CN114282118A

CN114282118A - Method and device for determining push user of public service item

Info

Publication number: CN114282118A
Application number: CN202210200675.5A
Authority: CN
Inventors: 张爱军; 黄定存
Original assignee: Beijing Qingsongchou Information Technology Co ltd
Current assignee: Beijing Easy Yikang Information Technology Co ltd
Priority date: 2022-03-03
Filing date: 2022-03-03
Publication date: 2022-04-05
Anticipated expiration: 2042-03-03
Also published as: CN114282118B

Abstract

The application provides a method and a device for determining a push user of a public service project, wherein the method comprises the following steps: acquiring a tweet description text of a public welfare item to be recommended; determining a tweet vector of the tweet description text; determining at least one target individual project with higher similarity between the text vector and the tweet vector based on the text vectors of the individual projects in the business service platform; determining at least one candidate user containing a target individual project from a plurality of users based on individual project lists corresponding to the users in a business service platform, wherein the individual project lists corresponding to the users comprise at least one recommendable individual project which is interested in the users; and determining at least one target user to which the public interest item needs to be pushed from the at least one candidate user. The scheme of the application can be used for pushing the public welfare items to the user more pertinently, and the opening rate and the conversion rate of the public welfare items are improved.

Description

Method and device for determining push user of public service item

Technical Field

The present application relates to the field of data processing technologies, and in particular, to a method and an apparatus for determining a push user of a public welfare item.

Background

With the continuous development of the internet, the amount of information generated on the internet is also greatly increased. When a user faces massive information, it is difficult to obtain information really useful for the user from the massive information.

For public interest items, the public interest items focused by different users may also be different. At present, when the text of the public welfare items is pushed, the text pushing of the public welfare items which are interesting to the user can not be recommended to the user in a targeted manner, so that the opening rate and the conversion rate of pushing the public welfare items are not improved, and groups needing help in the public welfare items can not be helped in time.

Disclosure of Invention

The application provides a method and a device for determining a public welfare item pushing user, so that the public welfare item is pushed to the user more pertinently, and the opening rate and the conversion rate of the public welfare item are improved.

In order to achieve the above object, in one aspect, the present application provides a method for determining a push user of a public service item, where the method is applied to a business service platform, and the method includes:

acquiring a tweet description text of a public welfare item to be recommended;

determining a tweet vector of the tweet description text;

determining a preset number of target individual raising projects with high similarity between a text vector and the tweet vector based on the text vectors of the individual raising projects in the business service platform, wherein the individual raising projects are crowd-raising projects initiated by individuals, and the text vectors of the individual raising projects are vectors generated based on project description texts of the individual raising projects;

determining at least one candidate user containing the target individual project from the multiple users on the basis of individual project lists corresponding to the multiple users in the business service platform, wherein the individual project lists corresponding to the users comprise at least one recommendable individual project which is of interest to the users;

and determining at least one target user to which the public interest item needs to be pushed from the at least one candidate user.

In a possible implementation manner, the individual chip item list corresponding to the user is obtained by:

obtaining historical transaction details for each of the plurality of staged projects, the historical transaction details for the respective staged project including at least user information contributing to the respective staged project;

determining that the user contributed at least one contributed research item based on historical transaction details for respective research items, the contributed research item belonging to the plurality of research items;

for each donated staged project contributed by the user, determining a similar items list of the donated staged projects from the plurality of staged projects, the similar items list of the donated staged projects including: at least one similar item of the plurality of items having historical transaction details similar to the historical transaction details of the donated item and a similarity of the similar item to the item;

generating an individual project list of the user based on a similar project list of each donated individual project, wherein recommendable individual projects in the individual project list belong to the similar project list of the donated individual projects.

In yet another possible implementation manner, before the determining the list of similar items of the donated staged items, the method further includes:

respectively determining the similarity between two project research items based on the respective historical transaction details of the project research items;

for each project, determining at least one similar project from the project according to the similarity of the project to other projects, and generating a similar project list of the project, wherein the similar project list of the project comprises: at least one similar individual project similar to the individual project, and a similarity of the similar individual project to the individual project;

the determining the list of similar items of the donated staged item includes:

determining the list of similar items of the donated project from the list of similar items of the plurality of project.

In another possible implementation manner, the list of the respective project stages corresponding to the user includes: at least one recommendable individual project similar to the at least one donated individual project contributed by the user and the corresponding similarity of each recommendable individual project, wherein the corresponding similarity of the recommendable individual project is: a similarity between the recommendable individual project and donated individual projects similar to the recommendable individual project;

the determining, from the at least one candidate user, at least one target user to which the public welfare item needs to be pushed includes:

for each candidate user, determining the maximum similarity from the similarities corresponding to target individual project contained in the individual project list of the candidate user, and determining the maximum similarity as the score of the candidate user;

and determining at least one target user required to be recommended to the public welfare items from the at least one candidate user based on the grade of the at least one candidate user from high to low.

In another possible implementation manner, before the obtaining the tweet description text of the public welfare item to be recommended, the method further includes:

acquiring a project recommendation request submitted by a service terminal, wherein the project recommendation request carries a tweet description text of a public welfare item to be recommended or a tweet link address of the public welfare item to be recommended;

the method for obtaining the tweet description text of the public welfare items to be recommended comprises the following steps:

obtaining a recommendation description text carried in the item recommendation request;

or,

and obtaining a tweet description text of the public welfare item based on the tweet link address.

