CN116012110A

CN116012110A - Information recommendation method, device, equipment and storage medium

Info

Publication number: CN116012110A
Application number: CN202310065719.2A
Authority: CN
Inventors: 陈华杰; 徐伟盛; 聂双喜; 何吉元; 冯涛; 刘铭
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2023-01-13
Filing date: 2023-01-13
Publication date: 2023-04-25

Abstract

The specification discloses a method, a device, equipment and a storage medium for information recommendation, which can determine commodities which are selected by a user currently and predict the intention of the user according to the commodities which are selected by the user, so that some commodities which accord with the intention of the user in the next step can be selected from other commodities in a collocation relationship or a replacement relationship with the commodities which are selected by the user, the commodities are recommended to the user, and the accuracy of the commodities recommended to the user can be improved.

Description

Information recommendation method, device, equipment and storage medium

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a storage medium for information recommendation.

Background

With the development of electronic commerce, in order to improve the shopping experience of users, a platform may provide the users with goods according to the user's preference.

At present, the main mode of recommending services to users by a platform is as follows: according to the method, the user preference is obtained from the past historical behavior records, and the commodity is recommended to the user according to the user preference, so that the commodity possibly preferred by the user can be recommended to the user, but the method relies on the historical behavior records of the user, and the recommendation effect is poor when the historical behavior records of the user are less.

Disclosure of Invention

The present disclosure provides a method, apparatus, device, and storage medium for information recommendation, so as to partially solve the foregoing problems in the prior art.

The technical scheme adopted in the specification is as follows:

the specification provides a method for recommending information, which comprises the following steps:

determining commodities required by a user at present as basic commodities;

according to the feature data corresponding to the basic commodity, determining a candidate commodity set, wherein the candidate commodity set comprises: the characteristic data corresponding to the basic commodity comprises collocation characteristics determined through a pre-established first relation diagram and replacement characteristics determined through a pre-established second relation diagram, wherein the first relation diagram is used for representing collocation relations among the commodities, the second relation diagram is used for representing replacement relations among the commodities, and the first relation diagram and the second relation diagram are established through historical ordering behaviors and historical searching behaviors of each user;

and recommending information to the user according to the feature data corresponding to each candidate commodity in the candidate commodity set.

Optionally, constructing the first relationship diagram and the second relationship diagram specifically includes:

constructing each order node according to the historical ordering file of each user and constructing each search node according to the historical searching behavior of each user, wherein the order node comprises behavior information of one historical ordering behavior, and the search node comprises behavior information of one historical searching behavior;

for each commodity, constructing a commodity node corresponding to the commodity, determining an order node corresponding to the commodity and a search node corresponding to the commodity, connecting the commodity node corresponding to the commodity with the order node corresponding to the commodity, determining a commodity collocation relation diagram, connecting the commodity node of the commodity with the search node corresponding to the commodity, and determining a commodity replacement relation diagram;

taking a commodity corresponding to a commodity node as a constraint condition, and fusing the commodity collocation relation diagram and the commodity replacement relation diagram to obtain a commodity relation diagram;

and constructing the first relation diagram and the second relation diagram according to the commodity relation diagram.

Optionally, constructing the first relationship diagram according to the commodity relationship diagram specifically includes:

Determining at least one order node connected with each commodity node in the commodity relation diagram aiming at each commodity node;

determining at least one commodity node connected with the order node except the commodity node as a collocation node corresponding to the commodity node according to the at least one order node;

determining the first relation graph according to each commodity node and the collocation node corresponding to each commodity;

constructing the second relation diagram according to the commodity relation diagram, wherein the construction method specifically comprises the following steps:

determining at least one searching node connected with each commodity node in the commodity relation diagram aiming at each commodity node;

determining at least one commodity node connected with the searching node except the commodity node according to the at least one searching node, and taking the commodity node as a replacement node corresponding to the commodity node;

and constructing the second relation graph according to the commodity node and the replacement node corresponding to the commodity node.

Optionally, determining the matching features corresponding to each commodity through a pre-constructed first relation diagram specifically includes:

acquiring initial characteristic representations corresponding to all commodities;

inputting the initial characteristic representations corresponding to the commodities into a characteristic extraction model to be trained, adjusting the initial characteristic representations corresponding to the commodities through training the characteristic extraction model, and obtaining the characteristic representations corresponding to the commodities after the characteristic extraction model is trained;

And determining collocation features and replacement features corresponding to each commodity according to the feature representation, the first relation diagram and the second relation diagram corresponding to each commodity.

