CN113254633B

CN113254633B - Message document generation method and device

Info

Publication number: CN113254633B
Application number: CN202110514552.4A
Authority: CN
Inventors: 刘轶琳; 王静; 姚晓璐; 董鲁豫; 袁宇茜; 闫炳琪
Original assignee: China Minsheng Banking Corp Ltd
Current assignee: China Minsheng Banking Corp Ltd
Priority date: 2021-05-10
Filing date: 2021-05-10
Publication date: 2024-05-03
Anticipated expiration: 2041-05-10
Also published as: CN113254633A

Abstract

The application provides a message document generation method and a message document generation device. The method comprises the following steps: and the server determines a document creation rule corresponding to the scene theme according to the scene theme in the scene information, and acquires related phrases from the phrase library. The text creation rule comprises text elements, wherein the text elements are obtained by finely disassembling and decomposing the whole text. The phrases in the phrase library are obtained by disassembling and decomposing historical document data. And the server constructs and generates a message document according to the related phrase and document model. The document model comprises document preferences of different client groups, and personalized message documents can be generated for the different client groups. The server pushes the corresponding message file to the target device. The target device is a device used by clients in the target client group. The method of the application improves the attraction of the message file to the customer group and increases the click rate of the message file.

Description

Message document generation method and device

Technical Field

The present application relates to the field of computers, and in particular, to a method and apparatus for generating a message document.

Background

In the era of mobile interconnection, self-service channels (APP, weChat public number, short message) have become the most important tie for enterprises to connect customers with products and services. The enterprise can guide clients to the APP, the applet or the desktop application through push messages to participate in activities or transact business. After the client views the information pushed by the enterprise in the self-service channel, the client can participate in the activity or transact business by clicking the information.

In the prior art, after determining the activity content or business content, the enterprise needs to write the message document first. The content of the message document is directly related to the active content or business content. When the message file is determined, the message file is sent to the client. At present, the message document is mostly written by staff according to activity content or business content.

However, this message document has problems such as poor attraction and low click rate.

Disclosure of Invention

The application provides a method and a device for generating a message document, which are used for solving the problems of poor attraction, low click rate and the like of the message document in the prior art.

In a first aspect, the present application provides a method for generating a message document, including:

Acquiring related phrases from a phrase library according to scene information and scene document creation rules, wherein the scene information comprises scene topics, and the phrases in the phrase library are obtained by decomposing historical document data;

Constructing and generating a message document according to the related phrase and a document model, wherein the document model comprises document preferences of different client groups, and the document preferences are used for indicating the preferences of different client groups on different expression modes in the same document element;

And pushing the message file to target equipment, wherein the target equipment is equipment used by clients in a client group corresponding to the message file.

Optionally, the acquiring related phrases from the phrase library according to the scene information and the scene document creation rule includes:

determining a scene theme and the scene document creation rule corresponding to the scene theme according to the scene information, wherein the scene document creation rule comprises document elements;

acquiring related phrases matched with the scene subjects from a phrase library;

and matching the related phrases corresponding to the text elements from the related phrases according to the text elements included in the scene text creation rules.

Optionally, the text element is determined according to the action of the phrase in the message text, the phrase is obtained by disassembling and constructing the message text, and the text element comprises at least one of statement color, emotion style, personality title, word description, specific rights, action encouragement and presentation position.

Optionally, the method further comprises:

Acquiring historical document data, wherein the historical document data comprises a plurality of documents; dividing the text in the text data into phrases according to a preset dividing rule;

Marking the scene theme and the text element corresponding to each phrase according to the phrase, the preset scene theme and the preset text element;

and adding the marked phrases into the phrase library.

Optionally, the marking, according to a preset scene theme and a preset document element, the scene theme and the document element corresponding to each phrase includes:

determining a word shift distance matrix of the phrase according to the phrase;

determining the text elements to which each phrase belongs according to the word shift distance matrix;

clustering the phrases by using a clustering algorithm, and determining the scene subject to which each phrase belongs;

the phrases are labeled according to the document element and the scene topic of each of the phrases.

Optionally, the method further comprises:

and using a natural language processing algorithm to expand the phrase library by using a Chinese open source word library.

Optionally, the method further comprises:

generating a plurality of test cases according to the scene information, the scene case creation rule, the related phrases corresponding to the scene information and a preset screening algorithm, wherein the plurality of test cases comprise all the related phrases;

Sending the test documents to the test equipment of a customer group, wherein each test equipment uniquely corresponds to a test customer in the customer group;

counting the reflow data of each test file;

determining the document preference of the client group according to the reflow data and the client images of the test clients in the client group;

And generating a document model according to the document preference of different client groups, wherein the document model is used for generating corresponding message documents for different client groups.

Optionally, generating a plurality of test cases according to the scene information, the scene case creation rule, the related phrase corresponding to the scene information and a preset screening algorithm, including:

Generating a plurality of first test cases according to the scene information, the scene case creation rule and related phrases corresponding to the scene information, wherein the plurality of first test cases comprise all arrangement modes of the related phrases;

and screening to obtain a plurality of second test patterns according to the preset screening algorithm and the plurality of first test patterns, wherein the number of the second test patterns is smaller than that of the first test patterns.

Optionally, the sending all the test documents to the test devices of a customer group includes:

Dividing the test clients of the client group into preset number groups according to a grouping rule, wherein the preset number is related to the number of the test documents;

pushing corresponding test files to the test equipment of each grouping.

Optionally, the method further comprises:

obtaining reflow data of the message file of each client group, wherein the reflow data is determined according to the clicking condition of the message file corresponding to each client group;

And continuously and iteratively optimizing the document model according to the reflow data.