In yet another possible implementation manner, the item recommendation request further indicates that: the target number of similar items and the target number of recommended users;

the determining a preset number of target individual project of which the similarity between the text vector and the tweet vector is higher based on the text vector of the individual project in the business service platform comprises:

determining a front target number of target individual project with higher similarity between a text vector and the tweet vector based on the text vector of a plurality of individual project in the business service platform;

if the difference value between the number of the at least one candidate user and the target number of people exceeds a set threshold value, adjusting the target number, and returning to execute the operation of determining the target number of project of the front target number with higher similarity between the text vector and the tweet vector based on the adjusted target number;

and if the difference value between the number of the at least one candidate user and the target number of the users does not exceed the set threshold value, determining that the public service items need to be pushed to the target users from the at least one candidate user in combination with the target number of the users.

In yet another possible implementation manner, the item link request further indicates a public number for publishing the public service item, where the public number belongs to public numbers under the business service platform;

the determining, from the plurality of users, at least one candidate user including the target individual project in the corresponding individual project list based on the individual project lists corresponding to the plurality of users in the business service platform includes:

and determining at least one candidate user which contains the target individual project and pays attention to the public number in the corresponding individual project list from the plurality of users based on the individual project list corresponding to the plurality of users in the business service platform.

In yet another possible implementation manner, after the determining at least one target user to which the public welfare item needs to be pushed, the method further includes:

sending a push notification to a push service program, wherein the push notification is used for indicating that the public service item is pushed to the at least one target user.

In another aspect, the present application further provides an apparatus for determining a push user of a public welfare project, which is applied to a service platform, and includes:

the public welfare information obtaining unit is used for obtaining a tweet description text of a public welfare item to be recommended;

the tweet vector determining unit is used for determining a tweet vector of the tweet description text;

the similar item determining unit is used for determining a preset number of target project blocks with higher similarity between a text vector and the tweet vector based on the text vector of a plurality of project blocks in the business service platform, wherein the project blocks are crowd-funded projects initiated by individuals, and the text vector of each project block is a vector generated based on the project description text of the project block;

the candidate user determining unit is used for determining at least one candidate user containing the target multiple project from the multiple users on the basis of multiple project lists corresponding to the multiple users in the business service platform, wherein the multiple project lists corresponding to the users comprise at least one recommendable multiple project interested by the users;

and the target user determining unit is used for determining at least one target user to which the public welfare item needs to be pushed from the at least one candidate user.

In one possible implementation, the apparatus further includes:

the list generating unit is used for generating the project list corresponding to the user by the following modes:

for each donated project contributed by the user, determining a similar project list of the donated projects, wherein the similar project list of the donated projects comprises: at least one similar item of the plurality of items having historical transaction details similar to the historical transaction details of the donated item and a similarity of the similar item to the item;

According to the scheme, after the text description of the tweet of the public service item to be recommended is determined, at least one target individual project with the project description text similar to the text description of the tweet is determined. If the target individual project is included in the individual project list which is interested by the user, the target individual project is interested by the user, and the target individual project is similar to the public service project, the public service project is interested by the user. On the basis, the target user for pushing the public service item is determined based on the users including the target individual project in the individual project list, so that the users interested in the public service item can be determined as a pushing user group of the public service item, and the opening rate of the text description of the public service item and the conversion rate of the public service item are naturally 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 schematic flowchart of a method for determining a public welfare item pushing user according to an embodiment of the present application;

FIG. 2 is a flow chart illustrating an implementation of determining a list of project stages of interest to a user according to an embodiment of the present disclosure;

fig. 3 is a schematic flowchart of another method for determining a public welfare item pushing user according to an embodiment of the present application;

fig. 4 is a schematic structural diagram of an apparatus for determining a public welfare item pushing user according to an embodiment of the present disclosure.

The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings described above, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It should be understood that the data so used may be interchanged under appropriate circumstances such that embodiments of the application described herein may be practiced otherwise than as specifically illustrated.

Detailed Description

The scheme of the application can be applied to a business service platform, and the business service platform can be a platform for providing related services of individual project and determining push user groups of individual project and public welfare project. For example, the business service platform is a crowd funding service platform.

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 inventive step, are within the scope of the present disclosure.

As shown in fig. 1, which shows a flowchart of a method for determining a push user of a public welfare item provided by the present application, the method of the present embodiment is applied to a business service platform, and includes:

and S101, acquiring a tweet description text of the public welfare item to be recommended.

The public welfare project is an investment project which aims at seeking social effects and generally has the characteristics of large scale, more investment, wide benefit range, long service life, profound influence and the like. In the present application, a public interest item may refer to a public interest item related to a fund class.

Unlike individual financing items which are only financed for individuals, the equity items are generally items initiated by the non-assisted group and larger assisted groups.

For example, a public interest item is typically some public interest item that is initiated by an authoritative individual fund through a fund platform or the like. For example, the commonweal items may be book donation commonweal items for mountain children to read books, and the like, as well as a desk plan for donating desks for rural schools, and the like.

The text description of the public welfare project refers to the text content for describing the specific plan and content of the public welfare project, and the specific content of the public welfare project can be understood through the text description of the public welfare project.

In one implementation, the tweet description of the public welfare item may be text introduced to the item content.