Optionally, the training of the feature extraction model is used to adjust the initial feature representation corresponding to each commodity, and after the feature extraction model is trained, the feature representation corresponding to each commodity is obtained, which specifically includes:

aiming at each commodity, training the feature extraction model by taking the larger the similarity between the matching feature corresponding to the commodity and the matching feature corresponding to the matching commodity of the commodity and the larger the similarity between the replacement feature corresponding to the commodity and the replacement feature corresponding to the replacement commodity of the commodity as an optimization target, and adjusting the feature representation to be adjusted corresponding to the commodity so as to obtain the feature representation corresponding to each commodity after the feature extraction model is trained.

And training the feature extraction model by alternately adopting a first training mode and a second training mode so as to adjust initial feature representations corresponding to the commodities, wherein the first training mode is to train the feature extraction model by using the feature representations of the matched commodities, and the second training mode is to train the feature extraction model by using the feature representations of the replaced commodities.

Optionally, determining the collocation feature and the replacement feature corresponding to each commodity according to the feature representation, the first relationship diagram and the second relationship diagram corresponding to each commodity specifically includes:

determining the commodity with the collocation relation with each commodity according to the first relation diagram, and fusing the characteristic representation corresponding to the commodity with the collocation relation with the commodity with the characteristic representation corresponding to the commodity to obtain collocation characteristics corresponding to the commodity; and

and determining the commodity with the replacement relation according to the second relation diagram for each commodity, and fusing the characteristic representation corresponding to the commodity with the replacement relation with the characteristic representation corresponding to the commodity to obtain the replacement characteristic corresponding to the commodity.

The present specification provides an apparatus for information recommendation, including:

the acquisition module is used for determining the commodity required by the user at present as a basic commodity;

the determining module is used for determining a candidate commodity set according to the feature data corresponding to the basic commodity, wherein the candidate commodity set comprises: the characteristic data corresponding to the basic commodity comprises collocation characteristics determined through a pre-established first relation diagram and replacement characteristics determined through a pre-established second relation diagram, wherein the first relation diagram is used for representing collocation relations among the commodities, the second relation diagram is used for representing replacement relations among the commodities, and the first relation diagram and the second relation diagram are established through historical ordering behaviors and historical searching behaviors of each user;

and the recommending module is used for recommending information to the user according to the characteristic data corresponding to each candidate commodity in the candidate commodity set.

The present specification provides a computer readable storage medium storing a computer program which when executed by a processor implements the method of information recommendation described above.

The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of information recommendation described above when executing the program.

The above-mentioned at least one technical scheme that this specification adopted can reach following beneficial effect:

in the information recommendation method provided by the specification, firstly, a commodity required by a user at present is determined as a basic commodity, and a candidate commodity set is determined according to feature data corresponding to the basic commodity, wherein the candidate commodity set comprises: the feature data corresponding to the basic commodity comprises matching features determined through a pre-built first relation diagram and replacement features determined through a pre-built second relation diagram, wherein the first relation diagram is used for representing the matching relation among the commodities, the second relation diagram is used for representing the replacement relation among the commodities, the first relation diagram and the second relation diagram are built through historical ordering behaviors and historical searching behaviors of the users, and information recommendation is carried out on the feature data corresponding to the candidate commodities in the candidate commodity set.

According to the method, the commodities which are selected by the user at present can be determined, the intention of the user is predicted according to the commodities which are selected by the user, so that some commodities which accord with the intention of the user in the next step can be selected from other commodities which have a collocation relation or a replacement relation with the commodities which are selected by the user, the commodities are recommended to the user, and the accuracy of the commodities recommended to the user can be improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the specification, illustrate and explain the exemplary embodiments of the present specification and their description, are not intended to limit the specification unduly. In the drawings:

FIG. 1 is a flow chart of a method for recommending information provided in the present specification;

FIG. 2 is a schematic diagram of the construction process of the commodity relation diagram provided in the present specification;

FIG. 3 is a schematic diagram of an apparatus for recommending information according to the present disclosure;

fig. 4 is a schematic diagram of an electronic device corresponding to fig. 1 provided in the present specification.