In a second aspect, the present application provides a message document generating apparatus, comprising:

The phrase acquisition module is used for acquiring related phrases from a phrase library according to scene information and scene document creation rules, wherein the scene information comprises scene topics, and the phrases in the phrase library are obtained by decomposing historical document data;

The document generation module is used for constructing and generating a message document according to the related phrase and a document model, wherein the document model comprises document preferences of different client groups, and the document preferences are used for indicating the preferences of different client groups on different expression modes in the same document element;

and the document pushing module is used for pushing the message document to target equipment, wherein the target equipment is equipment used by clients in a target client group corresponding to the message document.

Optionally, the phrase obtaining module is specifically configured to determine a scene theme and the scene document creation rule corresponding to the scene theme according to the scene information, where the scene document creation rule includes document element components; acquiring related phrases matched with the scene subjects from a phrase library; and matching the related phrases corresponding to the text elements from the related phrases according to the text elements included in the scene text creation rules.

Optionally, the apparatus further includes a phrase library generation module, the phrase library generation module including:

The acquisition sub-module is used for acquiring historical document data, wherein the historical document data comprises a plurality of documents;

The segmentation sub-module is used for segmenting the text in the text data into phrases according to a preset segmentation rule;

The marking sub-module is used for marking the scene theme and the text element corresponding to each phrase according to the phrases, the preset scene theme and the preset text element;

And the adding sub-module is used for adding the marked phrases into the phrase library.

Optionally, the marking submodule is specifically configured to determine a word shift distance matrix of the phrase according to the phrase; determining the text elements to which each phrase belongs according to the word shift distance matrix; clustering the phrases by using a clustering algorithm, and determining the scene subject to which each phrase belongs; the phrases are labeled according to the document element and the scene topic of each of the phrases.

Optionally, the phrase library generating module further includes:

and the expansion sub-module is used for expanding the phrase library by using a natural language processing algorithm and utilizing a Chinese open source word library.

Optionally, the apparatus further includes a model generation module, the model generation module including:

The first generation sub-module is used for generating a plurality of test cases according to the scene information, the scene case creation rule, the related phrases corresponding to the scene information and a preset screening algorithm, wherein the plurality of test cases comprise all the related phrases;

the sending submodule is used for sending the test text to the test equipment of a customer group, and each test equipment uniquely corresponds to one test customer in the customer group;

the statistics sub-module is used for counting the reflux data of each test document;

a second determining submodule, configured to determine a document preference of the customer base according to the reflow data and the customer images of the test customers in the customer base;

and the second generation submodule is used for generating a document model according to the document preference of different customer groups, and the document model is used for generating corresponding message documents for different customer groups.

Optionally, the first generating sub-module is specifically configured to generate a plurality of first test cases according to the scene information, the scene case creation rule, and related phrases corresponding to the scene information, where the plurality of first test cases include all arrangements of the related phrases; and screening to obtain a plurality of second test patterns according to the preset screening algorithm and the plurality of first test patterns, wherein the number of the second test patterns is smaller than that of the first test patterns.

Optionally, the sending submodule is specifically configured to divide the test clients of the client group into a preset number of groups according to a grouping rule, where the preset number is related to the number of test documents; pushing corresponding test files to the test equipment of each grouping.

Optionally, the device further includes a model optimization module, configured to obtain reflow data of the message documents of each client group, where the reflow data is determined according to click conditions of the message documents corresponding to each client group; and continuously and iteratively optimizing the document model according to the reflow data.

In a third aspect, the present application provides a server comprising: a memory and a processor;

The memory is used for storing program instructions;

the processor is configured to invoke program instructions in the memory to perform the message document generation method of the first aspect and any of the possible designs of the first aspect.

In a fourth aspect, the present application provides a readable storage medium having stored thereon a computer program which, when executed by at least one processor of a server, performs the message document generating method of the first aspect and any of the possible designs of the first aspect.

In a fifth aspect, the present application provides a computer program product comprising a computer program which, when executed by at least one processor of a server, performs the message document generation method of the first aspect and any of the possible designs of the first aspect.

The method and the device for generating the message text provided by the application are characterized in that scene topics and scene text creation rules corresponding to the scene topics are determined; acquiring related phrases from a phrase library according to the scene theme and scene document creation rules, wherein the phrases in the phrase library are obtained by decomposing historical document data; constructing and generating a message document according to related phrases and scene document creation rules, wherein a document model comprises document preferences of a target customer group, and the document preferences are used for indicating preferences of different customer groups on different expression modes in the same document element; the method for pushing the message file to the target equipment corresponding to the target customer group achieves the effects of improving the accuracy of capturing the preference of the customer group by the file model, improving the attraction of the message file to the customer group and increasing the click rate of the message file.

Drawings

In order to more clearly illustrate the application or the technical solutions of the prior art, the following description of the embodiments or the drawings used in the description of the prior art will be given in brief, it being obvious that the drawings in the description below are some embodiments of the application and that other drawings can be obtained from them without inventive effort for a person skilled in the art.

Fig. 1 is a schematic diagram of a scenario for pushing a message document according to an embodiment of the present application;

FIG. 2 is a flowchart of a message document generating method according to an embodiment of the present application;

FIG. 3 is a flowchart of another message document generating method according to an embodiment of the present application;

FIG. 4 is a flowchart of another message document generating method according to an embodiment of the present application;

FIG. 5 is a flowchart of another message document generation method according to an embodiment of the present application;

Fig. 6 is a schematic structural diagram of a message document generating device according to an embodiment of the present application;

Fig. 7 is a schematic structural diagram of another message document generating device according to an embodiment of the present application;

fig. 8 is a schematic hardware structure of a server according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

It should be understood that although the terms first, second, third, etc. may be used herein to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope herein.