For example, taking the public interest item "desk plan" as an example, the "desk plan" is the item name of the public interest item, there are also the starting time and the ending time of the public interest item, and there are also some description texts for the "desk plan", for example: the novel school desk is a novel school desk, and makes efforts to enable children to have a novel school desk, wherein the school desk has uneven small holes and even defects on the surface, which influence the desire of the children to write.

In yet another implementation, the tweet description of the public interest item may also be a text that is spliced by text content of portions of the public interest item. For example, after the pictures and the special characters related to the item content of the public service item are removed, the text obtained by splicing the remaining text content is the text description content.

The specific implementation manner of triggering and determining the public welfare items to be recommended and specifically obtaining the tweet description text of the public welfare items can be various, and the application is not limited.

In a possible implementation manner, the method and the system can initiate an item recommendation request to a service platform through a service terminal, wherein the item recommendation request is used for indicating the public welfare items to be recommended, and the item recommendation request can request the service platform to determine that the public welfare items are suitable for a user group to be recommended to.

Specifically, the service platform may obtain an item recommendation request submitted by the service terminal, where the item recommendation request carries a tweet description text of the public welfare item to be recommended, or a tweet link address of the public welfare item to be recommended. The text pushing link address can be a link address for positioning text of text pushing description of the public welfare item. For example, the tweet link address may be a link address of the item content publishing the public welfare item, or the like

Correspondingly, if the item recommendation request carries the tweet description text of the public service item, the recommendation description text can be directly obtained from the item recommendation request.

If the item recommendation request carries the text pushing link address of the public welfare item to be recommended, the text pushing description text of the public welfare item can be obtained according to the text pushing link address. For example, the tweet description text of the public welfare item is extracted from the page pointed to by the tweet connection address.

S102, determining a tweet vector of the tweet description text.

Wherein the tweet vector refers to a vector representation of the tweet description text.

There are many possible specific ways to determine the vector of the tweet description text, which is not limited in this application.

In a possible implementation manner, the word segmentation may be performed on the tweet description text to obtain at least one word segmentation of the tweet description text. Then, word vectors of all the participles of the text are determined respectively, or a keyword is determined from at least one participle of the text and then the word vector of the keyword is determined; and finally, combining word vectors of all the participles or keywords in the tweet description text into a vector of the tweet description text, namely the tweet vector.

The method for segmenting the text of the tweet description may be various, for example, segmenting the text content by using a Chinese segmentation algorithm or other segmentation algorithms, and extracting the keywords of the text of the tweet description by using the function of extracting the keywords of the Chinese segmentation algorithm.

For example, the ending participle supports four participle modes:

1. the accurate mode is used for trying to cut the sentence most accurately, and is suitable for text analysis;

2. in the full mode, all words which can be formed into words in a sentence are scanned, so that the speed is very high, but ambiguity cannot be solved;

3. and the search engine mode is used for segmenting long words again on the basis of the accurate mode, so that the recall rate is improved, and the search engine mode is suitable for word segmentation of the search engine.

4. And in the paddlemode, a paddlepaddledeep learning framework is utilized, and a sequence annotation (bidirectional GRU) network model is trained to realize word segmentation.

In the application, in order to more accurately split the sentence into corresponding atomic words and not generate redundant words, an accurate mode can be adopted for word segmentation, a user-defined dictionary of a crowd funding industry is added, stop words are manually specified, and the first 50 keywords of the text description text are extracted by using the crust TextRank algorithm.

The basic idea of TextRank may include the following:

1. segmenting the text of the keywords to be extracted;

2. constructing a graph according to the co-occurrence relationship between words and the fixed window size (default is 5, and the adjustment is carried out through span attributes);

3. PageRank of nodes in the graph is calculated, noting that the graph is a undirected weighted graph.

Of course, the above description is given by taking the ending word segmentation as an example, and the way of segmenting the textual description text by adopting other word segmentation ways is also applicable to the embodiment, and is not described herein again.

The vector for determining the word segmentation or the keyword can be obtained by combining a word vector library. The word vector library used in the present application may be some word vector libraries that are currently used and trained. In order to better describe a public service project or a subsequent individual project, a word vector library according to which word vectors of participles or keywords are determined in the application may be: and training by combining the corpora in the crowd funding field to obtain a word vector library. For example, Word vector libraries which are more close to the public welfare and crowd funding fields related to the application are trained by using Word2Vec, Doc2Vec, Bert and the like.

The method of aggregating the word vectors of the participles or keywords may also be multiple possibilities, and may be specifically set according to actual needs, which is not limited.

S103, determining a preset number of target project files with high similarity between the text vector and the tweet vector based on the text vectors of the project files in the business service platform.

It can be understood that the business service platform may provide services for a plurality of project stages, and therefore, the business service platform may store information related to a plurality of project stages.

Wherein, each project is a crowd-funding project initiated by an individual. In an alternative mode, the individual research projects in the application can be large illness rescue projects such as large illness crowd research initiated by individuals.

In the present application, the individual chip items may be individual chip items which are being processed and completed.

The text vector of the respective staged project is a vector generated based on the project description text of the respective staged project. The text vectors of the multiple project are predetermined and stored in the business service platform.

The item description text of each project is extracted from item detail data recorded in the business service platform, for example, the item detail data may record information such as a project name, a project content, a patient age, a disease, and a hospital. Of course, if some data are not normal in the project detail data of each project, the business system can correct the data. For example, assuming that the gender of the patient for each project is not filled, it may be determined from the "son", "daughter", or "mom" etc. appearing in the contents of the project.