Detailed Description

For the purposes of making the objects, technical solutions and advantages of the present specification more apparent, the technical solutions of the present specification will be clearly and completely described below with reference to specific embodiments of the present specification and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present specification. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.

The following describes in detail the technical solutions provided by the embodiments of the present specification with reference to the accompanying drawings.

Fig. 1 is a flow chart of a method for recommending information, which includes the following steps:

s101: and determining the commodity currently required by the user as a basic commodity.

In the present specification, the execution body for implementing the information recommendation method may refer to a designated device such as a server provided on a service platform, or may refer to a terminal device such as a desktop computer or a notebook computer, and for convenience of description, the information recommendation method provided in the present specification will be described below by taking the server as an example.

In the specification, the server can determine the commodity required by the user according to the service behaviors of the user in different service scenes, and can further recommend the commodity for the user according to the determined commodity required by the user.

The service scenario may include: and the commodity detail page browses the service scenes such as a shopping cart management scene, a commodity searching recommendation scene and the like.

Specifically, the user can enter a commodity searching recommendation scene by inputting a search word in a designated area, the server can determine the commodity matched with the search word input by the user according to the search word input by the user, and recommend commodity information of the commodity matched with the search word input by the user to the user.

For example: the search term input by the user is "A-brand beef", at this time, the server can determine commodity information of A-brand beef on shelves of each merchant corresponding to the "A-brand beef" input by the user, in addition, the server can also recommend the A-brand beef to the user as a basic commodity, further determine matched commodities (such as potatoes, persimmons and other commodities matched with the A-brand beef) and replaced commodities (such as A-brand mutton and other commodities capable of replacing the A-brand beef) of the basic commodity, further can take the matched commodities and the replaced commodities of the A-brand beef as part of recall results of the search term input by the user, further can sequence the commodities matched with the search term input by the user through a sequencing model, and recommend the replaced commodities and the matched commodities to the user after sequencing.

Similarly, when the user opens the commodity detail page of a commodity to browse, the server can recommend some replacement commodities of the commodity to the user at the bottom of the commodity detail page of the commodity, at this time, the commodity displayed on the commodity detail page opened by the user is the commodity required by the user at present, and the server can take the commodity displayed on the commodity detail page opened by the user as the basic commodity.

For example: the user is browsing the product detail page of the brand B orange juice, at this time, a replacement product of the brand B orange juice may be recommended to the user at the bottom of the product detail page of the brand B orange juice, such as: orange juice brand C, lemon water brand D, etc.

Similarly, after a certain commodity is added to the shopping cart by the user, when the user views the commodity in the shopping cart page, the server can recommend some collocation commodities matched with the commodity currently added to the shopping cart by the user through the shopping cart page assembly such as the popup frame, at this time, the commodity added to the shopping cart by the user is the commodity currently required by the user, and the server can take the commodity added to the shopping cart by the user as a basic commodity.

For example: after the tomatoes are added to the shopping cart by the user, when the user opens the shopping cart management interface, some matched commodities of the tomatoes added to the shopping cart by the user can be recommended to the user through popup windows and other components under the commodities added to the shopping cart by the user, for example: can be matched with tomatoes to prepare eggs for stir-frying the eggs with the tomatoes, etc.

S102: according to the feature data corresponding to the basic commodity, determining a candidate commodity set, wherein the candidate commodity set comprises: the feature data corresponding to the basic commodity comprises collocation features determined through a pre-established first relation diagram and replacement features determined through a pre-established second relation diagram, wherein the first relation diagram is used for representing collocation relations among the commodities, the second relation diagram is used for representing replacement relations among the commodities, and the first relation diagram and the second relation diagram are established through historical ordering behaviors and historical searching behaviors of users.

Further, after determining the basic commodity, the server may determine a candidate commodity set of the basic commodity according to feature data corresponding to the basic commodity, where the candidate commodity set includes: at least one of a commodity having a collocation relationship with the base commodity (i.e., a collocation commodity of the base commodity) and a commodity having a replacement relationship with the base commodity (i.e., a replacement commodity of the base commodity).