Furthermore, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise.

It will be further understood that the terms "comprises," "comprising," "includes," and/or "including" specify the presence of stated features, steps, operations, elements, components, items, categories, and/or groups, but do not preclude the presence, presence or addition of one or more other features, steps, operations, elements, components, items, categories, and/or groups.

The terms "or" and/or "as used herein are to be construed as inclusive, or meaning any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: A. B. C. a and B. A and C. B and C. A. B and C). An exception to this definition will occur only when a combination of elements, functions, steps or operations are in some way inherently mutually exclusive.

In the era of mobile interconnection, self-service channels (APP, weChat public number, short message) have become the most important tie for enterprises to connect customers with products and services. The enterprise can guide clients to the APP, the applet or the public number through push messages to participate in activities or transact business. After the client views the information pushed by the enterprise in the self-service channel, the client can participate in the activity or transact business by clicking the information. The click-through promotion of the message directly affects the activity participation level or the business handling scale. In the enterprise operation process, the activity participation degree or the change of the business handling scale can have important influence on a series of key operation indexes such as income, client viscosity, daily activity at the mobile end, monthly activity and the like.

In the prior art, after determining the activity content or business content, the enterprise needs to write the message document first. The content of the message document is directly related to the active content or business content. When the message file is determined, the message file is sent to the client. At present, the message document is mostly written by staff according to activity content or business content. Since the message document is pushed towards all clients, the primary requirement is generally clear of the content during the composition of the message document. After receiving the push of the message file, the client can explicitly recommend the participating activities or recommend the transacted business through the message file. For the writing requirements, the content of the message file is mainly in a parallel and straight-forward description mode, and the problem that the message file cannot catch eyes exists due to lack of personalized design is caused. Further, the lack of personalized design and attractive message text will lead to the problem of low click-through rates for the activity or business.

In view of the above problems, the prior art proposes a method for generating a more attractive message document by means of advanced techniques such as natural language processing methods and machine learning algorithms. However, the message document generation method proposed in the prior art is mainly used for directly generating the target message document according to the document template, the marketing campaign and the customer portrait. The generation process of the message document lacks the function of generating the document from the aspects of quantitative analysis and an application data driven mode according to the preference of a client to the document element constitution and the document expression.

The application provides a message document generating method aiming at the problems. According to the application, through carrying out finer splitting and deconstructing on the message file, a scientific test design is used for pushing test, so that the client preference is more accurately captured, and the attraction of the message file is improved. For example, statistics find push message "[ exclusive welfare ]," multipoint APP shopping, full 60 minus 30-! The click rate of the' is high. By disassembling the structure, the message file is divided into "[ exclusive welfare ]," multipoint APP "and" full 60 minus 30 ]! "three parts". In the prior art, the message document is directly used as a piece of training data to optimize the document model. The optimization process will lose the analysis of the three parts of the message document. The application can obtain the preference of the client more accurately by testing and analyzing the preference of the client to different parts in the message file. The server adds the client preference to the document model, so that the effective knowledge for improving the document attractive force is solidified, thereby realizing the effects of enhancing the message document attractive force and improving the click rate of the message document.

As in the case described above, a message document may be split into three parts. In the present application, each part obtained by message document disassembly construction is defined as a document element. After the document data is collected, the server can determine the document elements commonly used in the design process by analyzing the document data. The text elements may include, among other things, sentence blurriness, emotional style, personality title, word description, specific rights, action encouragement, presentation location, and the like. Wherein, the mood color is used for the beginning part of the message document. The language elements of the language-qi moisturizing brings focus to customers by using emoticons or Chinese-English language-qi words. Wherein the emotion styles typically include web hotwords or simple sentences. The emotion style document element is used to further draw the attention or curiosity of the customer. Wherein the personality designation is used to pull a distance from the client, so that the client perceives that the message is closely related to him (her). Wherein the text description is used to illustrate the body and events of the message. Wherein specific rights are used to describe messages related to the customer's own interests. Wherein the action encourages the user to act to achieve the goal of directing the task for clicking.

In the application, in order to improve the generation efficiency and diversity of the message text, the server also needs to construct a phrase library on the basis of acquiring a large amount of text data. In the process of constructing the phrase library, the server marks scene topics and document elements for each phrase by utilizing a multi-term machine learning algorithm such as a topic model (LATENT DIRICHLET Allocation, LDA), word2Vec Word vectors, word move distance algorithm (WMD) and Affinity Propagation clustering algorithm. Meanwhile, the server expands the phrase semantic library by using the open source word library. The application increases the expression mode of each document element and the diversity of the message document by constructing a rich phrase library.

The technical scheme of the application is described in detail below by specific examples. The following embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.

Fig. 1 is a schematic diagram of a scenario of message document pushing according to an embodiment of the present application. As shown, the message file is generated by the server. After the message file is generated, the message file is pushed to the corresponding target device by the server. The target device may be a mobile terminal, tablet, computer, etc. The target device is provided with front-end software for receiving the message file, such as short message, APP, desktop application and the like. The front-end software displays the message document using a display of the target device. The client can view the message document through the display. When the client is interested in the content in the message file, the client can jump to the active interface or the business handling interface corresponding to the message file by clicking the message file. The active interface or the business handling interface may be an active interface of an APP, applet or desktop application corresponding to the front-end software. Or the active interface or the business transaction interface may also be an active interface for an APP, applet or desktop application specified in the message document.

In the application, the message document method of the following embodiment is executed by taking the server as an execution body. In particular, the execution body may be a hardware device of a server, or a software application implementing the embodiments described below in the server, or a computer-readable storage medium on which a software application implementing the embodiments described below is installed, or code of a software application implementing the embodiments described below.

