CN113158056A

CN113158056A - Recommendation language generation method and device

Info

Publication number: CN113158056A
Application number: CN202110458165.3A
Authority: CN
Inventors: 王秉慧; 杜佳琦; 吴宇娟; 王怡; 刘记平; 李拓; 吴静姗; 赵伟; 王梦麟; 罗与天; 潘民兰; 王杰; 张大宗; 李云龙; 雷云; 施展; 裴武扬; 陈灵龙; 蒋本朋; 邵晶
Original assignee: Koubei Shanghai Information Technology Co Ltd
Current assignee: Koubei Shanghai Information Technology Co Ltd
Priority date: 2021-04-27
Filing date: 2021-04-27
Publication date: 2021-07-23

Abstract

The application discloses a recommendation generating method and device, and relates to the technical field of information. The method comprises the following steps: collecting historical association data between the user and the object under different dimensions; determining association weights between the user and different dimensionality recommenders of the object according to the historical association data; acquiring a target dimension of which the associated weight meets a preset condition; and generating a recommendation of the object for the user according to the target dimension. The recommendation language generation method and device can generate the recommendation language matched with the user under the target dimension, so that the generated recommendation language is more accurate and suitable for the user, and the selection rate of the object can be improved.

Description

Recommendation language generation method and device

Technical Field

The present application relates to the field of information technologies, and in particular, to a method and an apparatus for generating a recommendation.

Background

With the rapid development of science and technology, the living material level is continuously improved, the user has more and more requirements on object selection, and in order to facilitate the user to select a corresponding object, a recommendation corresponding to the object needs to be added and displayed to the user, so that the user can quickly match the required object according to the recommendation.

At present, a recommendation of an object is generated and displayed to a user according to a description provided by an object publisher. However, most of the descriptors provided by the object publisher are related to the object, and the relevance between the descriptors and the user is not large, so that the method is difficult to give accurate and appropriate recommenders from the perspective of the user, and in addition, the generated recommenders have single dimension and are not beneficial to attracting the attention of the user, thereby causing the selection efficiency of the object to be low.

Disclosure of Invention

In view of this, the present application provides a method and an apparatus for generating a recommended language, and a main objective of the method and the apparatus is to generate a recommended language in a target dimension matching with a user, so as to ensure that the generated recommended language is more accurate and suitable for the user, and thus, the selection rate of an object can be improved.

According to an aspect of the present application, there is provided a method for generating a recommendation, the method including:

collecting historical association data between the user and the object under different dimensions;

determining association weights between the user and different dimensionality recommenders of the object according to the historical association data;

acquiring a target dimension of which the associated weight meets a preset condition;

and generating a recommendation of the object for the user according to the target dimension.

Optionally, the obtaining of the target dimension of which the associated weight meets the preset condition includes:

determining the basic weight of the object aiming at the recommendation words with different dimensions;

adding the basic weight and the associated weight to obtain a total weight between the user and the recommendation words with different dimensions of the object;

and determining the dimension of which the total weight is greater than the preset weight as a target dimension.

Optionally, the different dimensions include a primary dimension, or a sum of the primary dimension and a secondary dimension, where the secondary dimension is a sub-dimension corresponding to the primary dimension.

Optionally, the collecting historical association data between the user and the object in different dimensions includes:

collecting first behavior data between the user and the object in different dimensions;

determining a target crowd to which the user belongs, and collecting second behavior data of the target crowd and the object under different dimensions;

determining similar objects corresponding to the objects, and collecting third behavior data between the user and the similar objects in different dimensions and fourth behavior data between the target population and the similar objects in different dimensions;

determining the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data as the history correlation data.

Optionally, the determining, according to the historical association data, association weights between the user and different-dimensional recommenders of the object includes:

determining associated weights corresponding to the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data, respectively;

and adding the associated weights respectively corresponding to the first behavior data, the second behavior data, the third behavior data and/or the fourth behavior data to obtain the associated weights between the user and the different-dimension recommenders of the object.

Optionally, the generating a recommendation of the object for the user according to the target dimension includes:

and generating grammar through a preset recommendation corresponding to the target dimension, and generating the recommendation of the object for the user.

Optionally, if the target dimension includes at least one dimension of a weight, a material, and a taste corresponding to the object, the generating a grammar by using a preset recommendation corresponding to the target dimension to generate the recommendation of the object for the user includes:

obtaining raw data of at least one dimension of the weight, the material and the taste;

and generating a recommendation corresponding to the object for the user according to the original data.

Optionally, if the target dimension includes comment information, the generating a grammar by using a preset recommendation corresponding to the target dimension to generate the recommendation of the object for the user includes:

obtaining comment information related to the object;

filtering the comment information by using a preset front lexicon to obtain filtered comment information;

splicing the filtered comment information by using a preset grammar model to obtain a recommendation language corresponding to the object; or the like, or, alternatively,

and performing word positioning on the comment words in the comment information by using a preset front word bank, and splicing the positioned comment words to obtain a recommendation word corresponding to the object.

Optionally, if the target dimension includes at least one of an order placing behavior, a good comment behavior, and a browsing behavior, the generating a grammar by using a preset recommendation corresponding to the target dimension to generate the recommendation of the object for the user includes:

counting original data under at least one dimension of ordering behavior, favorable behavior and browsing behavior to obtain a first statistical result corresponding to the object;

and generating a recommendation corresponding to the object based on the first statistical result and a first preset grammar model.

Optionally, if the target dimension includes at least one of a regional ranking, an in-store ranking, and a quality ranking, the generating a grammar from a preset recommendation corresponding to the target dimension to generate the recommendation of the object for the user includes:

counting the regional ranking, the in-store ranking or the quality ranking of the object within a preset time period to obtain a second statistical result corresponding to the object;

and generating a recommended language corresponding to the object according to the second statistical result and a second preset grammar model.