It is understood that the process of determining the item description text based on the item record details of the respective project is similar to the process of determining the tweet description text, and will not be described herein again.

The process of generating the text vector of each project is similar to the process of generating the tweet vector of the tweet description text, and is not repeated here.

The similarity between the text vector and the tweet vector may adopt any vector similarity calculation method, which is not limited in this respect.

The set quantity of the selected target project can be set according to needs, such as the setting of a business service platform, and the setting of the set quantity can also be set when a business terminal initiates a project recommendation request.

For example, a plurality of research projects may be sorted in an order from high similarity to low similarity between a text vector of each research project and a tweet vector, and a set number of research projects ranked in the top may be selected as target research projects; or, all the pre-set number (for example, the target number) of stage items of which the similarity between the text vector and the tweet vector exceeds a set threshold are determined as the target stage items. S104, determining at least one candidate user containing a target individual project from the corresponding individual project lists based on the individual project lists corresponding to the multiple users in the business service platform.

The individual project list corresponding to the user comprises at least one recommendable individual project which is interested by the user. The recommendable individual project which is interested by the user belongs to the plurality of individual projects in the service platform, and for convenience of distinguishing, the individual projects contained in the individual project list of the user are called recommendable individual projects.

The method for determining the recommendable project in which the user is interested may have various possibilities, and is not limited to this. The following description will be made in conjunction with the possible case of determining the individual project listings in which the user is interested, and will not be described herein again.

For convenience of distinction, a user who includes a target individual project in the corresponding individual project list is referred to as a candidate user. It is understood that the project research list of the candidate user may include only one target project research, or may include a plurality of target project research, without limitation.

It can be understood that a target chip item is included in the chip item list of the candidate user, which indicates that the target chip item is a chip item of interest to the candidate user. Since the text vector of the target individual research item is similar to the tweet vector of the public service item to be recommended, that is, the target individual research item and the public service item have the same characteristics, the public service item is also an item that the candidate user is interested in under the condition that the candidate user is interested in the target individual research item.

It is understood that in practical applications, public service items may be selected to be published on a public number under the business service platform. The item link request of the service terminal obtained in step S101 may further indicate that there is a public number issuing the public welfare item, where the public number belongs to a public number under the service platform. Correspondingly, based on the respective project planning lists corresponding to the multiple users in the business service platform, at least one candidate user which contains the target project planning and pays attention to the public account in the corresponding project planning list can be determined from the multiple users in the business service platform.

And S105, determining at least one target user to which the public welfare item needs to be pushed from at least one candidate user.

For example, all candidate users may be determined as target users to which the public welfare project needs to be pushed. Alternatively, a part (for example, a number of persons) of the candidate users is selected as the target user from among the at least one candidate user.

In one possible implementation manner, the respective chip item list corresponding to the user includes: at least one recommendable individual project similar to the at least one donated individual project contributed by the user and a corresponding similarity for each recommendable individual project.

In the application, the similarity between each chip item (which may recommend each chip item) and each donated chip item may be that the similarity between each chip item and each donated chip item exceeds a set threshold; or, based on the similarity between different individual project and the donated individual project, sorting from high to low according to the similarity, wherein the similarity between the individual project and the donated individual project belongs to the top N similarities of the top ranking, and the value of N can be set as required. Of course, there may be other possibilities for determining whether each of the donated project is a recommendable project similar to the donated project based on the similarity between the donated project and the donated project, which is not limited herein.

Wherein, the similarity corresponding to each recommendable research item is: a similarity between the recommendable individual project and donated individual projects similar to the recommendable individual project. For example, if the recommendable individual piece item is similar to the user-contributed individual piece item a, the recommendable individual piece item may be represented by a similarity between the recommendable individual piece item and the user-contributed individual piece item a.

In particular, if a certain recommendable research item is similar to a plurality of research items contributed by the user, the maximum similarity between the recommendable research item and the plurality of research items may be determined as the similarity corresponding to the recommendable research item. For example, the individual project a may be recommended to be similar to the individual project S1 that the user has contributed, and the similarity is 0.8; and the similarity between the recommendable individual project a and the individual project S2 contributed by the user is 0.9, it may be determined that the similarity corresponding to the recommendable individual project a is 0.9.

Of course, the average of the similarity between the recommendable individual research items and the plurality of donated individual research items contributed by the user may be determined as the similarity corresponding to the recommendable individual research items.

Correspondingly, for each candidate user, the maximum similarity can be determined from the similarities corresponding to the target project chips included in the project chip list of the candidate user, and the maximum similarity is determined as the score of the candidate user. Then, at least one target user required to be recommended for the public welfare items is determined from the at least one candidate user based on the grade of the at least one candidate user from high to low. For example, a set number of target users ranked in the top are selected from at least one candidate user in the order of the scores of the candidate users from high to low. Or, all the at least one candidate user is determined as the target user, and the ranking of the at least one candidate user is taken as the recommended ranking of the target user.

It can be understood that the historical data of the public welfare items is less, and the difference between different public welfare items is larger, so that it is difficult to determine the interested user group of the public welfare items to be recommended based on the participating users of the historical public welfare items and the data of the historical public welfare items. Based on the method and the system, the data volume of the individual project in the business service platform is large and rich, and the user interested in the public service project is determined by innovatively combining the individual project and the participating users.