Fig. 2 is a flowchart of a message document generating method according to an embodiment of the present application. On the basis of the embodiment shown in fig. 1, as shown in fig. 2, with the server as the execution body, the method of this embodiment may include the following steps:

s101, acquiring related phrases from a phrase library according to scene information and scene document creation rules, wherein the scene information comprises scene subjects, and the phrases in the phrase library are obtained by decomposing historical document data.

In this embodiment, the server acquires scene information. The scene information includes scene subjects. The scene topics may include, among other things, holiday activities, preferential activities, staging activities, card handling activities, financial activities, and the like.

Wherein the server may comprise an interactive interface. The client selects the target scene through the interactive interface. The server determines scene information according to the target scene selected by the client. Or the server may also obtain keyword information. The server selects one scene from the scenes stored in advance by the server as a target scene according to the keyword information. The server determines scene information from the target scene. The keyword information may be keyword information determined by the server according to the client input information. Or the key information can also be key word information determined by the server according to a preset rule. For example, the server may obtain current date information from which the server determines holiday activity.

And the server determines scene document creation rules corresponding to the scene topics according to the scene topics. The scene document creation rule comprises a plurality of document elements and an arrangement mode of the plurality of document elements. The server may select a phrase from a library of related phrases based on a plurality of document elements included in the scene document authoring rules. The phrase forms a message document according to the arrangement mode of the document elements. The text elements are obtained by splitting and deconstructing the whole text according to different actions of different phrases in the message text. The text elements include at least one of sentence color rendering, emotional style, personality title, word description, specific rights, action encouragement, and presentation location. Message documents of different scene subjects may be composed of different document elements. The composition of the file elements in the message text of different scene subjects can be obtained through data science and marketing psychology construction. For example, the document elements of the confectionary offer scene consist of: statement color + word description + specific rights + action encouragement.

In one example, the specific step of the server obtaining related phrases from the phrase library according to the scene information and scene document authoring rules may include:

step 1, a server determines a scene theme and a scene document creation rule corresponding to the scene theme according to scene information, wherein the scene document creation rule comprises document elements.

And 2, the server acquires related phrases matched with the scene theme from the phrase library according to the scene theme.

And 3, the server matches the related phrases corresponding to each document element from the related phrases obtained by screening in the step 2 according to the document elements included in the scene document creation rule.

In this step, the number of related phrases corresponding to each document element in the related phrases obtained by the server screening can be selected according to actual requirements. For example, 4 text elements may be included in a dessert offer scene. Wherein each document element may include a plurality of related phrases therein. Taking the example of the "statement" element, the related phrase may include "Woow-! "," [ limited name ] "," is? "," dingdong! "etc.

S102, constructing and generating a message document according to related phrases and document models, wherein the document models comprise document preferences of different customer groups, and the document preferences are used for indicating preferences of different customer groups on different expression modes in the same document element.

In this embodiment, the document model may include document preferences for a plurality of customer groups. After the server enters the phase Guan Duanyu into the document model, the document model may generate a plurality of message documents according to document preferences of a plurality of customer groups therein. Each message document is a personalized document generated for one customer group. The personalized document includes preferences of the customer group for the document.

S103, pushing the message file to target equipment, wherein the target equipment is equipment used by clients in a client group corresponding to the message file.

In this embodiment, the server may determine, according to a message document generated in S102, a client group corresponding to the message document. The server determines the devices of the clients in the client group according to the client group. The device is the target device of the message document. The server pushes the message file to each target device. The client can view the message document on his device. When the client is interested in the content in the message file, the client can participate in the activity in the message file or transact the business in the message file by clicking on the message file.

In the message case method provided by the application, the server determines the scene theme and the scene case creation rule corresponding to the scene theme according to the scene information. And the server acquires related phrases from the phrase library according to the scene theme and scene document creation rules. The phrases in the phrase library are obtained by disassembling and decomposing historical document data. The document model may include document preferences for a plurality of customer groups. After the server enters the phase Guan Duanyu into the document model, the document model may generate a plurality of message documents according to document preferences of a plurality of customer groups therein. The server determines its corresponding client group and target device according to a document message. The server pushes the message file to each target device. According to the application, the message file is disassembled and constructed into file elements, so that the file model can more accurately capture the effective information in the message file. Meanwhile, the text model is trained by carrying out reflux data analysis on different client groups, so that the text model can capture the preference of the different client groups. Furthermore, the personalized message document is generated by using the document model, so that the attraction of the message document to different client groups is improved, and the click rate of the message document is increased.

In addition, the click rate of the message file and the original message file generated by the method after being sent under various scene subjects is obtained. According to the click rate, the server can calculate the click lifting degree of the generated document, and the calculation formula of the click lifting degree is as follows:

click-through promotion = (first click rate-second click rate)/second click rate

The first click rate is the click rate of the message file generated by the method, and the second click rate is the click rate of the original message file. The click rate may be a click condition within a preset time period from the transmission of the message file. Compared with the original message file, the click rate of the message file generated by the method is obviously improved in the early-stage landing scene, and the click rate can reach 70%. Therefore, the method for generating the message file can effectively improve the click rate of the message file and increase the attraction of the message file.

Fig. 3 is a flowchart of another method for generating a message document according to an embodiment of the present application. On the basis of the embodiments shown in fig. 1 to 2, as shown in fig. 3, the server is taken as an execution subject, and in this embodiment, the phrase library generating process may include the following steps:

S201, acquiring historical document data, wherein the historical document data comprises a plurality of documents.

In this embodiment, the server obtains pushed history document data from each large mainstream APP. The historical document data may include a plurality of industries, such as finance, electronic commerce, video, and the like. The historical document data may relate to a variety of scenario topics of campaigns, holiday campaigns, hairpin campaigns, and the like. The historical document data may include a plurality of documents therein.