Optionally, after generating the recommendation of the object for the user according to the target dimension, the method further includes:

if the object has a plurality of target dimension recommenders, respectively counting the historical click rate corresponding to the object marked with the plurality of target dimension recommenders;

and sequencing the plurality of target dimension recommenders based on the historical click rate, and determining the target dimension recommenders to be displayed according to a sequencing result.

judging whether the recommendation language of the object needs to be highlighted or not according to the target dimension;

and if the recommendation language of the object needs to be highlighted, adding a highlighting mark corresponding to the recommendation language, and displaying the recommendation language added with the mark.

According to another aspect of the present application, there is provided a recommendation generation apparatus, including:

the collecting unit is used for collecting historical association data between the user and the object under different dimensions;

the determining unit is used for determining association weights between the user and different dimensionality recommenders of the object according to the historical association data;

the acquisition unit is used for acquiring a target dimension of which the associated weight meets a preset condition;

and the generating unit is used for generating the recommendation language of the object for the user according to the target dimension.

Optionally, the obtaining unit includes: a determination module and an addition module for determining the sum,

the determining module is used for determining the basic weight of the object aiming at the recommendation words with different dimensions;

the adding module is used for adding the basic weight and the associated weight to obtain a total weight between the user and the recommendation words with different dimensions of the object;

the determining module is further configured to determine, as a target dimension, a dimension in which the total weight is greater than a preset weight.

Optionally, the collection unit comprises: a collection module and a determination module, wherein the collection module and the determination module,

the collection module is used for collecting first behavior data between the user and the object under different dimensions;

the collection module is further used for determining a target crowd to which the user belongs and collecting second behavior data of the target crowd with the object under different dimensions;

the collection module is further configured to determine a similar object corresponding to the object, and collect third behavior data between the user and the similar object in different dimensions, and fourth behavior data between the target population and the similar object in different dimensions;

the determining module is configured to determine the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data as the history associated data.

Optionally, the determining unit includes: a determination module and an addition module for determining the sum,

the determining module is configured to determine associated weights corresponding to the first behavior data, the second behavior data, the third behavior data, and/or the fourth behavior data, respectively;

the adding module is configured to add the association weights corresponding to the first behavior data, the second behavior data, the third behavior data, and/or the fourth behavior data, respectively, to obtain association weights between the user and the object with different dimensions.

Optionally, the generating unit is specifically configured to generate a grammar according to a preset recommendation corresponding to the target dimension, and generate the recommendation of the object for the user.

Optionally, if the target dimension includes at least one of a weight, a material, and a taste corresponding to the object, the generating unit includes: an acquisition module and a generation module, wherein the acquisition module and the generation module,

the acquisition module is used for acquiring original data of at least one dimension of the weight, the material and the taste;

and the generating module is used for generating a recommendation corresponding to the object for the user according to the original data.

Optionally, if the target dimension includes comment information, the generating unit includes: an acquisition module, a filtering module and a splicing module,

the acquisition module is used for acquiring comment information related to the object;

the filtering module is used for filtering the comment information by utilizing a preset front lexicon to obtain filtered comment information;

the splicing module is used for splicing the filtered comment information by using a preset grammar model to obtain a recommendation language corresponding to the object; or, word positioning is carried out on the comment words in the comment information by utilizing a preset front word bank, and the positioned comment words are spliced to obtain the recommendation words corresponding to the object.

Optionally, if the target dimension includes at least one of an order taking behavior, a good comment behavior, and a browsing behavior, the generating unit includes: a statistic module and a generating module, wherein the statistic module is used for generating a statistic,

the statistical module is used for carrying out statistics on original data under at least one dimensionality of ordering behavior, favorable behavior and browsing behavior to obtain a first statistical result corresponding to the object;

and the generating module is used for generating a recommendation corresponding to the object based on the first statistical result and a first preset grammar model.

Optionally, if the target dimension includes at least one of a regional ranking, an in-store ranking, and a quality ranking, the generating unit includes: a statistic module and a generating module, wherein the statistic module is used for generating a statistic,

the statistical module is used for carrying out statistics on the regional ranking, the in-store ranking or the quality ranking of the object within a preset time period to obtain a second statistical result corresponding to the object;

and the generating module is used for generating a recommendation corresponding to the object according to the second statistical result and a second preset grammar model.

Optionally, the apparatus further comprises: a statistical unit for counting the number of the data units,

the statistical unit is used for respectively counting the historical click rate corresponding to the object marked with the target dimension recommenders if the object has the target dimension recommenders;

the determining unit is further configured to sort the plurality of target dimension recommenders based on the historical click rate, and determine the displayed target dimension recommenders according to a sorting result.

Optionally, the apparatus further comprises: a judging unit and a display unit, wherein,

the judging unit is used for judging whether the recommendation language of the object needs to be highlighted or not according to the target dimension;

and the display unit is used for adding a highlight display mark corresponding to the recommended word and displaying the recommended word added with the mark if the recommended word of the object needs to be highlighted.

According to still another aspect of the present application, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the above-described method of generating a recommended word.

According to still another aspect of the present application, there is provided a computer device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the method for generating the recommendation language when executing the computer program.

By means of the technical scheme, compared with the current mode of generating the recommendation language according to the description language provided by the object publisher, the recommendation language generation method and device provided by the application collect historical associated data of the user and the object under different dimensions; determining association weights between the user and different dimensionality recommenders of the object according to the historical association data; meanwhile, acquiring a target dimension of which the associated weight meets a preset condition; and finally, generating the recommended words of the objects for the users according to the target dimensions, so that the recommended words under the target dimensions matched with the users can be generated by collecting historical associated data of the users under different dimensions and the objects, the generated recommended words can be ensured to be closely associated with the users, the recommended words are more suitable for the users, the attraction of the objects to the users is increased, and the selection rate of the objects is improved.