As an optional manner, after determining the target user, the application may further send a push notification to the push service program, where the push notification is used to instruct to push the public interest item to at least one target user.

The push service program may be a program on a specific push service platform, or may be a program on the service platform, which is not limited in the present application. The purpose of sending the push notification to the push service program is to make the push service program know that the public service item is suitable for being pushed to the target user,

As can be seen from the above, after determining the tweet description text of the public service item to be recommended, at least one target individual research item whose item description text is similar to the tweet description text is determined. If the target individual project is included in the individual project list which is interested by the user, the target individual project is interested by the user, and the target individual project is similar to the public service project, the public service project is interested by the user. On the basis, the target user for pushing the public service item is determined based on the users including the target individual project in the individual project list, so that the users interested in the public service item can be determined as a pushing user group of the public service item, and the opening rate of the text description of the public service item and the conversion rate of the public service item are naturally improved.

In order to facilitate understanding, in the present application, an implementation process of determining a chip item list of a user is determined, and a possible implementation manner is combined to introduce generation of a chip item list in which the user is interested in the present application.

As shown in fig. 2, which shows a flowchart of an implementation manner of generating a project list of individual research projects of a user provided in the present application, the method of this embodiment may include:

s201, obtaining historical transaction details of each project in the business service platform.

Wherein the historical transaction details for the individual project include at least user information contributing to the individual project.

The historical transaction details for the respective project include details of transaction records generated by the respective project. Each transaction record may record a user identification (e.g., a user name or user ID) of the user (i.e., contributing to the respective project) and an identification of the respective project for each transaction of the respective project. Of course, the transaction materials and the transaction time can be recorded.

For example, taking individual project of big disease rescue as an example, each transaction record detail of the individual project obtains a transaction record for the individual project, and thus, the transaction record detail may include a user identifier and a project identifier for contributing money to the individual project. And may also include the amount of the donation, the time of the donation, etc.

S202, determining at least one donated project contributed by the user based on the historical transaction details of the project.

For the sake of convenience of distinction, the individual stage items contributed by the user are referred to as donated individual stage items. The donated project belongs to a plurality of project of the business service platform.

It can be understood that, according to the historical transaction details of a plurality of works in the business service platform, the project of the information of the user contained in the historical transaction details can be determined, and the project is the donated project contributed by the user.

S203, aiming at each donated project contributed by the user, determining a similar project list of the donated projects.

Wherein, the similar item list of donated project includes: at least one similar item of the plurality of items, the historical transaction details of which are similar to the historical transaction details of the donated item, and the similarity of the similar item to the item.

There are many possible specific implementations of determining a list of similar items that have contributed to a project.

For example, in one possible implementation, the application may determine, for each donated project of the user, a similarity between the donated project and each project based on historical transaction details of the donated projects and historical transaction details of the donated projects. Then, based on the similarity of the donated project to each project, at least one recommendable project similar to the donated project is determined from a plurality of projects, a similar project list corresponding to the donated project is generated, and the similar project list comprises the at least one recommendable project similar to the donated project.

In practice, for each user, when a similar item list of each donated project needs to be determined, the similarity between the donated project and other project is calculated in real time. In practical applications, however, the donated project of different users may be repeated, and the data may be calculated in a large amount by using the above possible manner.

In yet another possible implementation manner, the application may determine a similar item list of each contributed item in the business service platform before analyzing the contributed items of the user. Specifically, determining the similar item list of each chip item may include the following steps:

(1) respectively determining the similarity between two project planning based on the respective historical transaction details of the project planning;

(2) the method comprises the steps of determining at least one similar individual project similar to each project from the plurality of projects according to the similarity of the project to other projects, generating a similar project list of the project, wherein the similar project list of the project comprises at least one similar individual project similar to the project, and the similarity of each similar individual project to the project.

Accordingly, in step S203, a similar item list corresponding to the donated project may be determined from the similar item list of the multiple project.

In the above two manners of the present application, for any two project planning, there are many possibilities for the method of calculating the similarity between the two project planning based on the historical transaction details of the two project planning, which is not limited by the present application.

For example, in one possible implementation, the present application may employ a Swing algorithm based on a graph structure. Swing refers to a Swing, and in this algorithm, it is considered that there will be many such swings in a bipartite graph of users and items, e.g., (u, v, i1), i.e., users u and v have both purchased item i1, which make up a Swing (the triangle lacks one side). This is actually a 3 rd order interaction. Traditional heuristic neighbor methods only focus on second-order interactions between users and items. Swing will focus on this 3-order relationship. One intuition of this approach stems from: if multiple users click on i1 and share only one other i2, i1 and i2 must be strongly correlated, and the unknown strong correlation is communicated through the users. On the other hand, if more Swing structures are formed between two user pairs, each structure is weaker, and each node is assigned a lower weight on this user pair. The formula for Swing is as follows:

in the context of the present application, it is,

representing individual project

And an individual project

The similarity of (c).

Representing individual project

Is provided (or is to donate individual research projects) to a user's set of users included in the historical transaction specification

The set of users of),

for an individual project

The historical transaction details of (1) is included in the user's collection of users.

Wherein,

including users for historical transaction details

Of individual project, i.e. users

A collection of donated individual project;

including users for historical transaction details

Of individual project, i.e. users

A collection of donated individual project blocks. Accordingly, the method can be used for solving the problems that,

refers to the user

And the user

Number of individual donated items.