S202, dividing the historic document into phrases according to preset dividing rules.

In this embodiment, the server disassembles and constructs the collected historical document data according to a preset segmentation rule to obtain phrases. The specific process of the server for disassembling and constructing the historical document data according to the preset segmentation rule can comprise the following steps:

And step 1, the server disassembles the text into short sentences by using punctuation marks. Wherein punctuation marks may include. ","? ", I! ", and: ",". "etc.

And 2, the server performs word segmentation processing on the text by using a word segmentation method to obtain words. The word segmentation method can be an existing word segmentation method or an improved word segmentation method.

In one example, the server reviews the disassembled phrases according to preset review rules. The phrase includes phrases and words. When a sensitive word is included in the phrase, the server may delete the phrase.

S203, marking the scene theme and the text element corresponding to each phrase according to the phrases, the preset scene theme and the preset text element.

In this embodiment, after splitting the document data into phrases, the server may use a natural language processing algorithm such as a document topic generation model (LATENT DIRICHLET Allocation, LDA), word2Vec Word vector, word move' S DISTANCE, WMD, affinity Propagation clustering algorithm, and the like to mark scene topics and document elements for the phrases.

In one example, a particular process of the server marking scene topics and document elements for phrases may include the steps of:

Step 1, determining a word shift distance matrix of the phrase according to the phrase.

In this step, the server calculates the word shift distance between every two phrases. And the server determines a word shift matrix of the phrase according to the word shift distance.

And step 2, determining the text elements to which each phrase belongs according to the word shift distance matrix.

In this step, the server may use a neighbor propagation algorithm to cluster the phrases according to the word shift distance matrix of the phrases. According to the clustering result, the server can respectively classify each phrase into different document elements. In order to ensure the accuracy of classification of the document elements, an administrator can check the document elements corresponding to each classification after the server finishes classifying each document template. After determining the document element of a phrase, the server tags the phrase with the document element.

And step 3, clustering the phrases by using a clustering algorithm, and determining scene topics to which each phrase belongs.

In this step, the server may perform vectorization identification on the word obtained by word segmentation, so as to obtain a word vector of each word. The server may train the word vector using the topic model. The topic model may perform topic clustering on the word vectors, clustering all word vectors into one or more topics. In order to ensure the accuracy of classification of the document elements, an administrator can check the subject words of each type of word vectors after the server completes clustering of each word vector. After determining the scene topic of a phrase, the server marks the phrase with the scene topic.

S204, adding the marked phrases into a phrase library.

In this embodiment, the server stores phrases that are tagged with text elements and scene topics into a phrase library. The server distributes the phrases to different phrase libraries according to the scene theme corresponding to each phrase. Each phrase library corresponds to a scene topic. One phrase may correspond to multiple scene topics. When a phrase corresponds to a plurality of scene topics, the phrase may be a related phrase of the plurality of scene topics.

S205, using a natural language processing algorithm, expanding a phrase library by using an open source word library.

The server expands the phrase semantic library by using the open source word library. The application increases the expression mode of each document element and the diversity of the message document by constructing a rich phrase library.

According to the message document generation method provided by the application, the server disassembles and constructs the collected document data according to the preset segmentation rules to obtain the phrase library. The server marks the scene theme and the text element corresponding to each phrase according to the phrases, the preset scene theme and the preset text element. The server stores phrases that tag scene topics and document elements into a phrase library. According to the application, the richness of the document data is ensured by acquiring the document data from each large main stream APP, so that the diversity of the message document is ensured in the process of generating the message document.

Fig. 4 is a flowchart of still another method for generating a message document according to an embodiment of the present application. On the basis of the embodiments shown in fig. 1 to 3, as shown in fig. 4, the method for determining the document model in this embodiment may include the following steps:

S301, generating a plurality of test cases according to scene information, scene case creation rules, related phrases corresponding to the scene information and a preset screening algorithm, wherein the plurality of test cases comprise all the related phrases.

In this embodiment, the server first determines the scene information of the test. The server may determine scene topics from the scene information. And the server determines a document creation rule according to the scene theme. The document creation rule comprises a plurality of document elements and arrangement modes thereof. The server matches related phrases from a phrase library according to the scene subject and the document element. The number of phrases in each document element can be selected according to actual requirements.

And the server selects phrases from all the document elements according to the scene document creation rules and forms a plurality of test documents. The plurality of test documents includes all collocations of selected phrases in each document element. In the prior art, the test patterns are typically generated using permutation and combination. For example, the test dessert preference scene includes 4 document elements, each comprising 3 related phrases. The server can compose 81 test documents according to the 4 document elements and 12 related phrases. The 81 test cases cover all collocations of related phrases.

And the server adopts a preset screening algorithm to screen the plurality of test texts. The preset screening algorithm adopts a scientific test design method, comprehensively considers the information quantity and the complexity of a test group, adopts Fedorov, k-exchange and other technologies, and determines the collocation mode of the test text by the methods of maximizing an information matrix and minimizing a distance matrix variance. For example, after the above 81 test cases are screened by using a preset screening algorithm, the number of test cases in the test scene is 20.

S302, sending test files to test equipment of a customer group, wherein each test equipment uniquely corresponds to one test customer in the customer group.

In this embodiment, when a plurality of client groups are included, the server generates test documents of the client groups one by one, and transmits the test documents to devices of the corresponding client groups. The server determining a customer group as a target customer group, and the process of sending the test document to the target customer group by the server may include:

step 1, the server divides target test clients into preset number groups according to grouping rules, wherein the preset number is related to the number of test files. Wherein the target test clients are all clients in the target client group.