The above description is only an overview of the technical solutions of the present application, and the present application may be implemented in accordance with the content of the description in order to make the technical means of the present application more clearly understood, and the following detailed description of the present application is provided in order to make the above and other objects, features, and advantages of the present application more clearly understandable.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:

fig. 1 is a flowchart illustrating a method for generating a recommendation provided in an embodiment of the present application;

FIG. 2 is a flowchart illustrating another method for generating a recommended phrase according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of a recommender provided in an embodiment of the present application;

FIG. 4 is a schematic diagram of another recommender provided in embodiments of the present application;

fig. 5 is a schematic structural diagram illustrating a recommendation generation apparatus according to an embodiment of the present application;

fig. 6 is a schematic structural diagram illustrating another recommendation generation apparatus according to an embodiment of the present application.

Detailed Description

The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

The method aims to solve the problems that relevance between a generated recommendation and a user is not large and the dimension is single in the prior art. The present embodiment provides a method for generating a recommendation, as shown in fig. 1, the method includes:

step 101, collecting historical association data between the user and the object in different dimensions.

The execution main body of the embodiment is a client device or a server device capable of generating a recommendation language, and the client side includes an Application (APP) with a transaction function, an applet, a public number, a WEB Application, and the like installed on a smart terminal such as a smart phone, a tablet computer, and the like.

The object may be a dish, or may be other types of goods, or services. The different dimensions may include a first-level dimension or a sum of the first-level dimension and a second-level dimension, the second-level dimension is a sub-dimension corresponding to the first-level dimension, the first-level dimension may include dimensions such as object attributes, user evaluation, user behaviors and list ranking, and the second-level dimension corresponding to the object attributes specifically includes each attribute tag of the object.

For example, the object is specifically a dish, and if the primary dimension corresponding to the dish is the attribute of the dish, the secondary dimension corresponding to the dish includes the quality, heat, raw materials, belonging cuisine, making mode, taste and the like corresponding to the dish; the secondary dimensionalities corresponding to the user behaviors specifically comprise ordering behaviors, favorable behaviors, browsing behaviors and the like of the user; the second-level dimensions corresponding to the list ranking specifically include dimensions such as area list data and in-store list data, and the next-level dimensions corresponding to the area list data specifically include sales ranking list, popularity promotion ranking list, good-evaluation ranking list and the like corresponding to the object.

For this embodiment, when a user browses an object at a client or the client recommends an object to the user, a recommendation corresponding to the object is generated and displayed to the user, so that the user can quickly match a desired object according to the recommendation, and the selection rate of the object is improved.

Wherein the historical associated data may include: the behavior data of the user between the user and the object in different dimensions, the behavior data of the crowd to which the user belongs between the user and the object in different dimensions, the behavior data of the user between the user and the same kind of object in different dimensions, the behavior data of the crowd to which the user belongs between the user and the same kind of object in different dimensions, and the like. The behavior data may specifically include purchase adding behavior data, purchase resuming behavior data, order placing behavior data, click behavior data, and the like of the user, and the association data between the user and the object in different dimensions can be determined according to the behavior data.

For example, the object may be dishes, and when dish recommendation is performed for the user, historical association data between the user and dishes marked with dish attributes, user evaluations, user behaviors and a list ranking data recommendation are collected respectively. Particularly, when historical association data between the user and the dishes marked with the dish attribute recommenders is collected, the dish purchasing behavior data, the dish repurchasing behavior data, the ordering behavior data and the clicking behavior data of the user aiming at the dishes marked with the dish attribute recommenders can be firstly collected, then the crowd to which the user belongs is determined, the purchasing behavior data, the repurchasing behavior data, the ordering behavior data and the clicking behavior data of the crowd aiming at the dishes marked with the dish attribute recommenders are collected, and finally, collecting the purchase adding behavior data, the purchase repeating behavior data, the order placing behavior data and the click behavior data of the crowd aiming at the type of the dish marked with the dish attribute recommending language.

According to the collected behavior data, historical association data between the user and dishes marked with dish attribute recommenders can be determined, and further according to the mode, historical association data between the user and dishes marked with user evaluation recommenders, historical association data between the user and dishes marked with user behavior recommenders, and historical association data between the user and dishes marked with list ranking data recommenders can be respectively determined, so that association weights between the user and different-dimension recommenders of the dishes can be determined according to the historical association data.

And 102, determining the association weight between the user and the recommendation words of different dimensions of the object according to the historical association data.

For this embodiment, in order to generate a recommended language that matches the user best and ensure the degree of association between the user and the generated recommended language, it is necessary to calculate the association weights between the user and the recommended languages of different dimensions of the object.

Specifically, behavior data of a user between objects and different dimensions are collected, behavior data of a crowd to which the user belongs between objects and different dimensions, behavior data of the user between the user and the same kind of objects and behavior data of the crowd to which the user belongs between the user and the same kind of objects and different dimensions are collected, association weights corresponding to the behavior data are determined according to the size of the collected behavior data, and the association weights corresponding to the behavior data are added to obtain association weights between recommenders of the user and the objects in different dimensions.

For example, when determining the association weight between the user and the user behavior recommendation language of the dish, first, the association weight corresponding to the first behavior data is determined according to the size of the first behavior data between the user and the dish marked with the user behavior recommendation language.

Specifically, the association weight corresponding to the first behavior data may be determined according to the data range in which the first behavior data is located, for example, when the first behavior data is greater than 0 and less than 10, the association weight corresponding to the first behavior data is determined to be 1; when the first behavior data is greater than or equal to 10 and smaller than 20, determining that the associated weight corresponding to the first behavior data is 2; and when the first behavior data is more than or equal to 20 and less than 30, determining that the association weight corresponding to the first behavior data is 3, and if the accumulated sum of the purchase adding behavior times, the purchase repurchasing behavior times, the order placing behavior times and the click behavior times of the user in the collected first behavior data is 25, determining that the association weight corresponding to the first behavior data is 3. Similarly, the association weight corresponding to the second behavior data can be determined according to the size of the second behavior data between the crowd to which the user belongs and the dishes marked with the user behavior recommendation, then the association weight corresponding to the third behavior data is determined according to the size of the third behavior data between the user and the similar dishes marked with the user behavior recommendation, finally the association weight corresponding to the fourth behavior data is determined according to the size of the fourth behavior data between the crowd to which the user belongs and the similar dishes marked with the user behavior recommendation, and the association weights corresponding to the behavior data are added to obtain the association weight between the user behavior recommendations of the user and the dishes.