In the above, the feature data corresponding to the basic commodity includes a collocation feature and a replacement feature, and the feature data corresponding to the basic commodity is determined in advance according to the first relationship diagram and the second relationship diagram that are constructed.

Specifically, the server may obtain the historical ordering behavior and the historical searching behavior of each user, and may further construct each order node according to the historical ordering behavior of each user, and construct each searching node according to the historical searching behavior of each user, where the order node includes behavior information of one historical ordering behavior, and the searching node includes behavior information of one historical searching behavior.

It should be noted that, in an actual scenario, there may be some commodities because of such things as: for reasons of new marketing, low sales and the like, abundant historical ordering behaviors and historical searching behaviors are lacking, and for the part of commodities, manually-arranged historical ordering behaviors and historical searching behaviors corresponding to the commodities with the replacement relation can be obtained, so that the historical ordering behaviors and the historical searching behaviors corresponding to the part of commodities can be generated according to the historical ordering behaviors and the historical searching behaviors corresponding to the replacement commodities of the part of commodities, and the richness of characteristic data corresponding to the part of commodities can be improved.

Further, the server may construct a commodity node corresponding to each commodity, determine an order node corresponding to the commodity and determine a search node corresponding to the commodity, connect the commodity node corresponding to the commodity with the order node corresponding to the commodity, determine a commodity collocation relationship graph, connect the commodity node of the commodity with the search node corresponding to the commodity, determine a commodity replacement relationship graph, and fuse the commodity collocation relationship graph and the commodity replacement relationship graph with one commodity node corresponding to one commodity as a constraint condition to obtain a commodity relationship graph, and construct a first relationship graph and a second relationship graph according to the commodity relationship graph, as shown in fig. 2.

Fig. 2 is a schematic diagram of a process for constructing a commodity relational graph provided in the present specification.

In fig. 2, white nodes are commodity nodes, light gray nodes are order nodes, dark gray nodes are search nodes, edges between commodity nodes and order nodes are single edges used for representing the ordering relation, when three commodities such as potatoes, beef and bell peppers are purchased simultaneously in an order corresponding to one order node, the three commodities are considered to have a collocation relation, and the order node corresponding to the order is respectively connected with the commodity node corresponding to the potatoes, the commodity node corresponding to the beef and the commodity node corresponding to the bell peppers through the three order edges.

Similarly, an edge between a commodity node and a search node is a search edge that is used to characterize a search relationship.

Further, after the commodity collocation relation diagram and the commodity replacement relation diagram are constructed, the commodity collocation relation diagram and the commodity replacement relation diagram can be fused by taking a commodity corresponding to a commodity node as a constraint condition, so that the commodity relation diagram is obtained, in other words, the commodity node corresponding to the commodity collocation relation diagram and the commodity node corresponding to the commodity replacement relation diagram of the same commodity are fused into one commodity node, and at the moment, one commodity node is connected with the next single node through a lower single side and is also connected with the search node through the search side.

Further, at least one order node connected with the commodity node in the commodity relation graph can be determined for each commodity node, at least one commodity node connected with the order node except the commodity node is determined according to the determined at least one order node, the commodity node is used as a collocation node corresponding to the commodity node, and the first relation graph is determined according to each commodity node and the collocation node corresponding to each commodity.

It should be noted that, the first relationship graph only includes commodity nodes, but does not include the ordering node and the searching node, and at the same time, an edge between two commodity nodes in the first relationship graph is only used for representing that a collocation relationship exists between commodities corresponding to the two commodity nodes.

Similarly, for each commodity node, at least one searching node connected with the commodity node in the commodity relation diagram can be determined, at least one commodity node connected with the searching node except the commodity node is determined according to the determined at least one searching node, the commodity node is used as a replacement node corresponding to the commodity node, and a second relation diagram is constructed according to the commodity node and the replacement node corresponding to the commodity node.

It should be noted that, the second relationship graph only includes commodity nodes, but does not include the ordering node and the searching node, and at the same time, an edge between two commodity nodes in the second relationship graph is only used for representing that a replacement relationship exists between commodities corresponding to the two commodity nodes.