In this step, the grouping rule may be a random grouping. The server randomly divides all clients into a preset number of groups. Or the grouping rules may also be to group the customer images according to specific test objectives. For example, it is necessary to verify the appeal of a document to a group of guests under 30 years old, and it is possible to separately extract the parts of guests under 30 years old as a group. The process also needs to calculate the minimum value of the sample size in each group by using a methodology of test design, and ensures that the number of clients in each group is larger than the minimum value so as to ensure the validity of test analysis.

The preset number of the test groups is related to the number of the test files. For example, 20 test papers are divided into 20 groups.

And step 2, pushing a corresponding test document to the test equipment of each group.

In this step, the server sends a test document to each group. The test patterns of each group are different.

S303, counting the reflow data of each test document.

In this embodiment, the server obtains the reflow data of each test document. The reflow data includes test notes, test clients, whether to click, etc.

S304, determining the document preference of the client group according to the reflow data and the client portraits of the test clients in the client group.

In this embodiment, the server analyzes the reflow data by dividing the clients into different client groups based on the reflow data and the client representation. For example, a female customer base, a male customer base, a 20-25 year old customer base, a 26-35 year old customer base, and the like. The return data of the message file may be analyzed using analysis of variance, performance analysis, logistic regression, etc. The analysis results may include analysis of click-through boosting of the message document and analysis of importance indicators for individual document elements in the message document. Based on the analysis results, the server may determine different guest file preferences.

S305, according to the document preference of different client groups, generating a document model, wherein the document model is used for generating corresponding message documents for the different client groups.

In this embodiment, the server counts the reflow data of each client group. And the server determines the characteristics of the test texts of interest to each client group according to the reflow data. For example: in a test of a preferential activity scene, young guests like more lively expression modes such as 'someone @ you' like a 'statement color' element part, a more favored network hotword and the like. The female guest group is in the "word description" element section, which favors the expression of romantic styles, such as "××" with you's sweet date "from the" Wheather' ". Further, the server generates a document model based on the client document preferences learned from the test. The document model may be used to generate different message documents for different customer groups.

According to the message document generation method provided by the application, the server determines the test document according to the scene information and the related phrases corresponding to the scene information, and the test document is obtained by collocating the related phrases of each document element in the scene information. The server sends a corresponding test document to each test device. And the server acquires the reflow data of each test document. And the server obtains the text preference of different client groups in different scenes according to the reflow data and the client portrait. The server generates a document model based on the different document preferences of the different client groups. According to the application, the generation of the document model is realized by acquiring the reflow data and the customer portrait, and the document model can generate the message document conforming to the preference of the document of the customer group for each customer group. The use of the document model can realize the effect of improving document message clicking by increasing the attraction of the message document.

Fig. 5 is a flowchart of another method for generating a message document according to an embodiment of the present application. Based on the embodiments shown in fig. 1 and fig. 4, as shown in fig. 5, with a server as an execution body, in this embodiment, the model iterative optimization method may include the following steps:

s401, acquiring related phrases from a phrase library according to scene information and scene document creation rules, wherein the scene information comprises scene subjects, and the phrases in the phrase library are obtained by decomposing historical document data.

S402, constructing and generating a message document according to related phrases and document models, wherein the document models comprise document preferences of different customer groups, and the document preferences are used for indicating preferences of different customer groups on different expression modes in the same document element.

S403, pushing the message file to target equipment, wherein the target equipment is equipment used by clients in the client group corresponding to the message file.

Steps S401 to S403 are similar to steps S101 to S103 in the embodiment of fig. 2, and are not repeated here.

S404, obtaining the reflow data of the message file of each client group, wherein the reflow data is determined according to the clicking condition of the message file corresponding to each client group.

In this embodiment, the server may obtain the click condition of the pushed message document. The click condition of the pushed message file may be the reflow data of the message file. Or the reflow data of the message file may be other business related indicators. Or the reflow data of the message document may include click conditions and other business related indicators of the pushed message document.

S405, optimizing a document model according to the reflow data.

In this embodiment, the server may count the reflow data of the message document of each client group. After the server obtains the reflow data, the method such as analysis of variance, efficiency analysis, logistic regression and the like can be used for analyzing the reflow data of the message file. The analysis results may include analysis of click-through boosting of the message document and analysis of importance indicators for individual document elements in the message document. Based on the analysis results, the server may determine client preferences. Meanwhile, the server can add the message file meeting the client preference into the training data according to the reflow data. The training data is used for optimizing the document model, so that the document model continuously and accurately generates the personalized message document.

According to the method for generating the message text, the server acquires related phrases from the phrase library according to the scene information, wherein the scene information comprises a scene theme and a plurality of text elements corresponding to the scene theme, and the related phrases comprise phrases matched with the scene theme and the text elements in the phrase library. The server generates a message document according to the related phrases and document models, the document models can generate personalized documents for different client groups, the document models are used for selecting target phrases of all document elements from the related phrases, and the message document consists of the target phrases of all the document elements. The server pushes the message document to a target device, the target device corresponding to a target customer group. The server acquires the reflow data of the message file of the target client group, and the reflow data is determined according to the condition of clicking the message file of the target client group. And the server optimizes the document model according to the reflow data. According to the application, the message file is generated by using the file model, so that the attraction of the message file to different client groups is improved, and the click lifting degree of the message file is increased. Meanwhile, in the application, the optimization and updating of the message file are realized by acquiring and analyzing the reflow data, and the precise personalized positioning of the file model is realized, so that the attraction and click rate of the message file are further improved.