Therefore, the association weight between the user and the recommendation in different dimensions of the object can be determined according to the above mode, so that the target dimension which is most matched with the user is determined according to the association weight, and the generated recommendation in the target dimension is most suitable for the user and is closely associated with the user.

And 103, acquiring a target dimension of which the associated weight meets a preset condition.

For this embodiment, after calculating the association weights between the user and the different-dimension recommenders of the object, the association weights may be sorted according to the calculated association weights, a dimension corresponding to an association weight with a ranking within a preset ranking range is determined as a target dimension, a dimension corresponding to an association weight with a top ranking may also be determined as a target dimension, a dimension corresponding to an association weight with an association weight within a preset weight range may also be determined as a target dimension, and a specific manner for determining a target dimension is not specifically limited in this embodiment.

And 104, generating a recommendation of the object for the user according to the target dimension.

For the embodiment, after the primary target dimension matched with the user is determined, the secondary target dimension matched with the user is continuously determined on the basis of the primary target dimension. Specifically, historical associated data of a user and an object under different secondary dimensions are collected, associated weights between the user and different secondary dimension recommenders of the object are determined according to the historical associated data, a secondary target dimension with the associated weights meeting preset conditions is obtained, and the recommenders of the object are generated for the user according to the secondary target dimension.

For example, the primary target dimension may be a user behavior, the secondary dimension corresponding to the user behavior includes a placing behavior, a favorable behavior and a browsing behavior of the user, and historical association data between the user and an object marked with a placing behavior recommendation, historical association data between the user and an object marked with a favorable recommendation and historical association data between the user and an object marked with a browsing behavior recommendation are collected respectively.

The historical associated data specifically includes: behavior data of the user between the user and the object under different secondary dimensions, and behavior data of the crowd to which the user belongs between the user and the object under different secondary dimensions; the method comprises the steps that behavior data of a user between the user and an object of the type under different secondary dimensions and behavior data of a crowd to which the user belongs between the user and the object of the type under different secondary dimensions are determined, association weights between the user and recommenders of the object of the type under different secondary dimensions are further determined according to the behavior data, a secondary target dimension with the association weights meeting preset conditions is determined, the determination process of the secondary target dimension is the same as the determination process of the primary dimension, and finally, recommenders are generated for the user according to the determined secondary target dimension. Therefore, for the same object, the recommended words seen by different users are different, and the generated recommended words and the users can be guaranteed to have higher matching degree and relevance degree.

In the process of generating the recommended words according to the secondary target dimensions, the recommended words corresponding to different secondary target dimensions are generated in different modes. For example, if the secondary target dimension is a dish raw material, the dish attribute information recorded by the merchant is directly obtained, and the raw material information in the dish attribute information is extracted and displayed to the user as a recommendation.

For another example, if the secondary target dimension is favorable behavior, obtaining comment information for the dish, determining favorable information of the user by positioning the associated words, accumulating the times of the favorable information, and generating a user behavior recommendation for the dish for the user according to the accumulated result and a specific grammar model.

For another example, if the secondary target dimension is user evaluation, obtaining comment information for the dish, and filtering and recombining the comment information to generate a user evaluation recommendation for the dish for the user.

For another example, if the secondary target dimension is the area list data, the area sales ranking, the area popularity surge ranking and the area good rating ranking corresponding to the dish can be counted, and the area list data recommendation for the dish is generated for the user according to the counting result and the specific grammar model.

Compared with the current mode of generating the recommendation according to the description provided by the object publisher, the method for generating the recommendation provided by the embodiment collects historical associated data between the user and the object in different dimensions; determining association weights between the user and different dimensionality recommenders of the object according to the historical association data; meanwhile, acquiring a target dimension of which the associated weight meets a preset condition; and finally, generating the recommended words of the objects for the users according to the target dimensions, so that the recommended words under the target dimensions matched with the users can be generated by collecting historical associated data of the users under different dimensions and the objects, the generated recommended words can be ensured to be closely associated with the users, the recommended words are more suitable for the users, the attraction of the objects to the users is increased, and the selection rate of the objects is improved.

Further, as a refinement and an extension of the specific implementation of the above embodiment, another recommendation generation method is provided, as shown in fig. 2, the method includes:

step 201, collecting historical association data between the user and the object in different dimensions.

For the present embodiment, in order to determine a target dimension that is most matched with an object, historical association data between a user and the object in different dimensions needs to be collected, and as an optional implementation manner, for a collection process of the historical association data, step 201 specifically includes: collecting first behavior data between the user and the object in different dimensions; determining a target crowd to which the user belongs, and collecting second behavior data of the target crowd and the object under different dimensions; determining a similar object corresponding to the object, and collecting third behavior data of the user and the similar object in different dimensions, and fourth behavior data of the target crowd and the similar object in different dimensions; determining the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data as the history correlation data.

For example, different dimensions are respectively user behavior and list ranking data, historical association data between the user and dishes marked with user behavior recommenders and historical association data between the user and dishes marked with list ranking data recommenders are collected. Specifically, when historical association data between a user and dishes marked with user behavior recommenders is collected, first behavior data of the user for dishes marked with the user behavior recommenders is collected, second behavior data of people to which the user belongs for dishes marked with the user behavior recommenders is collected, third behavior data of the user for dishes marked with the user behavior recommenders is collected, fourth behavior data of people to which the user belongs for dishes marked with the user behavior recommenders is collected, the first behavior data, the second behavior data, the third behavior data and the fourth behavior data respectively comprise purchase adding behavior data, purchase repeating behavior data, order placing behavior data and click behavior data, and historical association data between the user and the dishes marked with the user behavior recommenders is determined according to at least one of the first behavior data, the second behavior data, the third behavior data and the fourth behavior data, similarly, historical association data between the user and the dishes marked with the list ranking data can be determined in the above mode, so that association weights between the user and different-dimension recommenders of the dishes are determined according to the historical association data.

Step 202, determining the association weight between the user and the recommendation words of different dimensions of the object according to the historical association data.

For this embodiment, in order to determine the associated weights between the user and the recommenders for different dimensions of the object, step 202 specifically includes: determining associated weights corresponding to the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data, respectively; and adding the associated weights respectively corresponding to the first behavior data, the second behavior data, the third behavior data and/or the fourth behavior data to obtain the associated weights between the user and the different-dimension recommenders of the object. When determining the association weights corresponding to the different-dimension recommenders, the association weights corresponding to the different-dimension recommenders may be determined according to the association weight corresponding to any behavior data, or the association weights corresponding to any several behavior data in the first behavior data, the second behavior data, the third behavior data, and the fourth behavior data may be added to obtain the association weights corresponding to the different-dimension recommenders.

For example, when determining the association weight between the user and the list ranking data recommender of the dish, first, the association weight corresponding to the first behavior data is determined according to the size of the first behavior data between the user and the dish marked with the area list ranking data recommender. Specifically, the association weight corresponding to the first behavior data may be determined according to the data range in which the first behavior data is located, for example, when the first behavior data is greater than 0 and less than 10, the association weight corresponding to the first behavior data is determined to be 1; when the first behavior data is greater than or equal to 10 and smaller than 20, determining that the associated weight corresponding to the first behavior data is 2; when the first behavior data is more than or equal to 20 and less than 30, determining that the associated weight corresponding to the first behavior data is 3, if the accumulated sum of the times of purchase adding behavior, the times of repurchase behavior, the times of order release behavior and the times of click behavior of the users in the collected first behavior data is 25, determining that the associated weight corresponding to the first behavior data is 3, similarly, determining the associated weight corresponding to the second behavior data according to the size of the second behavior data between the crowd to which the users belong and the dishes marked with the regional list data recommenders, then determining the associated weight corresponding to the third behavior data according to the size of the third behavior data between the users and the similar dishes marked with the regional list data recommenders, and finally determining the associated weight corresponding to the fourth behavior data according to the size of the fourth behavior data between the crowd to which the users belong and the similar dishes marked with the regional list data recommenders, and adding the associated weights corresponding to the behavior data to obtain the associated weight between the user and the regional list data recommendation of the dishes.

And 203, acquiring a target dimension of which the associated weight meets a preset condition.

For the embodiment, if the user is a new user, the historical associated data of the user cannot be collected, so that the target dimension matched with the user cannot be determined according to the historical associated data, and in order to generate corresponding recommenders for the new user as well, when the target dimension is determined, the target dimension matched with the user can be determined according to the basis weights of the object for recommenders with different dimensions, and then the recommenders are generated for the new user; if the user is not a new user, jointly determining a target dimension matched with the user according to the calculated association weight and the basis weight, and further generating a corresponding recommended word for the user, based on which step 203 specifically includes: determining the basic weight of the object aiming at the recommendation words with different dimensions; adding the basic weight and the associated weight to obtain a total weight between the user and the recommendation words with different dimensions of the object; and determining the dimension of which the total weight is greater than the preset weight as a target dimension.

Specifically, when determining a target dimension corresponding to a new user, since the historical associated data of the user cannot be collected, historical associated data of the same type of population between the object and the same type of population in different dimensions can be collected, the basic weights of the object for recommenders in different dimensions can be determined according to the historical associated data corresponding to the same type of population, for example, the new user is a fat-reduction user, the historical associated data of the fat-reduction type population between the object and the same type of population in different dimensions is collected, the basic weights of the object for recommenders in different dimensions are determined according to the historical associated data corresponding to the fat-reduction type population, the target dimension with the basic weight meeting preset conditions is obtained, for example, the basic weight corresponding to the ranking data is the largest, the ranking data is determined as the target dimension, and for example, the ranking data and the basic weight corresponding to the user behavior are within a preset weight range, determining the ranking data of the list and the user behavior as target dimensions.

Further, if the user is not a new user, the association weight and the corresponding basis weight may be added to obtain a total weight between the user and the recommendation of the object in different dimensions, for example, the basis weight corresponding to the list ranking data and the association weight between the user and the recommendation of the list ranking data are added to obtain a total weight between the user and the list ranking data, and similarly, the total weight between the user and the user behavior and the total weight between the user and the commodity attribute may be obtained, and the dimension corresponding to the maximum total weight is determined as the target dimension, and the corresponding recommendation is generated for the user according to the target dimension. Therefore, according to the mode, the recommendation language can be generated not only for the old user, but also for the new user.

And 204, generating grammar through a preset recommendation corresponding to the target dimension, and generating the recommendation of the object for the user.

In this embodiment, after the primary target dimension matched with the user is determined, the secondary target dimension matched with the user needs to be continuously determined, and a specific determination process of the secondary target dimension is the same as that of the primary target dimension, and is not repeated again.

In a specific application scenario, if the secondary target dimension includes at least one dimension of a weight, a material, and a taste corresponding to an object, the generating a grammar by using a preset recommendation corresponding to the target dimension to generate a recommendation of the object for the user includes: obtaining raw data of at least one dimension of the weight, the material and the taste; and generating a recommendation corresponding to the object for the user according to the original data.

For example, if the primary target dimension is a dish attribute, and the secondary target dimension is a material corresponding to a dish, dish description information issued by a merchant may be acquired, the dish description information is used as original data, the material corresponding to the dish is extracted from the original data, and a recommendation corresponding to the dish is generated for the user. If the tomato scrambled eggs comprise tomatoes and eggs, the material labels 'tomatoes and eggs' can be displayed to the user as recommenders, furthermore, historical behavior data of the user for the tomato scrambled eggs can be collected, the historical behavior data comprise purchase times, order placing times and click times of the user for the tomato scrambled eggs, and by analyzing the historical behavior data of the user for the tomato scrambled eggs, if it is determined that the purchase times, order placing times and click operations of the user are more frequent when the material label 'tomatoes' is preferentially displayed, which indicates that the user has a higher interest level in the tomatoes, the material label 'tomatoes' can be preferentially displayed to the user, and then the 'eggs' are displayed.

In a specific application scenario, if a target secondary dimension includes comment information, the generating a grammar by using a preset recommendation corresponding to the target dimension to generate the recommendation of the object for the user includes: obtaining comment information related to the object; filtering the comment information by using a preset front lexicon to obtain filtered comment information; splicing the filtered comment information by using a preset grammar model to obtain a recommendation language corresponding to the object; or word positioning is carried out on the comment words in the comment information by utilizing a preset front word bank, and the positioned comment words are spliced to obtain the recommendation words corresponding to the object. Wherein the pre-set positive word library comprises positive words.

As shown in fig. 3, the secondary target dimension is comment information, the dish is roasted fish with sweet fragrance, comment information related to the roasted fish with sweet fragrance is collected, a preset front lexicon is used for locating front comment words in the comment information, sentences containing the front comment words are proposed, and then the extracted sentences are spliced by using a preset grammar model.

For example, the extracted sentence includes "wonderful roasted fish", "good taste", and "next return", which are composed using a preset grammar model to generate a recommendation "wonderful roasted fish, good taste, and next return" for the user. In addition, the positioned positive comment words can be directly extracted and spliced, for example, the extracted positive comment words comprise 'good words' and 'very tender', and after the comment words are spliced with the name of the dish, a recommendation word 'wonderful roasted fish, very delicious and very tender' is generated for the user. Therefore, the comment information recommendation words can be generated according to the two modes and displayed to the user, so that the user can quickly find the required dishes according to the comment words, and the ordering rate of the dishes is improved.

In a specific application scenario, if the secondary target dimension includes at least one of a placing behavior, a good comment behavior, and a browsing behavior, the generating a grammar by using a preset recommendation corresponding to the target dimension to generate the recommendation of the object for the user includes: counting original data under at least one dimension of ordering behavior, favorable behavior and browsing behavior to obtain a first statistical result corresponding to the object; and generating a recommendation corresponding to the object based on the first statistical result and a first preset grammar model.

As shown in fig. 3, if the secondary target dimension is the ordering behavior, then order ordering behavior data of a large number of users for the object is collected, and statistics is performed on the order ordering behavior data, and if the number of the users ordering for the object within 7 days is collected to be 100, an order ordering behavior recommendation "100 orders ordering within 7 days" is generated for the user according to a first preset grammar model. For another example, if the secondary target dimension is favorable behavior, collecting a large amount of favorable behavior data of the user for the object, counting the favorable behavior data, if 200 people give favorable comments within 1 month, and generating a favorable behavior recommendation "200 people give favorable comments in the near future" for the user according to the first preset grammar model. For another example, if the secondary target dimension is recommended by the user, collecting a large amount of recommended behavior data of the user for the object, counting the recommended behavior data, and generating a comment recommendation "comment webfriend recommendation" for the user according to the first preset grammar model when the recommended behavior data exceeds a preset number.

In a specific application scenario, if the secondary target dimension includes at least one of a regional ranking, an in-store ranking, and a quality ranking, the generating a grammar from a preset recommendation corresponding to the target dimension to generate a recommendation of the object for the user includes: counting the regional ranking, the in-store ranking or the quality ranking of the object within a preset time period to obtain a second statistical result corresponding to the object; and generating a recommended language corresponding to the object according to the second statistical result and a second preset grammar model. The regional ranking comprises sales ranking, popularity surging ranking and good rating ranking of the objects in the region, the in-store ranking comprises sales ranking, popularity surging ranking and good rating ranking of the objects in the stores, and the stores in the embodiment can be traded online through combination of software and hardware.

Specifically, sales data, sales growth data and good comment data of the object in a certain area are counted in a preset time period, ranking is performed according to the statistical result, and if the ranking rank corresponding to the object is within a preset range, a corresponding recommendation is generated according to the ranking result.

For example, counting sales data of all dishes in a Putuo area within one month, ranking according to the sales data from high to low, generating corresponding regional list data recommenders for the user according to a corresponding ranking result for the dishes with top 5, as shown in fig. 4, ranking the sales data of potato beef with top 5 in the Putuo area, generating a recommendation "potato beef order with top 5" for the user according to the ranking result, similarly counting sales increase data of the dishes within one week, ranking according to the sales increase data from high to low, generating corresponding regional list data recommenders for the user according to the ranking result corresponding to the dishes with the ranking within a preset ranking number, such as "potato beef number with Putuo area skyihui 1", and further, ranking according to the statistical result of good evaluation data, for the dishes with the preset ranking, and generating a corresponding area list data recommendation word for the user according to the corresponding sequencing result, such as '2 nd best ranking list of potato and beef in Putuo area'.

Similarly, the sales data, the sales growth data and the good appraisal data of the object in the shop are counted in a preset time period, ranking is carried out according to the statistical result, and if the ranking rank corresponding to the object is in a preset range, corresponding recommenders are generated according to the ranking result, for example, "the first name of the order of the local shop", "the 2 nd good appraisal of the local shop", and "the 1 st businessness of the local shop".

Further, whether the object belongs to a high-quality class or not can be judged according to sales data and good evaluation data corresponding to the object, a recommendation corresponding to the object is generated according to a judgment result and displayed to the user, specifically, the object with the sales data larger than the preset sales data and the good evaluation rate larger than the preset good evaluation rate can be screened out, if the sales ranking and the good evaluation rate ranking corresponding to the object can be in a preset ranking range, the high-quality class of the object is determined, and when the object is recommended to the user, a recommendation corresponding to the object, such as a high-quality pasta commodity, can be generated. Therefore, the list ranking data recommendation can be generated according to the method, so that the needed objects can be quickly matched through the list ranking data recommendation, and the selection rate of the objects is improved.

Step 205, according to the target dimension, determining whether the recommendation of the object needs to be highlighted.

For the embodiment, in order to facilitate the user to distinguish the recommendation and other information of the object, the generated recommendation may be displayed in a different display manner from other information, for example, the recommendation is displayed in a form of yellow-background red characters, and in addition, in order to further arouse the attention of the user, the recommendation of some dimensions needs to be highlighted, for example, if the generated recommendation is an area list data recommendation, the recommendation needs to be highlighted so as to improve the attention of the user.

And step 206, if the recommendation of the object needs to be highlighted, adding a highlighting mark corresponding to the recommendation, and displaying the recommendation after the mark is added.

For the embodiment, if the object has a plurality of target dimension recommenders, respectively counting the historical click volumes corresponding to the object marked with the plurality of target dimension recommenders; and sequencing the plurality of target dimension recommenders based on the historical click rate, and determining the target dimension recommenders to be displayed according to a sequencing result.

For example, if the historical click rates of the objects marked with the regional list data recommendation and the in-store list data recommendation are both greater than the preset click rate, the regional list recommendation and the in-store list recommendation are sorted according to the size of the historical click rate, only the first recommendation is displayed to the user, the regional list recommendation and the in-store list recommendation can be ranked and displayed in the ranking order, further, if the historical click rates of the objects marked with the regional list data recommendation and the in-store list data recommendation are both less than or equal to the preset click rate, the regional list recommendation or the in-store list recommendation can be randomly selected and displayed to the user, and the regional list recommendation and the in-store list recommendation can be randomly ranked, and displaying the arranged recommendation words to the user.

Compared with the current mode of generating the recommendation according to the description provided by the object publisher, the generation method of the recommendation provided by the embodiment collects historical association data between the user and the object in different dimensions; determining association weights between the user and different dimensionality recommenders of the object according to the historical association data; meanwhile, acquiring a target dimension of which the associated weight meets a preset condition; and finally, generating the recommended words of the objects for the users according to the target dimensions, so that the recommended words under the target dimensions matched with the users can be generated by collecting historical associated data of the users under different dimensions and the objects, the generated recommended words can be ensured to be closely associated with the users, the recommended words are more suitable for the users, the attraction of the objects to the users is increased, and the selection rate of the objects is improved.

Further, as a specific implementation of the method shown in fig. 1 and fig. 2, this embodiment provides a recommendation generating device, as shown in fig. 5, the device includes: a collection unit 31, a determination unit 32, an acquisition unit 33, and a generation unit 34.

The collecting unit 31 may be configured to collect historical association data between the user and the object in different dimensions.

The determining unit 32 may be configured to determine, according to the historical association data, association weights between the user and different dimension recommenders of the object.

The obtaining unit 33 may be configured to obtain a target dimension of which the associated weight meets a preset condition.

The generating unit 34 may be configured to generate a recommendation of the object for the user according to the target dimension.

In a specific application scenario, in order to determine the dimension as the target dimension, as shown in fig. 6, the obtaining unit 33 includes a determining module 331 and an adding module 332.

The determining module 331 may be configured to determine the basis weights of the object for the recommenders with different dimensions.

The adding module 332 may be configured to add the basic weight and the associated weight to obtain a total weight between the user and the different-dimension recommenders of the object.

The determining module 331 may be further configured to determine, as a target dimension, a dimension in which the total weight is greater than a preset weight.

Further, in order to collect the historical behavior data, the collecting unit 31 includes: a collection module 311 and a determination module 312.

The collecting module 311 may be configured to collect first behavior data between the user and the object in different dimensions.

The collecting module 311 may be further configured to determine a target group to which the user belongs, and collect second behavior data between the target group and the object in different dimensions.

The collecting module 311 may be further configured to determine a similar object corresponding to the object, and collect third behavior data of the user and the similar object in different dimensions, and fourth behavior data of the target group and the similar object in different dimensions.

The determining module 312 may be configured to determine the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data as the history correlation data.

Further, in order to determine the association weight between the user and the recommendation in different dimensions of the object, the determining unit 32 includes: a determination module 321 and an addition module 322.

The determining module 321 may be configured to determine the associated weights corresponding to the first behavior data, the second behavior data, the third behavior data, and/or the fourth behavior data, respectively.

The adding module 322 may be configured to add the association weights respectively corresponding to the first behavior data, and/or the second behavior data, and/or the third behavior data, and/or the fourth behavior data, so as to obtain the association weights between the user and the different-dimension recommenders of the object.

In a specific application scenario, the generating unit 34 may be specifically configured to generate a grammar by using a preset recommendation corresponding to the target dimension, and generate the recommendation of the object for the user.

Further, if the target dimension includes at least one of a weight, a material, and a taste corresponding to the object, the generating unit 34 includes: an acquisition module 341 and a generation module 342.

The obtaining module 341 may be configured to obtain raw data of at least one dimension of the weight, the material, and the taste.

The generating module 342 may be configured to generate a recommendation corresponding to the object for the user according to the original data.

Further, if the target dimension includes comment information, the generating unit 34 includes: an acquisition module 341, a filtering module 343, and a splicing module 344.

The obtaining module 341 may be configured to obtain comment information related to the object.

The filtering module 343 may be configured to filter the comment information by using a preset front lexicon, so as to obtain filtered comment information.

The splicing module 344 may be configured to splice the filtered comment information by using a preset grammar model to obtain a recommendation corresponding to the object.

The concatenation module 344 may be further configured to perform word location on the comment words in the comment information by using a preset front lexicon, and concatenate the located comment words to obtain the recommendation corresponding to the object.

Further, if the target dimension includes at least one of an order placing behavior, a good comment behavior, and a browsing behavior, the generating unit 34 includes: a statistics module 345 and a generation module 346.

The statistical module 345 may be configured to perform statistics on the original data in at least one dimension of the ordering behavior, the favorable behavior, and the browsing behavior to obtain a first statistical result corresponding to the object.

The generating module 346 may be configured to generate a recommended language corresponding to the object based on the first statistical result and the first preset grammar model.

Further, if the target dimension includes at least one of a regional rank, an in-store rank, and a quality rank, the generating unit 34 includes: a statistics module 345 and a generation module 346.

The statistical module 345 may be configured to perform statistics on the regional ranking, the in-store ranking, or the quality ranking of the object within a preset time period to obtain a second statistical result corresponding to the object.

The generating module 346 may be configured to generate the recommended word corresponding to the object according to the second statistical result and a second preset grammar model.

Further, a target dimension recommendation for presentation is determined, and the apparatus further includes a statistical unit 35.

The statistical unit 35 may be configured to, if the object has multiple target dimension recommenders, respectively count historical click volumes corresponding to the objects marked with the multiple target dimension recommenders.

The determining unit 32 may be further configured to sort the multiple target dimension recommenders based on the historical click rate, and determine the target dimension recommenders to be displayed according to a sorting result.

Further, in order to highlight the recommended word, the apparatus further includes a determination unit 36 and a presentation unit 37.

The determining unit 36 may be configured to determine whether the recommendation of the object needs to be highlighted according to the target dimension.

The displaying unit 37 may be configured to add a highlighting mark corresponding to the recommended word if it is determined that the recommended word of the object needs to be highlighted, and display the recommended word after the mark is added.

It should be noted that other corresponding descriptions of the functional units related to the apparatus for generating a recommendation provided in this embodiment may refer to the corresponding descriptions in fig. 1 and fig. 2, and are not described herein again.

Based on the methods shown in fig. 1 and fig. 2, correspondingly, the embodiment of the present application further provides a storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the method for generating the recommendation language shown in fig. 1 and fig. 2.

Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.), and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method of the embodiments of the present application.

Based on the method shown in fig. 1 and fig. 2 and the virtual device embodiment shown in fig. 5 and fig. 6, in order to achieve the above object, an embodiment of the present application further provides a computer device, which may specifically be a tablet computer, a smart phone, a smart watch, a smart bracelet, or other network devices, and the computer device includes a storage medium and a processor; a storage medium for storing a computer program; a processor for executing a computer program to implement the method for generating a recommended word as shown in fig. 1 and 2.

Optionally, the above entity devices may further include a user interface, a network interface, a camera, a Radio Frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, and the like. The user interface may include a Display screen (Display), an input unit such as a keypad (Keyboard), etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface), etc.

It will be understood by those skilled in the art that the client device architecture provided in the present embodiment does not constitute a limitation of such physical devices, and may include more or fewer components, or some components in combination, or a different arrangement of components.

The storage medium may further include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the two physical devices described above, supporting the operation of the information processing program as well as other software and/or programs. The network communication module is used for realizing communication among components in the storage medium and communication with other hardware and software in the information processing entity device.

Through the above description of the embodiments, those skilled in the art can clearly understand that by collecting historical association data between users and objects in different dimensions; determining association weights between the user and different dimensionality recommenders of the object according to the historical association data; meanwhile, acquiring a target dimension of which the associated weight meets a preset condition; and finally, generating the recommended words of the objects for the users according to the target dimensions, so that the recommended words under the target dimensions matched with the users can be generated by collecting historical associated data of the users under different dimensions and the objects, the generated recommended words can be ensured to be closely associated with the users, the recommended words are more suitable for the users, the attraction of the objects to the users is increased, and the selection rate of the objects is improved.

Those skilled in the art will appreciate that the figures are merely schematic representations of one preferred implementation scenario and that the blocks or flow diagrams in the figures are not necessarily required to practice the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario may be distributed in the devices in the implementation scenario according to the description of the implementation scenario, or may be located in one or more devices different from the present implementation scenario with corresponding changes. The modules of the implementation scenario may be combined into one module, or may be further split into a plurality of sub-modules.

The above application serial numbers are for description purposes only and do not represent the superiority or inferiority of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of the present application, but the present application is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present application.

Claims

1. A method for generating a recommended language is characterized by comprising the following steps:

2. The method according to claim 1, wherein the obtaining of the target dimension of which the associated weight meets a preset condition comprises:

3. The method of claim 1, wherein the different dimensions comprise a primary dimension or a sum of a primary dimension and a secondary dimension, and wherein the secondary dimension is a sub-dimension corresponding to the primary dimension.

4. The method of claim 1, wherein collecting historical association data between the user and the object in different dimensions comprises:

5. The method of claim 4, wherein determining the association weight between the user and the object's different dimensional recommenders based on the historical association data comprises:

6. The method of claim 1, wherein generating the recommendation of the object for the user according to the target dimension comprises:

7. The method of claim 6, wherein if the target dimension includes at least one of a weight, a material, and a taste of the object, the generating a grammar from a preset recommendation corresponding to the target dimension to generate a recommendation of the object for the user comprises:

8. An apparatus for generating a recommended phrase, comprising:

9. A storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method of any of claims 1 to 7.

10. A computer device comprising a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, characterized in that the processor implements the method of any one of claims 1 to 7 when executing the computer program.