Wherein,

is a fixed coefficient with a value as a set value.

It is understood that although the example of swing calculating the similarity of two project stages is described here, in practical application, the calculation of the similarity of the project stages by combining the historical transaction details of the project stages and other algorithms is also applicable to this embodiment.

In the present application, the similar individual project similar to the individual project refers to another individual project whose similarity with the individual project exceeds a set threshold, or may be a previously set individual project whose similarity with the individual project is high, and may be specifically set as needed, which is not limited herein.

S204, generating a project item list of the user based on the similar project list of the donated project items.

Recommendable ones of the research items in the research item list belong to a similar item list of the donated research items.

For example, each similar project in the similar project list of each donated project may be aggregated, so as to obtain the individual project list of interest to the user. It will be appreciated that if there are similar item duplicates in a similar item list of different donated items, duplicate item duplicates need to be removed.

In the application, the individual donated items in the similar item list of the individual donated items can be removed, and the remaining similar items in the similar item list of the individual donated items can be used as recommendable individual items in the individual project list of the user.

For example, after the user u1 donates the respective chip items i1 and i9, after obtaining the similar item lists L1 and L9 of the respective chip items i1 and i9, the similar chip items in L1 and L9 may be deduplicated and the chip items i1 and i9 that the user u1 has donated may be rejected, and the remaining similar chip items are the recommendable chip items in the chip item list Lu1 of the user u 1.

In this application, the individual project list of the user may include not only recommendable individual projects, but also: a degree of similarity between the individual project and the donated individual project for which the individual project is similar to the recommendable individual project may be recommended.

In the above embodiments of the present application, the number of the target users who determine to push the public service item may be set by the service platform. In order to enable each service terminal to set the number of target users who wish to push public welfare projects according to needs, the application can also be used for inputting target data of the desired target users by the terminal of the service terminal. Accordingly, the target number of users to be recommended to the public service item may also be indicated in the item recommendation request.

Particularly, the business terminal in the application can also set the target number of target individual project similar to the public welfare project to be matched according to the requirement. Accordingly, the item recommendation request further indicates that: a target number of similar items and a target number of recommended users. On this basis, in the above embodiment of the application, according to the target number, the target number of the target project stages with the text vector and the text-pushing vector of the public welfare project having higher similarity can be determined. Similarly, it is finally determined that a plurality of target users of the target person are used as a group of push users for pushing the public welfare project.

In the following, the method for determining the push users of the public welfare items according to the present application is described by taking the example that the service terminal side sets the target number of target users, and combining a possible situation.

As shown in fig. 3, which shows a flowchart of another implementation of the method for determining a push user of a public welfare item according to the present application, the method of this embodiment may include:

s301, acquiring a project recommendation request submitted by a service terminal.

The item recommendation request carries the tweet link address of the public service item to be recommended.

The item recommendation request indicates a target number of similar items, a target number of recommended users, and a public number to publish a public welfare item. For example, the item recommendation request carries an identifier of the public number that issued the public welfare item or needs to issue the public welfare item. The public number belongs to the public number under the business service platform.

It is understood that the business service platform may develop an interactive page, such as a page based on H5 or an interactive page presented by an applet, etc. After the business terminal logs in the business service platform, the business terminal can present the interactive page.

If the public service item to be recommended is not released, a tweet description text of the public service item to be recommended can be input in a text box of the interactive page; if the public welfare item to be recommended is published, a tweet link address of the public welfare item to be recommended can be input. In the present embodiment, for convenience of understanding and description, a service terminal submits a tweet link address of a public service item to be recommended as an example.

The interactive page may further have a public number pull-down menu, and the public number pull-down menu may include a plurality of public numbers under the service platform, on which basis, a user of the service terminal may select a desired public number in the pull-down menu.

Similarly, an input box for setting the number of similar items and an input box for setting the number of the users to push can be provided in the interactive page, and on the basis, the user at the service terminal side can input the target number for limiting the target individual project similar to the public service project and the target number for limiting the determined target user.

After the user of the service terminal finishes the input or selection of the related information on the interactive page, the user can click the designated button to trigger the sending of the item recommendation request.

Of course, the above is an example of one possible case of the interactive page, and the method of sending the item recommendation request by the service terminal in other manners is also applicable to the embodiment.

S302, based on the text pushing link address, a text pushing description of the public welfare item is obtained.

It can be understood that, for the case that the recommendation description text of the public service item is carried in the item recommendation request, the recommendation description text can be directly obtained from the item recommendation request, and details are not repeated here.

S303, determining a tweet vector of the tweet description text.

S304, determining a front target number of target project files with high similarity between the text vector and the tweet vector based on the text vectors of the project files in the business service platform.

In this embodiment, since the service terminal side sets the target number of the target individual project to be selected, the step directly selects the target individual project with the front target number, which has a higher similarity between the text vector and the tweet vector, according to the target number.

The similarity between the text vector of each research item and the tweet vector of the public welfare item can be obtained by calculating the cosine distance or the Euclidean distance between the two vectors, which is not limited to this.

As an alternative, in consideration of calculating the similarity of each text vector and the tweet vector, the vector calculation of a high-dimensional space belonging to a large data volume is large in calculation amount and may take a long time. Based on the method, the target project search can be realized by means of open-source vector search libraries.