Fig. 6 is a schematic structural diagram of a message document generating device according to an embodiment of the present application, as shown in fig. 6, a message document generating device 10 according to the present embodiment is configured to implement operations corresponding to a server in any of the above method embodiments, where the message document generating device 10 according to the present embodiment includes:

The phrase obtaining module 11 is configured to obtain related phrases from a phrase library according to scene information and scene document creation rules, where the scene information includes scene topics, and the phrases in the phrase library are obtained by splitting and constructing historical document data.

The document generation module 12 is configured to obtain reflow data of the message documents of each client group, where the reflow data is determined according to clicking conditions of the message documents corresponding to each client group.

The document pushing module 13 is configured to push the message document to a target device, where the target device is a device used by a client in the client group corresponding to the message document.

In an example, the phrase obtaining module 11 is specifically configured to determine, according to the scene information, a scene theme and a scene document creation rule corresponding to the scene theme, where the scene document creation rule includes a document element composition. And acquiring related phrases matched with the scene theme from a phrase library according to the scene theme. According to the text elements included in the scene text creation rules, matching the related phrases corresponding to each text element from the related phrases.

In one example, the document element is determined according to the role of the phrase in the message document, the phrase is formed by disassembling the message document, and the document element comprises at least one of sentence color, emotion style, personality title, word description, specific rights, action encouragement and presentation position.

The message document generating device 10 provided in the embodiment of the present application may execute the above method embodiment, and the specific implementation principle and technical effects of the method embodiment may be referred to the above method embodiment, which is not described herein again.

Fig. 7 is a schematic structural diagram of another message document generating device according to an embodiment of the present application, where, on the basis of the embodiment shown in fig. 6, as shown in fig. 7, the message document generating device 10 of this embodiment is configured to implement operations corresponding to a server in any of the above method embodiments, and the message document generating device 10 of this embodiment further includes: phrase library generation module 14, model generation module 15, model optimization module 16.

The phrase library generation module 14 specifically includes:

The obtaining sub-module 141 is configured to obtain historical document data, where the historical document data includes a plurality of documents.

The segmentation sub-module 142 is configured to segment the text in the text data into phrases according to a preset segmentation rule.

The marking sub-module 143 is configured to mark the scene theme and the document element corresponding to each phrase according to the phrase, the preset scene theme and the preset document element.

An add sub-module 144 for adding the tagged phrases to the phrase library.

In one example, the labeling sub-module 143 is specifically configured to determine a word shift distance matrix for the phrase based on the phrase. And determining the text element to which each phrase belongs according to the word shift distance matrix. And clustering the phrases by using a clustering algorithm, and determining the scene subject to which each phrase belongs. The phrases are tagged according to the document elements and scene topics of each phrase.

In one example, the phrase library generation module 14 further includes:

an expansion sub-module 145 for expanding the phrase library with the Chinese open source word library using natural language processing algorithms.

The model generating module 15 specifically includes:

The first generation sub-module 151 is configured to generate a plurality of test cases according to the scene information, the scene case creation rule, the related phrases corresponding to the scene information, and the preset screening algorithm, where the plurality of test cases include all the related phrases.

The sending sub-module 152 is configured to send the test document to the test devices of a customer group, where each test device uniquely corresponds to a test customer in the customer group.

The statistics sub-module 153 is configured to count the reflow data of each test document.

A second determination submodule 154 for determining the document preference of the customer base based on the reflow data and the customer representation of each test customer in the customer base.

A second generation sub-module 155, configured to generate a document model according to document preferences of different customer groups, where the document model is configured to generate corresponding message documents for different customer groups.

In an example, the first generating sub-module 151 is specifically configured to generate a plurality of first test documents according to the scene information, the scene document creation rule, and the related phrases corresponding to the scene information, where the plurality of first test documents includes all arrangements of the related phrases. And screening to obtain a plurality of second test patterns according to a preset screening algorithm and a plurality of first test patterns, wherein the number of the second test patterns is smaller than that of the first test patterns.

In one example, the sending sub-module 152 is specifically configured to divide the test clients of the client group into a preset number of groups according to the grouping rule, where the preset number is related to the number of test documents. And pushing the corresponding test text to the test equipment of each group.

The model optimization module 16 is configured to obtain reflow data of the message documents of each client group, where the reflow data is determined according to clicking conditions of the message documents corresponding to each client group. And according to the reflow data, continuously iterating and optimizing the text model.

Fig. 8 shows a schematic hardware structure of a server according to an embodiment of the present application. As shown in fig. 8, the server 20, configured to implement operations corresponding to the server in any of the above method embodiments, the server 20 of this embodiment may include: a memory 21, a processor 22 and a communication interface 24.

A memory 21 for storing a computer program. The Memory 21 may include a high-speed random access Memory (Random Access Memory, RAM), and may further include a Non-Volatile Memory (NVM), such as at least one magnetic disk Memory, and may also be a U-disk, a removable hard disk, a read-only Memory, a magnetic disk, or an optical disk.

A processor 22 for executing a computer program stored in the memory to implement the message document generating method in the above embodiment. Reference may be made in particular to the relevant description of the embodiments of the method described above. The Processor 22 may be a central processing unit (Central Processing Unit, CPU), or may be other general purpose Processor, digital signal Processor (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present invention may be embodied directly in a hardware processor for execution, or in a combination of hardware and software modules in a processor for execution.

Alternatively, the memory 21 may be separate or integrated with the processor 22.

When memory 21 is a separate device from processor 22, server 20 may also include a bus 23. The bus 23 is used to connect the memory 21 and the processor 22. The bus 23 may be an industry standard architecture (Industry Standard Architecture, ISA) bus, an external device interconnect (PERIPHERAL COMPONENT INTERCONNECT, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, among others. The buses may be divided into address buses, data buses, control buses, etc. For ease of illustration, the buses in the drawings of the present application are not limited to only one bus or to one type of bus.