Further, after the first relation diagram and the second relation diagram are constructed, the server can acquire initial feature representations corresponding to all commodities, input the initial feature representations corresponding to all commodities into a feature extraction model to be trained, adjust the initial feature representations corresponding to all the commodities through training of the feature extraction model, acquire feature representations corresponding to all the commodities after the feature extraction model is trained, and determine matching features and replacement features corresponding to all the commodities according to the feature representations corresponding to all the commodities, the first relation diagram and the second relation diagram.

For each commodity, the initial feature representation corresponding to the commodity can be obtained by encoding according to each commodity information of the commodity in advance to obtain encoded data corresponding to each commodity information, and then splicing and fusing the encoded data corresponding to each commodity information of the commodity.

The server can train the feature extraction model and update the initial feature representation of each commodity, so that the feature representation corresponding to each commodity can be obtained after the feature extraction model is trained.

In the foregoing, the method for adjusting the initial feature representation of each commodity while training the feature extraction model by the server may be that, for each commodity, the greater the similarity between the matching feature corresponding to the commodity and the matching feature corresponding to the matching commodity of the commodity, and the greater the similarity between the replacement feature corresponding to the commodity and the replacement feature corresponding to the replacement commodity of the commodity are used as optimization targets, the feature extraction model is trained, and the feature representation to be adjusted corresponding to the commodity is adjusted, so that after the feature extraction model is trained, the feature representation corresponding to each commodity is obtained.

The server may alternatively adopt a first training mode and a second training mode, where the first training mode is to train the feature extraction model by using the feature representation of the matching commodity, and the second training mode is to train the feature extraction model by using the feature representation of the replacement commodity, so as to adjust the initial feature representation corresponding to each commodity.

Specifically, the server can adjust the initial feature representation corresponding to each commodity through multiple rounds of training of the feature extraction model, so as to obtain the feature representation corresponding to each commodity after the feature extraction model is trained, and each round of training of the feature extraction model can adjust the initial feature representation of each commodity.

And determining matching features corresponding to the commodities according to the similarity between the matching features corresponding to the commodities and matching features corresponding to the matching commodities of the commodities according to a first relation diagram, determining to-be-adjusted feature representation of each commodity (the to-be-adjusted feature representation corresponding to the commodity is obtained by iterating an initial feature representation corresponding to the commodity to a previous training round), wherein N is an integer not smaller than 0, determining the matching features corresponding to the commodity according to the to-be-adjusted feature representation corresponding to the commodity and the feature representation corresponding to the matching commodity of the commodity for each commodity, and training the feature extraction model and adjusting the to-be-adjusted feature representation corresponding to the commodity according to the similarity between the matching features corresponding to the commodity and the matching features corresponding to the matching commodity compared with the similarity between the matching features corresponding to the commodity and the matching features of other commodities except the matching commodity.

And aiming at 2N+2 round adjustment, determining a replacement commodity of each commodity according to a second relation diagram through a characteristic extraction model, determining a to-be-adjusted characteristic representation of each commodity, aiming at each commodity, determining a replacement characteristic corresponding to the commodity based on the to-be-adjusted characteristic representation corresponding to the commodity and a characteristic representation corresponding to the replacement commodity of the commodity, determining the similarity between the replacement characteristic corresponding to the commodity and the replacement characteristic corresponding to the replacement commodity of the commodity, training the characteristic extraction model and adjusting the to-be-adjusted characteristic representation corresponding to the commodity according to the similarity between the replacement characteristic corresponding to the commodity and the replacement characteristic corresponding to the replacement commodity of the commodity, wherein the greater the similarity between the replacement characteristic corresponding to the commodity and the replacement characteristic of other commodities except the replacement commodity of the commodity is an optimization target.

In the foregoing, the server may further preset a first weight corresponding to the first training manner and a second weight corresponding to the second training manner, train the feature vector model according to the first weight and the second weight, and adjust the feature representation corresponding to each commodity.

In the foregoing, the method for determining the matching feature and the replacement feature corresponding to each commodity by the server according to the feature representation, the first relationship diagram and the second relationship diagram corresponding to each commodity may be that, for each commodity, a commodity having a matching relationship with the commodity is determined according to the first relationship diagram, and the feature representation corresponding to the commodity having the matching relationship with the commodity is fused with the feature representation corresponding to the commodity, so as to obtain the matching feature corresponding to the commodity.

The following detailed description of the process of fusing the feature representation corresponding to the commodity with the matching relationship with the feature representation corresponding to the commodity is given by combining the formula, and the following formula may be specifically referred to:

wherein, the liquid crystal display device comprises a liquid crystal display device,

namely, the matching characteristics corresponding to the obtained commodity after fusion are->

And->

Super-parameters for learning the feature extraction model during training>

For the characteristic representation of the commodity, < > a->

In order to represent the feature corresponding to the commodity with collocation relation, concate is a splicing function.

As can be seen from the above formula, the server may fuse the feature representation corresponding to each commodity with the feature representation corresponding to the matching commodity of each commodity, so as to obtain the matching feature of each commodity.

Further, the server may further perform normalization processing on the collocation feature of each commodity after fusing the feature representation corresponding to each commodity and the feature representation corresponding to the collocation commodity of each commodity to obtain the collocation feature of each commodity, so as to obtain the collocation feature after normalization processing, and use the collocation feature after normalization processing as the collocation feature corresponding to each commodity finally.

It should be noted that, the server may further perform fusion on the feature representation corresponding to the commodity and the feature representation corresponding to the matched commodity of the commodity through multiple rounds of fusion, where for each round of fusion, the feature to be adjusted in the round of fusion is determined (where the feature to be adjusted is obtained by iterating the corresponding feature representation of the commodity to the previous round), the commodity having the matching relationship with the commodity is determined according to the first relationship diagram, and the feature representation corresponding to the commodity having the matching relationship with the feature representation corresponding to the commodity is fused with the feature to be adjusted corresponding to the commodity, so as to obtain the matching feature corresponding to the commodity output after the round of fusion.

The method for determining the replacement feature of each commodity is identical to the method for determining the matching feature of each commodity, and will not be described in detail herein.

S103: and recommending information to the user according to the feature data corresponding to each candidate commodity in the candidate commodity set.

Further, after determining the candidate commodity set according to the feature data corresponding to the basic commodity, the server may recommend information to the user according to the feature data corresponding to each candidate commodity in the candidate commodity set.

Specifically, if the candidate commodity set includes commodities having a matching relationship with the basic commodity, determining matching probability of the basic commodity and each candidate commodity according to matching characteristics corresponding to each candidate commodity in the candidate commodity set, and recommending information to a user according to the matching probability.

If the candidate commodity set contains commodities with a replacement relation with the basic commodity, determining the replacement probability of the basic commodity and each candidate commodity according to the replacement characteristics corresponding to each candidate commodity in the candidate commodity set, and recommending information to the user according to the replacement probability.

If the candidate commodity set comprises commodities with collocation relation with the basic commodity and commodities with replacement relation with the basic commodity, determining the replacement probability of the basic commodity and each candidate commodity according to the replacement characteristics corresponding to each candidate commodity in the candidate commodity set, and recommending information to a user according to the collocation probability and the replacement probability.

It should be noted that, the method for recommending information to the user by the server according to the collocation probability and/or the replacement probability may be that the server determines the weight of each commodity in the candidate commodity set according to the historical purchasing behavior and the historical searching behavior of the user and the collocation probability and/or the replacement probability of each commodity in each candidate commodity set through a preset ordering model, so that each commodity can be ordered according to the weight of each commodity, and then information recommendation can be performed to the user according to each ordered commodity.

If the current business scenario of the user is a browsing scenario of the commodity detail page, the server needs to determine a replacement commodity of the basic commodity for the basic commodity as a candidate commodity set, and at this time, the server may determine the replacement commodity of the basic commodity from the commodities according to the replacement characteristics of the basic commodity and the replacement characteristics of the commodities.

Similarly, if the current business scenario of the user is a shopping cart management scenario, the server needs to determine the matching commodity of the basic commodity for the basic commodity as the candidate commodity set, and at this time, the server can determine the matching commodity of the basic commodity from the commodities according to the matching characteristics of the basic commodity and the matching characteristics of the commodities.

Similarly, if the current business scenario of the user is a commodity search recommendation scenario, the server needs to determine the matched commodity and the replaced commodity of the basic commodity for the basic commodity as the candidate commodity set, and at this time, the server can determine the matched commodity and the replaced commodity of the basic commodity from the commodities according to the matched characteristic and the replaced characteristic of the basic commodity and the matched characteristic and the replaced characteristic of each commodity.

