 604, and a training module 605.

The second obtaining module 601 is configured to obtain sample user data, where the sample user data includes: sample user attributes and sample user behavior. In an embodiment, the second obtaining module 601 may be configured to perform the operation S401 described above, which is not described herein again.

The second tracking module 602 is configured to track the wisdom knowledge according to the sample user data to obtain a wisdom knowledge sample, where the wisdom knowledge includes at least one of: text, video, sound. In an embodiment, the second tracking module 602 may be configured to perform the operation S402 described above, which is not described herein again.

The third recommending module 603 is configured to process the wisdom base knowledge samples by using an individual recommendation algorithm to obtain an initial recommendation result, where the initial recommendation result includes a plurality of initially recommended wisdom base knowledge samples. In an embodiment, the third recommending module 603 may be configured to perform the operation S403 described above, which is not described herein again.

The training sample construction module 604 is configured to use the sample user data and the initial recommendation result as samples in a sample pool, perform classification prediction on the samples in the sample pool by using a classifier to obtain labeled samples, and add the labeled samples into a training set to construct a training sample set. In an embodiment, the training sample constructing module 604 may be configured to perform the operation S404 described above, which is not described herein again.

The training module 605 is configured to train the untrained recommendation model by using the training sample set to obtain a trained recommendation model, in an embodiment, the training module 605 may be configured to perform the operation S405 described above, which is not described herein again.

According to the embodiment of the present disclosure, any of the second obtaining module 601, the second tracking module 602, the third recommending module 603, the training sample constructing module 604 and the training module 605 may be combined and implemented in one module, or any one of them may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of the other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the second obtaining module 601, the second tracking module 602, the third recommending module 603, the constructing training sample module 604 and the training module 605 may be at least partially implemented as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging a circuit, or implemented by any one of three implementations of software, hardware and firmware, or by a suitable combination of any several of them. Alternatively, at least one of the second obtaining module 601, the second tracking module 602, the third recommending module 603, the construct training sample module 604 and the training module 605 may be at least partially implemented as a computer program module, which when executed, may perform a corresponding function.

FIG. 7 schematically illustrates a block diagram of an electronic device adapted to implement a method of intelligence-base knowledge recommendation, in accordance with an embodiment of the disclosure.

As shown in fig. 7, an electronic device 700 according to an embodiment of the present disclosure includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM)702 or a program loaded from a storage section 708 into a Random Access Memory (RAM) 703. The processor 701 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or associated chipset, and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), among others. The processor 701 may also include on-board memory for caching purposes. The processor 701 may comprise a single processing unit or a plurality of processing units for performing the different actions of the method flows according to embodiments of the present disclosure.

In the RAM703, various programs and data necessary for the operation of the electronic apparatus 700 are stored. The processor 701, the ROM 702, and the RAM703 are connected to each other by a bus 704. The processor 701 performs various operations of the method flows according to the embodiments of the present disclosure by executing programs in the ROM 702 and/or the RAM 703. It is noted that the programs may also be stored in one or more memories other than the ROM 702 and RAM 703. The processor 701 may also perform various operations of method flows according to embodiments of the present disclosure by executing programs stored in the one or more memories.

Electronic device 700 may also include input/output (I/O) interface 705, which input/output (I/O) interface 705 is also connected to bus 704, according to an embodiment of the present disclosure. The electronic device 700 may also include one or more of the following components connected to the I/O interface 705: an input portion 706 including a keyboard, a mouse, and the like; an output section 707 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 709 performs communication processing via a network such as the internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read out therefrom is mounted into the storage section 708 as necessary.

The present disclosure also provides a computer-readable storage medium, which may be contained in the apparatus/device/system described in the above embodiments; or may exist separately and not be assembled into the device/apparatus/system. The computer-readable storage medium carries one or more programs which, when executed, implement the method according to an embodiment of the disclosure.

According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, a computer-readable storage medium may include the ROM 702 and/or the RAM703 and/or one or more memories other than the ROM 702 and the RAM703 described above.

Embodiments of the present disclosure also include a computer program product comprising a computer program containing program code for performing the method illustrated in the flow chart. When the computer program product is run on a computer system, the program code is used for causing the computer system to implement the method for intelligence knowledge recommendation provided by the embodiments of the present disclosure.

The computer program performs the above-described functions defined in the system/apparatus of the embodiments of the present disclosure when executed by the processor 701. The systems, apparatuses, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the present disclosure.

In one embodiment, the computer program may be hosted on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted in the form of a signal on a network medium, distributed, downloaded and installed via the communication section 709, and/or installed from the removable medium 711. The computer program containing program code may be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 709, and/or installed from the removable medium 711. The computer program, when executed by the processor 701, performs the above-described functions defined in the system of the embodiment of the present disclosure. The systems, devices, apparatuses, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the present disclosure.

In accordance with embodiments of the present disclosure, program code for executing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, these computer programs may be implemented using high level procedural and/or object oriented programming languages, and/or assembly/machine languages. The programming language includes, but is not limited to, programming languages such as Java, C + +, python, the "C" language, or the like. The program code may execute entirely on the user computing device, partly on the user device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).