The communication interface 24 may be connected to the processor 21 via a bus 23. The processor 22 may enable interaction with a target device through the communication interface 24.

The server provided in this embodiment may be used to execute the above-mentioned message document generating method, and its implementation manner and technical effects are similar, and this embodiment is not repeated here.

The present application also provides a computer-readable storage medium having a computer program stored therein, which when executed by a processor is adapted to carry out the methods provided by the various embodiments described above.

The computer readable storage medium may be a computer storage medium or a communication medium. Communication media includes any medium that facilitates transfer of a computer program from one place to another. Computer storage media can be any available media that can be accessed by a general purpose or special purpose computer. For example, a computer-readable storage medium is coupled to the processor such that the processor can read information from, and write information to, the computer-readable storage medium. In the alternative, the computer-readable storage medium may be integral to the processor. The processor and the computer readable storage medium may reside in an Application SPECIFIC INTEGRATED Circuits (ASIC). In addition, the ASIC may reside in a client device. The processor and the computer-readable storage medium may also reside as discrete components in a communication device.

In particular, the computer readable storage medium may be implemented by any type or combination of volatile or non-volatile Memory devices, such as Static Random-Access Memory (SRAM), electrically erasable programmable Read-Only Memory (EEPROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), read-Only Memory (ROM), magnetic Memory, flash Memory, magnetic or optical disk. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.

The present application also provides a computer program product comprising a computer program stored in a computer readable storage medium. The computer program may be read from a computer-readable storage medium by at least one processor of the apparatus, and executed by the at least one processor, causes the apparatus to implement the methods provided by the various embodiments described above.

The embodiment of the application also provides a chip, which comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor is used for calling and running the computer program from the memory, so that the device provided with the chip executes the method in the various possible implementation modes.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of modules is merely a logical function division, and there may be additional divisions of actual implementation, e.g., multiple modules may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or modules, which may be in electrical, mechanical, or other forms.

Wherein the individual modules may be physically separated, e.g. mounted in different locations of one device, or mounted on different devices, or distributed over a plurality of network elements, or distributed over a plurality of processors. The modules may also be integrated together, e.g. mounted in the same device, or integrated in a set of codes. The modules may exist in hardware, or may also exist in software, or may also be implemented in software plus hardware. The application can select part or all of the modules according to actual needs to realize the purpose of the scheme of the embodiment.

When the individual modules are implemented as software functional modules, the integrated modules may be stored in a computer readable storage medium. The software functional modules described above are stored in a storage medium and include instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or processor to perform some of the steps of the methods of the various embodiments of the application.

It should be understood that, although the steps in the flowcharts in the above embodiments are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited in order and may be performed in other orders, unless explicitly stated herein. Moreover, at least some of the steps in the figures may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, the order of their execution not necessarily occurring in sequence, but may be performed alternately or alternately with other steps or at least a portion of the other steps or stages.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same. Although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments may be modified or some or all of the technical features may be replaced with equivalents. Such modifications and substitutions do not depart from the spirit of the application.

Claims

1. A method of generating a message document, the method comprising:

pushing the message file to target equipment, wherein the target equipment is equipment used by clients in a client group corresponding to the message file;

The method further comprises the steps of:

counting the reflow data of each test file;

2. The method of claim 1, wherein the obtaining related phrases from a phrase library according to the context authoring rules and the context information comprises:

acquiring related phrases matched with the scene theme from a phrase library according to the scene theme;

3. The method of claim 2, wherein the document element is determined based on the role played by the phrase in the message document, the phrase being formed by a breakdown of the message document, the document element including at least one of sentence color, emotional style, personality title, word description, specific rights, action encouragement, presentation location.

4. A method according to any one of claims 1-3, characterized in that the method further comprises:

Acquiring historical document data, wherein the historical document data comprises a plurality of documents;

dividing the text in the history text data into phrases according to a preset dividing rule;

and adding the marked phrases into the phrase library.

5. The method of claim 4, wherein the marking the scene topic and the document element corresponding to each phrase according to the phrase, the preset scene topic and the preset document element comprises:

determining a word shift distance matrix of the phrase according to the phrase;

6. The method according to claim 5, further comprising:

and expanding the phrase library by using a natural language processing algorithm and an open source word library.

7. The method of claim 1, wherein generating a plurality of test documents according to the scene information, the scene document authoring rules, the related phrases corresponding to the scene information, and a preset screening algorithm comprises:

8. The method of claim 7, wherein the sending the test document to a testing device of a customer segment comprises:

pushing corresponding test files to the test equipment of each grouping.

9. A method according to any one of claims 1-3, characterized in that the method further comprises:

10. A message document generation method apparatus, the apparatus comprising:

The phrase acquisition module is used for acquiring related phrases from a phrase library according to scene information, wherein the scene information comprises a scene theme and a plurality of text elements corresponding to the scene theme, the related phrases comprise phrases matched with the scene theme and the text elements in the phrase library, and the phrases in the phrase library are obtained by splitting historical text data;

the text pushing module is used for pushing the message text to target equipment, wherein the target equipment is equipment used by clients in a client group corresponding to the message text;

The apparatus further comprises: a model generation module;

the model generation module comprises:

And the second generation submodule is used for generating a document model according to the document preference of different client groups, and the document model is used for generating corresponding message documents for different client groups.

11. A server, the server comprising: a memory, a processor;

The memory is used for storing a computer program; the processor is configured to implement the message document generation method according to any one of claims 1 to 9 according to a computer program stored in the memory.

12. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program for realizing the message document generating method according to any of claims 1-9 when executed by a processor.

13. A computer program product, characterized in that the computer program product comprises a computer program, characterized in that the computer program, when executed by a processor, implements the message document generation method of any of claims 1-9.