For example, real-time computation of massive data vectors in a high-dimensional space can be achieved using the open source library Faiss. The Faiss library provides an efficient and reliable vector retrieval method and also reduces space usage. Correspondingly, the Faiss vector search service can be deployed on a business service platform, and the text vector of each project is imported into the Faiss. And the text vector with the newly added research projects is also imported into the Faiss in time so as to ensure that all research projects can be taken as the recalled objects.

S305, based on the project planning lists corresponding to the users in the business service platform, at least one candidate user which contains the target project planning list and pays attention to the public number in the corresponding project planning list is determined from the users.

Wherein, the individual project list that the user corresponds includes: at least one recommendable individual project similar to the at least one donated individual project contributed by the user and a corresponding similarity for each recommendable individual project. The similarity corresponding to each recommended research item is as follows: a similarity between the recommendable individual project and a donated individual project similar to the recommendable individual project.

S306, aiming at each candidate user, determining the maximum similarity from the similarities corresponding to all target project chips contained in the project chip list of the candidate user, and determining the maximum similarity as the score of the candidate user.

For example, the individual raising similarity list of the candidate user includes a target individual raising item a and a target individual raising item B, where the similarity corresponding to the target individual raising item a is 0.95, and the similarity corresponding to the target individual raising item B is 0.8, so that the maximum similarity is 0.95, and the score of the candidate user may be 0.95.

S307, determining target users to be recommended to the public welfare project from the at least one candidate user based on the sequence of the scores of the at least one candidate user from high to low.

It will be appreciated that, in general, the number of candidate users determined will be relatively large, and the number of users to whom the public welfare item is expected to be pushed is only a fraction of the candidate users. On the basis, considering that the scores of the candidate users can reflect the correlation degree of the individual project favored by the candidate users and the public welfare project, the application can select the target users with higher scores as the target users needing to push the public welfare project according to the scores of the candidate users, so that the pushing effect of the public welfare project is further improved.

It is understood that if the number of candidate users is less than the number of target persons, all candidate users may be the target users.

Particularly, when the number of the candidate users is excessive or insufficient, the value of the target number matched with the target project can be properly adjusted.

If the difference value between the number of the at least one candidate user and the target number of users exceeds a set threshold value, the target number is adjusted, and the operation of determining the target items of the front target number with higher similarity between the text vector and the tweet vector is returned to be executed based on the adjusted target number, so that the number of the matched target items is changed, and the number of the matched candidate users is further changed.

Correspondingly, if the difference value between the number of the at least one candidate user and the target number of the users does not exceed the set threshold value, the number of the target users is combined, and the fact that the public service items need to be pushed to the target users is determined from the at least one candidate user. For example, if the number of candidate users exceeds the number of target users, the target users with higher scores can be selected from at least one candidate user according to the scheme of the application and the scores of the candidate users. And if the number of the candidate users does not exceed the target number, all the candidate users are taken as the target users.

For example, assume that the business terminal sets a target number of people N =100w, and a target number of similar target-level items S = 100. If the number of candidate users is determined to be 200w based on the target number S, S may be halved. And then, recalculating the number of the candidate users finally determined when S =50, and if the number of the candidate users is still more than 100w, continuing to reduce the value of S for recalculation until the number of the candidate users approaches to N.

And if the number of the candidate users is less than 100w, increasing the S value for recalculation until the number of the candidate users approaches the N value.

On the basis, if the number of the candidate persons is larger than N, the target user can be finally determined according to the scores of the candidate users.

In practical application, the service platform may further send the adjusted target number S 'that makes the number of candidate users approach N to the service terminal, so that the service terminal determines whether to filter the candidate users based on the adjusted target number S'. And if the service terminal confirms that the candidate users are screened based on the adjusted target quantity S ', the candidate users screened according to the adjusted target quantity S' are used as final candidate users. If the service terminal still wants to keep the originally set target number unchanged, the target user can be finally determined by the candidate users screened out based on the original target number.

The application also provides a push user device for determining the public welfare items.

Fig. 4 is a schematic structural diagram illustrating a component of an apparatus for determining a push user of a public welfare item, which is applied to a business service platform, and includes:

a public welfare information obtaining unit 401, configured to obtain a tweet description text of a public welfare item to be recommended;

a tweet vector determining unit 402, configured to determine a tweet vector of the tweet description text;

a similar item determining unit 403, configured to determine, based on a text vector of multiple project funded in the business service platform, a preset number of target project funded items with a higher similarity between the text vector and the tweet vector, where the project funded items are crowd funded items initiated by individuals, and the text vector of each project is a vector generated based on an item description text of the project funded item;

a candidate user determining unit 404, configured to determine, based on respective chip lists corresponding to multiple users in the business service platform, at least one candidate user including the target chip from the multiple users, where the respective chip lists corresponding to the users include at least one recommendable chip that is of interest to the users;

a target user determining unit 405, configured to determine, from the at least one candidate user, at least one target user to which the public welfare item needs to be pushed.