It should be noted that, the server may perform mutual complementation of the relationship when determining the matching commodity and the replacement commodity of the basic commodity from the commodities by using the matching characteristic and the replacement characteristic of the basic commodity and the matching characteristic and the replacement characteristic of each commodity, so as to effectively recommend some commodities with poor correlation with other commodities.

For example: if some commodities are not matched or few commodities can be matched, the matched commodities of the commodities can be supplemented according to the matched commodities of the replacement commodities of the commodities, for example: milk and bread are matched with each other, and milk and goat milk are replaced with each other, so that the goat milk and bread can be pushed out to have a matching relationship.

In the foregoing, the method for determining the matched commodity and/or the replaced commodity of the basic commodity from each commodity by the server according to the matched characteristic and/or the replaced characteristic of the basic commodity and the matched characteristic and/or the replaced characteristic of each commodity may be as follows: and adopting algorithms such as a nearest neighbor search algorithm and the like, and determining the matched commodity and/or the replaced commodity of the basic commodity from the commodities according to the matched characteristic and/or the replaced characteristic of the basic commodity and the matched characteristic and/or the replaced characteristic of each commodity.

As can be seen from the above, the server can adjust the initial characteristic representation of each commodity by alternately using the characteristic representation of the matching commodity and the characteristic representation of the replacement commodity, so that the determined characteristic representation corresponding to each commodity can include information of other commodities with the matching relationship and information of other commodities with the replacement relationship.

In addition, the server can predict the intention of the user according to the selected commodity, so that some commodities conforming to the intention of the user can be selected from other commodities in collocation relation or replacement relation with the commodity selected by the user, the commodities are recommended to the user, and the accuracy of the commodities recommended to the user can be improved

Fig. 3 is a schematic diagram of an apparatus for information recommendation provided in the present specification, including:

an acquisition module 301, configured to determine an item currently required by a user as a basic item;

the determining module 302 is configured to determine a candidate commodity set according to the feature data corresponding to the basic commodity, where the candidate commodity set includes: the characteristic data corresponding to the basic commodity comprises collocation characteristics determined through a pre-established first relation diagram and replacement characteristics determined through a pre-established second relation diagram, wherein the first relation diagram is used for representing collocation relations among the commodities, the second relation diagram is used for representing replacement relations among the commodities, and the first relation diagram and the second relation diagram are established through historical ordering behaviors and historical searching behaviors of each user;

And a recommending module 303, configured to recommend information to the user according to feature data corresponding to each candidate commodity in the candidate commodity set.

Optionally, the apparatus further comprises: a construction module 304;

the building module 304 is specifically configured to build each order node according to the historical ordering list of each user, and build each search node according to the historical searching behavior of each user, where the order node includes behavior information of one historical ordering behavior, and the search node includes behavior information of one historical searching behavior; for each commodity, constructing a commodity node corresponding to the commodity, determining an order node corresponding to the commodity and a search node corresponding to the commodity, connecting the commodity node corresponding to the commodity with the order node corresponding to the commodity, determining a commodity collocation relation diagram, connecting the commodity node of the commodity with the search node corresponding to the commodity, and determining a commodity replacement relation diagram; taking a commodity corresponding to a commodity node as a constraint condition, and fusing the commodity collocation relation diagram and the commodity replacement relation diagram to obtain a commodity relation diagram; and constructing the first relation diagram and the second relation diagram according to the commodity relation diagram.

Optionally, the construction module 304 is specifically configured to determine, for each commodity node, at least one order node connected to the commodity node in the commodity relationship diagram; determining at least one commodity node connected with the order node except the commodity node as a collocation node corresponding to the commodity node according to the at least one order node; determining the first relation graph according to each commodity node and the collocation node corresponding to each commodity; determining at least one searching node connected with each commodity node in the commodity relation diagram aiming at each commodity node; determining at least one commodity node connected with the searching node except the commodity node according to the at least one searching node, and taking the commodity node as a replacement node corresponding to the commodity node; and constructing the second relation graph according to the commodity node and the replacement node corresponding to the commodity node.

Optionally, the apparatus further comprises: a feature extraction module 305;

the feature extraction module 305 is specifically configured to obtain an initial feature representation corresponding to each commodity; inputting the initial characteristic representations corresponding to the commodities into a characteristic extraction model to be trained, adjusting the initial characteristic representations corresponding to the commodities through training the characteristic extraction model, and obtaining the characteristic representations corresponding to the commodities after the characteristic extraction model is trained; and determining collocation features and replacement features corresponding to each commodity according to the feature representation, the first relation diagram and the second relation diagram corresponding to each commodity.

Optionally, the feature extraction module 305 is specifically configured to train the feature extraction model with respect to each commodity, and train the feature extraction model with respect to the greater the similarity between the matching feature corresponding to the commodity and the matching feature corresponding to the matching commodity of the commodity and with respect to the greater the similarity between the replacement feature corresponding to the commodity and the replacement feature corresponding to the replacement commodity of the commodity as an optimization target, and adjust the feature representation to be adjusted corresponding to the commodity, so as to obtain the feature representation corresponding to each commodity after the feature extraction model is trained.

Optionally, the feature extraction module 305 is specifically configured to alternately use a first training manner and a second training manner to train the feature extraction model, so as to adjust the initial feature representation corresponding to each commodity, where the first training manner is to train the feature extraction model by using the feature representation of the matching commodity, and the second training manner is to train the feature extraction model by using the feature representation of the replacement commodity.

Optionally, the feature extraction module 305 is specifically configured to determine, for each commodity according to the first relationship diagram, a commodity having the matching relationship with the commodity, and fuse a feature representation corresponding to the commodity and a feature representation corresponding to the commodity to obtain a matching feature corresponding to the commodity; and determining the commodity with the replacement relation according to the second relation diagram for each commodity, and fusing the characteristic representation corresponding to the commodity with the replacement relation with the characteristic representation corresponding to the commodity to obtain the replacement characteristic corresponding to the commodity.

The present specification also provides a computer readable storage medium storing a computer program operable to perform a method of information recommendation as provided in fig. 1 above.

The present specification also provides a schematic structural diagram of an electronic device corresponding to fig. 1 shown in fig. 4. At the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile storage, as described in fig. 4, although other hardware required by other services may be included. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs the same to implement the information recommendation method described above with respect to fig. 1. Of course, other implementations, such as logic devices or combinations of hardware and software, are not excluded from the present description, that is, the execution subject of the following processing flows is not limited to each logic unit, but may be hardware or logic devices.

In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable Gate Array, FPGA)) is an integrated circuit whose logic function is determined by the programming of the device by a user. A designer programs to "integrate" a digital system onto a PLD without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented by using "logic compiler" software, which is similar to the software compiler used in program development and writing, and the original code before the compiling is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but not just one of the hdds, but a plurality of kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), lava, lola, myHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers, and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller may thus be regarded as a kind of hardware component, and means for performing various functions included therein may also be regarded as structures within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.

The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functions of each element may be implemented in one or more software and/or hardware elements when implemented in the present specification.

It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present description is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.

It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.

The foregoing is merely exemplary of the present disclosure and is not intended to limit the disclosure. Various modifications and alterations to this specification will become apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, or the like, which are within the spirit and principles of the present description, are intended to be included within the scope of the claims of the present description.

Claims

1. A method of information recommendation, comprising:

determining commodities required by a user at present as basic commodities;

2. The method of claim 1, wherein constructing the first relationship graph and the second relationship graph specifically comprises:

3. The method of claim 2, wherein constructing the first relationship graph from the commodity relationship graph specifically comprises:

4. The method of claim 1, wherein determining the collocation feature corresponding to each commodity through the pre-constructed first relationship diagram specifically comprises:

5. The method of claim 4, wherein the training of the feature extraction model adjusts the initial feature representation corresponding to each commodity, and after training the feature extraction model, obtains the feature representation corresponding to each commodity, specifically comprising:

6. The method of claim 4, wherein the training of the feature extraction model adjusts the initial feature representation corresponding to each commodity, and after training the feature extraction model, obtains the feature representation corresponding to each commodity, specifically comprising:

7. The method of claim 4, wherein determining the collocation feature and the replacement feature for each commodity according to the feature representation, the first relationship diagram, and the second relationship diagram for each commodity, specifically comprises:

8. An apparatus for information recommendation, comprising:

9. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-7.

10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method of any of the preceding claims 1-7 when executing the program.