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 disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block 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 will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

It will be appreciated by a person skilled in the art that various combinations or/and combinations of features recited in the various embodiments of the disclosure and/or in the claims may be made, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, various combinations and/or combinations of the features recited in the various embodiments and/or claims of the present disclosure may be made without departing from the spirit or teaching of the present disclosure. All such combinations and/or associations are within the scope of the present disclosure.

The embodiments of the present disclosure have been described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described separately above, this does not mean that the measures in the embodiments cannot be used in advantageous combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be devised by those skilled in the art without departing from the scope of the present disclosure, and such alternatives and modifications are intended to be within the scope of the present disclosure.

Claims

1. A method of wisdom base knowledge recommendation, comprising:

acquiring target user data, wherein the target user data comprises: target user attributes and target user behavior;

tracking the wisdom base knowledge according to the target user data to obtain target wisdom base knowledge, wherein the wisdom base knowledge comprises at least one of the following: text, video, sound;

processing the target wisdom base knowledge by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom base knowledge;

and inputting the initial recommendation result into a trained recommendation model to obtain a final recommendation result, wherein the final recommendation result comprises a plurality of finally recommended wisdom knowledge.

2. The method of claim 1, wherein the recommendation model is trained by:

obtaining sample user data, wherein the sample user data comprises: sample user attributes and sample user behaviors;

tracking the intellectual property knowledge according to the sample user data to obtain an intellectual property knowledge sample;

processing the intellectual property samples by using an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended intellectual property samples;

taking the sample user data and the initial recommendation result as samples in a sample pool, carrying out classification prediction on the samples in the sample pool by using a classifier to obtain marked samples, and adding the marked samples into a training set to construct a training sample set;

and training the untrained recommendation model by using the training sample set to obtain the trained recommendation model.

3. The method of claim 2, further comprising:

and evaluating the trained recommendation model to obtain an evaluation result.

4. The method of claim 2, wherein the processing the wisdom knowledge samples by using the personality recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom knowledge samples comprises:

processing the intellectual property base knowledge sample based on a heat individual recommendation algorithm to obtain an intellectual property base knowledge sample based on heat initial recommendation;

processing the wisdom base knowledge sample based on a collaborative filtering personalized recommendation algorithm to obtain a wisdom base knowledge sample based on collaborative filtering initial recommendation;

processing the intellectual property sample based on a matrix decomposition individual recommendation algorithm to obtain an intellectual property sample based on matrix decomposition initial recommendation; and

and processing the intellectual property base knowledge sample based on a content individual recommendation algorithm to obtain the intellectual property base knowledge sample based on content initial recommendation.

5. The method of claim 2, wherein constructing the training sample set comprises:

setting a change rate threshold value of sample prediction class targets in a sample pool as mu;

calculating the uncertainty of each sample class in the sample pool when the change rate of the classifiers adjacent twice to the prediction class labels of the samples in the sample pool is greater than a threshold value;

marking the sample with the maximum uncertainty in the sample pool to obtain a marked sample;

removing the marked samples from the sample pool, and adding the samples into a training sample set to construct a training sample set; or

And when the change rate of the classifiers adjacent twice to the prediction class labels of the samples in the sample pool is smaller than a threshold value, finishing the construction of the training sample set.

6. An intelligence repository knowledge recommendation apparatus comprising:

a first obtaining module, configured to obtain target user data, where the target user data includes: target user attributes and target user behavior;

the first tracking module is used for tracking the wisdom knowledge according to the target user data to obtain the target wisdom knowledge, wherein the wisdom knowledge comprises at least one of the following: text, video, sound;

and the second recommending module is used for inputting the initial recommending result into a trained recommending model to obtain a final recommending result, wherein the final recommending result comprises a plurality of finally recommended wisdom knowledge.

7. The apparatus of claim 6, wherein the training means of the intellectual knowledge recommendation model comprises:

the second tracking module is used for tracking the intellectual property knowledge according to the sample user data to obtain an intellectual property knowledge sample, wherein the intellectual property knowledge comprises at least one of the following: text, video, sound;

the third recommendation module is used for processing the wisdom base knowledge samples by utilizing an individual recommendation algorithm to obtain an initial recommendation result, wherein the initial recommendation result comprises a plurality of initially recommended wisdom base knowledge samples;

a training sample building module, configured to use the sample user data and the initial recommendation result as samples in a sample pool, perform classification prediction on the samples in the sample pool by using a classifier to obtain labeled samples, and add the labeled samples into a training set to build a training sample set;

and the training module is used for training the untrained recommendation model by using the training sample set to obtain the trained recommendation model.

8. An electronic device, comprising:

one or more processors;

a storage device for storing one or more programs,

wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of any of claims 1-5.

9. A computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to perform the method of any one of claims 1 to 5.

10. A computer program product comprising a computer program which, when executed by a processor, implements a method according to any one of claims 1 to 5.