In one possible implementation, the apparatus may further include:

the list generating unit is used for generating a project list corresponding to the user by the following modes:

determining that a user contributed at least one contributed project based on historical transaction details for respective project, the contributed project belonging to the plurality of project;

for each donated staged project contributed by a user, determining a similar item list of the donated staged project from the plurality of staged projects, the similar item list of the donated staged projects including: at least one similar item of the plurality of items having historical transaction details similar to the historical transaction details of the donated item and a similarity of the similar item to the item;

In an alternative, the apparatus further comprises:

a similarity calculation unit configured to determine, before the list generation unit determines the similar item list of the donated project, a similarity between two project pieces based on respective historical transaction details of the project pieces;

the similarity item determining unit is used for determining at least one similarity individual project similar to the individual project from the plurality of individual projects according to the similarity of the individual project and other individual projects, and generating a similarity item list of the individual projects, wherein the similarity item list of the individual projects comprises at least one similarity individual project similar to the individual project and the similarity of the similarity individual project and the individual project;

the list generating unit is specifically configured to determine the similar item list of the donated project from the similar item lists of the multiple project when determining the similar item list of the donated project.

In one possible implementation manner, the respective chip item list corresponding to the user includes: at least one recommendable individual project similar to the at least one donated individual project contributed by the user and the corresponding similarity of each recommendable individual project, wherein the corresponding similarity of the recommendable individual project is: a similarity between the recommendable individual project and donated individual projects similar to the recommendable individual project;

a target user determination unit comprising:

the user scoring unit is used for determining the maximum similarity from the similarities corresponding to the target individual project contained in the individual project list of the candidate user aiming at each candidate user and determining the maximum similarity as the score of the candidate user;

and the user screening unit is used for determining at least one target user required to be recommended to the public welfare items from the at least one candidate user based on the order of the scores of the at least one candidate user from high to low.

In one possible implementation, the apparatus further includes:

the system comprises a request obtaining unit, a service terminal and a service information obtaining unit, wherein the request obtaining unit is used for obtaining an item recommendation request submitted by the service terminal before a public welfare information obtaining unit obtains a tweet description text of a public welfare item to be recommended, and the item recommendation request carries the tweet description text of the public welfare item to be recommended or a tweet link address of the public welfare item to be recommended;

public welfare information obtaining unit, including:

the text acquisition unit is used for acquiring a recommendation description text carried in the item recommendation request;

or,

and the text capturing unit is used for obtaining the text description text of the public welfare item based on the text link address.

In one possible implementation, the item recommendation request obtained by the request obtaining unit further indicates that: the target number of similar items and the target number of recommended users;

the similar item determining unit is specifically configured to determine, based on text vectors of multiple project files in the business service platform, a target number of target project files with higher similarity between the text vectors and the tweet vectors;

a target user determination unit comprising:

the adjusting unit is used for adjusting the target number if the difference value between the number of the at least one candidate user and the target number exceeds a set threshold value, and returning to execute the operation of determining the target number of the target project with high similarity between the text vector and the text pushing vector based on the adjusted target number;

and the target confirmation unit is used for determining that the public interest item needs to be pushed to the target user from the at least one candidate user in combination with the target number if the difference between the number of the at least one candidate user and the target number does not exceed the set threshold.

In another possible implementation manner, the item link request obtained by the request obtaining unit further indicates a public number for publishing the public welfare item, where the public number belongs to a public number under the service platform;

the candidate user determination unit is specifically configured to determine, from among the multiple users, at least one candidate user that includes the target individual project and pays attention to the public account in the corresponding individual project list based on the individual project lists corresponding to the multiple users in the business service platform.

In yet another possible implementation manner, the apparatus further includes:

and the push notification unit is used for sending a push notification to a push service program after the target user determination unit determines at least one target user to which the public interest item needs to be pushed, wherein the push notification is used for indicating to push the public interest item to the at least one target user.

It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other. Meanwhile, the features described in the embodiments of the present specification may be replaced or combined with each other, so that those skilled in the art can implement or use the present application. For the device-like embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining a push user of a public service item is applied to a business service platform and comprises the following steps:

acquiring a tweet description text of a public welfare item to be recommended;

determining a tweet vector of the tweet description text;

determining at least one candidate user containing the target individual project from the multiple users on the basis of individual project lists corresponding to the multiple users in the business service platform, wherein the individual project lists corresponding to the users comprise at least one recommendable individual project which is interested in the users;

2. The method of claim 1, wherein the user's respective list of staged items is obtained by:

determining at least one donated project contributed by the user based on historical transaction details of respective project, the donated project belonging to the plurality of project;

3. The method of claim 2, further comprising, prior to said determining said list of similar items to which said staged item has been donated:

the determining the list of similar items of the donated staged item includes:

4. The method of any one of claims 1 to 3, wherein the user's respective list of staged items comprises: at least one recommendable individual project similar to the at least one donated individual project contributed by the user and the corresponding similarity of each recommendable individual project, wherein the corresponding similarity of the recommendable individual project is: a similarity between the recommendable individual project and donated individual projects similar to the recommendable individual project;

5. The method of claim 1, prior to obtaining the tweet description text of the public welfare item to be recommended, further comprising:

or,

6. The method of claim 5, wherein the item recommendation request further indicates that there is: the target number of similar items and the target number of recommended users;

7. The method according to claim 5 or 6, wherein the item linking request further indicates a public number for publishing the public service item, and the public number belongs to a public number under the business service platform;

8. The method of claim 1, further comprising, after the determining at least one target user to which the public interest item needs to be pushed:

9. An apparatus for determining a push user of a public service item, applied to a business service platform, includes:

the candidate user determining unit is used for determining at least one candidate user containing the target individual project from the multiple users on the basis of individual project lists corresponding to the multiple users in the business service platform, wherein the individual project lists corresponding to the users comprise at least one recommendable individual project which is interested by the users;

10. The apparatus of claim 9, further comprising: