CN116340638A

CN116340638A - Method and device for determining interaction result

Info

Publication number: CN116340638A
Application number: CN202310339462.5A
Authority: CN
Inventors: 请求不公布姓名
Original assignee: Shenzhen Xumi Yuntu Space Technology Co Ltd
Current assignee: Shenzhen Xumi Yuntu Space Technology Co Ltd
Priority date: 2023-03-24
Filing date: 2023-03-24
Publication date: 2023-06-27

Abstract

The disclosure relates to the technical field of artificial intelligence, and provides a method and a device for determining an interaction result, computer equipment and a computer readable storage medium. In the process of determining the interaction result corresponding to the interaction information to be processed, the method considers not only the similarity in the feature domain of the original feature vectors corresponding to each interaction information to be processed, but also the relevance among the enhanced feature vectors corresponding to each original feature vector, so that the accuracy of the determined interaction result corresponding to the interaction information to be processed can be improved, the accuracy of CTR estimation in a recommendation scene can be effectively improved, and user experience and the efficiency of the whole recommendation system can be improved.

Description

Method and device for determining interaction result

Technical Field

The disclosure relates to the technical field of artificial intelligence, and in particular relates to a method and a device for determining an interaction result.

Background

The recommendation system plays an indispensable role in the life today, and has the physical and physical effects of online shopping, news reading and video watching. User click prediction (Click Through Rate, CTR) is a critical task in a recommendation system that can be implemented to estimate the probability that a user clicks on a recommended object (e.g., merchandise, news push message, push video). Taking a recommended object as an example, CTR estimation is used as a ring of keys of a ranking link of a recommendation system, and objects most likely to be clicked by a user can be preferentially pushed to the user, so that the satisfaction degree of the user and the efficiency of the whole recommendation system can be improved.

The existing CTR estimation model is a shallow model or a depth model, and only the intersections among different characteristic domains are considered in the CTR result prediction process; therefore, the accuracy of the whole CTR estimation is affected, the accuracy of the CTR estimation result is greatly reduced, and therefore commodities, news or videos recommended to the user are not really wanted commodities, news or videos by the user, further user experience is poor, and the efficiency of the whole recommendation system is affected.

Disclosure of Invention

In view of this, the embodiments of the present disclosure provide a method, an apparatus, a computer device, and a computer readable storage medium for determining an interaction result, so as to solve the problems in the prior art that, because only the intersections between different feature domains are considered in the process of predicting a CTR result, the accuracy of the overall CTR prediction is affected, resulting in a great decrease in the accuracy of the CTR prediction result, and thus, the recommended merchandise, news, or video for the user is not really the merchandise, news, or video that the user really wants, and further, the user experience is poor, and the efficiency of the whole recommendation system is affected.

In a first aspect of an embodiment of the present disclosure, a method for determining an interaction result is provided, where the method includes:

Acquiring interaction information to be processed;

according to the interaction information to be processed, determining a plurality of original feature vectors corresponding to the interaction information to be processed;

for each original feature vector, determining an enhanced feature vector corresponding to the original feature vector;

and obtaining an interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector.

In a second aspect of the embodiments of the present disclosure, there is provided an apparatus for determining an interaction result, the apparatus including:

the information acquisition unit is used for acquiring the interaction information to be processed;

the first determining unit is used for determining a plurality of original feature vectors corresponding to the interaction information to be processed according to the interaction information to be processed;

a second determining unit, configured to determine, for each original feature vector, an enhanced feature vector corresponding to the original feature vector;

and the result determining unit is used for obtaining the interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector.

In a third aspect of the disclosed embodiments, a computer device is provided, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the above method when the computer program is executed.

In a fourth aspect of the disclosed embodiments, a computer-readable storage medium is provided, which stores a computer program which, when executed by a processor, implements the steps of the above-described method.

Compared with the prior art, the embodiment of the disclosure has the beneficial effects that: the embodiment of the disclosure can firstly acquire the interaction information to be processed; then, according to the interaction information to be processed, a plurality of original feature vectors corresponding to the interaction information to be processed can be determined; next, for each original feature vector, an enhanced feature vector corresponding to the original feature vector may be determined; and then, according to the enhanced feature vectors corresponding to the original feature vectors, obtaining the interaction result corresponding to the interaction information to be processed. It can be seen that, in this embodiment, after the original feature vector corresponding to each piece of interaction information to be processed is obtained, the enhancement feature vector corresponding to each original feature vector can be determined first, that is, in this embodiment, the information modeling in the feature domain can be performed on the original feature vector corresponding to each piece of interaction information to be processed, so as to obtain the enhancement feature vector corresponding to the original feature vector, so that the enhancement feature vector corresponding to the original feature vector can fully embody the accuracy and richness of the feature domain semantics of the original feature vector; therefore, in the process of determining the interaction result corresponding to the interaction information to be processed according to the enhancement feature vectors respectively corresponding to the original feature vectors, the similarity in the feature domain of the original feature vector corresponding to the interaction information to be processed can be fully considered. In this way, in the process of determining the interaction result corresponding to the interaction information to be processed in this embodiment, not only the similarity in the feature domain of the original feature vectors corresponding to each respective interaction information to be processed is considered, but also the relevance between the enhanced feature vectors corresponding to each respective original feature vector is considered, so that the accuracy of the determined interaction result corresponding to the interaction information to be processed can be improved, the accuracy of CTR estimation in the recommendation scene can be effectively improved, and further the user experience can be improved, and the efficiency of the whole recommendation system can be improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings that are required for the embodiments or the description of the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and other drawings may be obtained according to these drawings without inventive effort for a person of ordinary skill in the art.

Fig. 1 is a scene schematic diagram of an application scene of an embodiment of the present disclosure;

FIG. 2 is a flow chart of a method of determining interaction results provided by embodiments of the present disclosure;

FIG. 3 is a block diagram of an apparatus for determining interaction results provided by an embodiment of the present disclosure;

fig. 4 is a schematic diagram of a computer device provided by an embodiment of the present disclosure.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system configurations, techniques, etc. in order to provide a thorough understanding of the disclosed embodiments. However, it will be apparent to one skilled in the art that the present disclosure may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present disclosure with unnecessary detail.

A method and apparatus for determining an interaction result according to embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

In the prior art, because the current CTR estimation model is a shallow model or a depth model, only the intersections among different feature domains are considered in the CTR result prediction process; therefore, the accuracy of the whole CTR estimation is affected, the accuracy of the CTR estimation result is greatly reduced, and therefore commodities, news or videos recommended to the user are not really wanted commodities, news or videos by the user, further user experience is poor, and the efficiency of the whole recommendation system is affected.

In order to solve the above problems. In the method, since the original feature vector corresponding to each piece of interaction information to be processed can be obtained, the enhancement feature vector corresponding to each original feature vector can be determined first, that is, the embodiment can model the information in the feature domain of the original feature vector corresponding to each piece of interaction information to be processed to obtain the enhancement feature vector corresponding to the original feature vector, so that the enhancement feature vector corresponding to the original feature vector can fully embody the accuracy and richness of the feature domain semantics of the original feature vector; therefore, in the process of determining the interaction result corresponding to the interaction information to be processed according to the enhancement feature vectors respectively corresponding to the original feature vectors, the similarity in the feature domain of the original feature vector corresponding to the interaction information to be processed can be fully considered. In this way, in the process of determining the interaction result corresponding to the interaction information to be processed in this embodiment, not only the similarity in the feature domain of the original feature vectors corresponding to each respective interaction information to be processed is considered, but also the relevance between the enhanced feature vectors corresponding to each respective original feature vector is considered, so that the accuracy of the determined interaction result corresponding to the interaction information to be processed can be improved, the accuracy of CTR estimation in the recommendation scene can be effectively improved, and further the user experience can be improved, and the efficiency of the whole recommendation system can be improved.

For example, the embodiment of the present invention may be applied to an application scenario as shown in fig. 1. In this scenario, a terminal device 1 and a server 2 may be included.

The terminal device 1 may be hardware or software. When the terminal device 1 is hardware, it may be various electronic devices having a display screen and supporting communication with the server 2, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like; when the terminal device 1 is software, it may be installed in the electronic device as described above. The terminal device 1 may be implemented as a plurality of software or software modules, or as a single software or software module, to which the embodiments of the present disclosure are not limited. Further, various applications, such as a data processing application, an instant messaging tool, social platform software, a search class application, a shopping class application, and the like, may be installed on the terminal device 1.

The server 2 may be a server that provides various services, for example, a background server that receives a request transmitted from a terminal device with which communication connection is established, and the background server may perform processing such as receiving and analyzing the request transmitted from the terminal device and generate a processing result. The server 2 may be a server, a server cluster formed by a plurality of servers, or a cloud computing service center, which is not limited in the embodiment of the present disclosure.

The server 2 may be hardware or software. When the server 2 is hardware, it may be various electronic devices that provide various services to the terminal device 1. When the server 2 is software, it may be a plurality of software or software modules providing various services to the terminal device 1, or may be a single software or software module providing various services to the terminal device 1, which is not limited by the embodiments of the present disclosure.

The terminal device 1 and the server 2 may be communicatively connected via a network. The network may be a wired network using coaxial cable, twisted pair wire, and optical fiber connection, or may be a wireless network that can implement interconnection of various communication devices without wiring, for example, bluetooth (Bluetooth), near field communication (Near Field Communication, NFC), infrared (Infrared), etc., which are not limited by the embodiments of the present disclosure.

Specifically, the user may input the interaction information to be processed through the terminal device 1; the terminal device 1 sends the interaction information to be processed to the server 2. The server 2 can firstly determine a plurality of original feature vectors corresponding to the interaction information to be processed according to the interaction information to be processed; then, the server 2 may determine, for each original feature vector, an enhanced feature vector corresponding to the original feature vector; then, the server 2 can obtain an interaction result corresponding to the interaction information to be processed according to the enhancement feature vectors corresponding to the original feature vectors; finally, the server 2 may return the interaction result corresponding to the interaction information to be processed to the terminal device 1, so that the terminal device 1 may display the interaction result corresponding to the interaction information to be processed to the user. In this way, in this embodiment, after the original feature vector corresponding to each piece of interaction information to be processed is obtained, the enhancement feature vector corresponding to each original feature vector can be determined first, that is, in this embodiment, the information modeling in the feature domain can be performed on the original feature vector corresponding to each piece of interaction information to be processed, so as to obtain the enhancement feature vector corresponding to the original feature vector, so that the enhancement feature vector corresponding to the original feature vector can fully embody the accuracy and richness of the feature domain semantics of the original feature vector; therefore, in the process of determining the interaction result corresponding to the interaction information to be processed according to the enhancement feature vectors respectively corresponding to the original feature vectors, the similarity in the feature domain of the original feature vector corresponding to the interaction information to be processed can be fully considered. In this way, in the process of determining the interaction result corresponding to the interaction information to be processed in this embodiment, not only the similarity in the feature domain of the original feature vectors corresponding to each respective interaction information to be processed is considered, but also the relevance between the enhanced feature vectors corresponding to each respective original feature vector is considered, so that the accuracy of the determined interaction result corresponding to the interaction information to be processed can be improved, the accuracy of CTR estimation in the recommendation scene can be effectively improved, and further the user experience can be improved, and the efficiency of the whole recommendation system can be improved.

It should be noted that the specific types, numbers and combinations of the terminal device 1 and the server 2 and the network may be adjusted according to the actual requirements of the application scenario, which is not limited in the embodiment of the present disclosure.

It should be noted that the above application scenario is only shown for the convenience of understanding the present disclosure, and embodiments of the present disclosure are not limited in any way in this respect. Rather, embodiments of the present disclosure may be applied to any scenario where applicable.

Fig. 2 is a flowchart of a method for determining an interaction result according to an embodiment of the present disclosure. A method of determining the result of the interaction of fig. 2 may be performed by the terminal device or the server of fig. 1. As shown in fig. 2, the method for determining the interaction result includes:

s201: and obtaining the interaction information to be processed.

In this embodiment, the to-be-processed interaction information may be understood as interaction information that needs to be predicted as an interaction result, and the interaction information may be understood as information that can reflect an interaction intention. In one implementation, the interaction information to be processed may include interaction user information, interaction object information, and interaction behavior information.

In this embodiment, a user who performs an interactive action may be referred to as an interactive user. The interactive user information may be understood as characteristic information capable of reflecting the attribute of the interactive user itself performing the interactive behavior. An interactive user may be understood as an account or client that generates an interactive behavior with respect to a target object, for example, the interactive user information is a feature that can reflect a model of a mobile phone used by the interactive user (i.e., a model of a mobile phone to which an account is logged in), a location of the interactive user (e.g., province, city), and so on. It can be understood that the interactive user information of the interactive user is characteristic information of the attribute of the target user.

An interactive object may be understood as an object to which an interactive action is performed, and for example, a target object may be a short video, a commodity, a service, or the like. The interactive object information may be understood as characteristic information capable of reflecting the attribute of the interactive object itself. For example, when the interactive object is a commodity or a service, the interactive object information may be a feature capable of reflecting the attribute of the commodity or the service, such as the price, the sales volume, the product type, the price of the commodity recently browsed by the target user, and the sales place.

The interactive behavior information between the interactive user and the interactive object may be understood as feature information capable of reflecting an operation performed by the interactive user on the interactive object, for example, assuming that the interactive object is a commodity, the interactive behavior information may include features capable of reflecting the number of clicks of the commodity by the interactive user, whether the interactive user is collected by the user, and/or purchased. In the online e-commerce scenario, a user often browses a plurality of goods or services in the same e-commerce website or mobile terminal application program, and the actions may be operations such as stay on a certain goods page, clicking on goods viewing details, and the like, and these operations may be collectively referred to as interaction actions.

S202: and determining a plurality of original feature vectors corresponding to the interaction information to be processed according to the interaction information to be processed.

After the interaction information to be processed is obtained, the original feature vector corresponding to each interaction information to be processed can be determined first, so that subsequent processing can be performed based on the original feature vector corresponding to the interaction information to be processed. It should be noted that, determining the original feature vector corresponding to the interaction information to be processed may be understood as mapping the interaction information to be processed into a feature vector of a preset type, for example, a feature vector of a discrete feature type and a feature vector of a numerical feature type.

Specifically, after a plurality of pieces of interaction information to be processed are obtained, each piece of interaction information to be processed can be mapped into a feature vector of a preset type, and then the original feature vector corresponding to each piece of interaction information to be processed can be obtained. For example, assuming that the to-be-processed interactive information includes one interactive user information, one interactive object information, and one interactive behavior information, an original feature vector corresponding to the interactive user information, an original feature vector corresponding to the interactive object information, and an original feature vector corresponding to the interactive behavior information may be determined.

S203: for each original feature vector, an enhanced feature vector corresponding to the original feature vector is determined.

In this embodiment, the enhanced feature vector corresponding to the original feature vector may be understood as a feature vector capable of reflecting the relationship between feature information of each dimension in the original feature vector. It can be understood that the enhanced feature vector corresponding to the original feature vector can fully embody the accuracy and richness of the feature domain semantics of the original feature vector. It should be noted that, the feature field of the original feature vector may be understood as all feature information included in the original feature vector, that is, one feature field represents one original feature vector.

The enhanced feature vector corresponding to the original feature vector is determined, which can be understood as performing information modeling in a feature domain on the original feature vector corresponding to the interaction information to be processed to obtain the enhanced feature vector corresponding to the original feature vector. In this way, the enhancement feature vectors corresponding to the original feature vectors can fully embody the accuracy and richness of the feature domain semantics of the original feature vectors, so that the similarity in the feature domain of the original feature vectors corresponding to each piece of interaction information to be processed can be fully considered in the process of determining the interaction result corresponding to the interaction information to be processed according to the enhancement feature vectors corresponding to each piece of original feature vectors.

For example, the interactive user information includes city information, and the original feature vector corresponding to the city information is a 32-dimensional vector, then by modeling information in a feature domain of the original feature vector (i.e. mining information on the feature domain), the relationship between the 32-dimensional vectors expressing the city information is actually learned, so that the relationship between the 32-dimensional vectors expressing the city information can be determined as an enhanced feature vector corresponding to the original feature vector.

S204: and obtaining an interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector.

In this embodiment, the interaction result corresponding to the interaction information to be processed may be understood as a conversion result of the interaction user with respect to the interaction object, that is, an operation content executed by the interaction user with respect to the interaction object. It will be appreciated that the interaction result corresponding to the interaction information to be processed may be a result of user click prediction (Click Through Rate, CTR), for example, the interaction result corresponding to the interaction information to be processed may be clicked, not clicked, or purchased, not purchased, or the like.

After the enhancement feature vectors corresponding to the original feature vectors are determined, the interaction results corresponding to the interaction information to be processed are predicted by combining the relevance among the enhancement feature vectors corresponding to the original feature vectors. It can be understood that in this embodiment, after the enhanced feature vectors corresponding to the original feature vectors are obtained, the interaction result corresponding to the interaction information to be processed may be predicted according to the relevance between the enhanced feature vectors corresponding to the original feature vectors.

For example, assume that the interaction information to be processed includes interaction user information as: the interactive user is account A and the interactive object information is: the interactive object is commodity a and the interactive behavior information is: the account A browses the commodity a but does not purchase or collect; if the predicted conversion result of the interaction user for the interaction object is not purchased, the interaction result corresponding to the interaction information to be processed can be determined as: not purchased. It should be noted that, assuming that the target object is a commodity, if the target user makes a further action such as ordering or reserving on the target commodity in a period of time after the target user interacts with the last commodity, the further action of the target user may be referred to as conversion.

In some embodiments, the method corresponding to fig. 2 may be applied to a trained interaction result prediction model, where the interaction result prediction model may include a feature extraction module, and the feature extraction module may include a convolutional neural network, and the feature extraction module may perform the step S202 described above. Next, two specific implementations of S202 will be described.

In a first implementation manner, if the type of the interaction information to be processed is a discrete feature type, the step of determining, in S202, a plurality of original feature vectors corresponding to the interaction information to be processed according to the interaction information to be processed may include the following steps:

s202a: and mapping the interaction information to be processed into a serial number corresponding to the interaction information to be processed according to a preset dictionary.

It should be noted that discrete features refer to features that cannot be represented by a meaningful number, such as gender, nationality, city, english word, item ID and user ID in a recommendation system, and the like. These discrete features are directly used without a direct input to the model.

In this embodiment, a preset dictionary may be pre-established, where the preset dictionary may store a mapping relationship between a category and a sequence number, for example, when the category is "gender", the sequence number of a mapping corresponding to a category "man" is "0", the sequence number of a mapping corresponding to a category "woman" is "1", and for example, when the category is "city", the sequence number of a mapping corresponding to a category "Beijing" is "0", the sequence number of a mapping corresponding to a category "Shanghai" is "1", and so on.

When the type of the interaction information to be processed is a discrete feature type, the interaction information to be processed can be mapped into a serial number corresponding to the interaction information to be processed according to a preset dictionary. For example, if the to-be-processed interaction information includes that the city information of the interaction user is Beijing, and the serial number corresponding to "Beijing" in the preset dictionary is "0", the to-be-processed interaction information is mapped to the serial number corresponding to "0" of the to-be-processed interaction information.

S202b: and carrying out coding processing on the serial numbers corresponding to the interaction information to be processed to obtain sparse vectors.

After the sequence number corresponding to the interaction information to be processed is obtained, the sequence number corresponding to the interaction information to be processed can be subjected to coding processing, and a sparse vector is obtained. In One implementation manner, an One hot encoding manner may be adopted to encode the sequence number corresponding to the interaction information to be processed, and the sequence number is mapped to the sparse vector of Cheng Gaowei.

S202c: the sparse vector is mapped to a dense vector.

In this embodiment, after determining the sparse vector corresponding to the interaction information to be processed, mapping may be continued to the sparse vector to be a dense vector. In one implementation, sparse vectors may be mapped to low-dimensional dense vectors using an emmbedding mapping.

It should be noted that the mapping manner of the ebedding is actually that a parameter matrix is multiplied by a sparse vector. For example, if a feature has 20 categories, it is desirable to represent the feature with a 5-dimensional dense vector. Then a 5 x 20 parameter matrix is required which is used to project the feature 20-dimensional sparse vector code into 5-dimensional space. It should be emphasized that the dense vectors may well describe the relationship between feature classes, for example, in movies, the distance between dense vectors corresponding to movies of the same class may be very close, whereas the distance between dense vectors corresponding to movies of different classes may be very far. It can be seen that the dense vector is able to distinguish the degree of information similarity between the interaction information to be processed.

S202d: and taking the dense vector as an original feature vector corresponding to the interaction information to be processed.

After determining the dense vector corresponding to the interaction information to be processed, the dense vector corresponding to the interaction information to be processed can be used as the original feature vector corresponding to the interaction information to be processed.

In a second implementation manner, if the type of the interaction information to be processed is a numerical feature type, the step of determining, in S202, a plurality of original feature vectors corresponding to the interaction information to be processed according to the interaction information to be processed may include the following steps:

S202A: and carrying out normalization processing on the interaction information to be processed to obtain a normalized numerical value feature vector corresponding to the interaction information to be processed.

It should be noted that, numerical features refer to features that are inherently a meaningful number, and often can be directly used as input of models, such as age of people, size and weight of objects, and the like.

And when the type of the interaction information to be processed is a numerical feature type, normalizing the interaction information to be processed to obtain a normalized numerical feature vector corresponding to the interaction information to be processed. For example, regularization processing may be performed on the interaction information to be processed, and the numerical value corresponding to the interaction information to be processed may be mapped to a certain range, for example, standard score is calculated through standard normal distribution, maximum and minimum scaling normalization is performed between the numerical values, and a discrete feature is changed into a discrete feature by adopting a bucket discretization manner, and then subsequent feature transformation (for example, one-hot conversion) is performed. Therefore, after the interaction information to be processed is normalized, the numerical value corresponding to the interaction information to be processed is scaled to a uniform standard scale, so that different characteristics can be conveniently compared. For example, the interaction information to be processed is "price: 9999 elements ", the interaction information to be processed can be converted into numbers between 0 and 1 as normalized numerical feature vectors.

S202B: and taking the normalized numerical value feature vector corresponding to the interaction information to be processed as an original feature vector corresponding to the interaction information to be processed.

After determining the normalized numerical feature vector corresponding to the interaction information to be processed, the normalized numerical feature vector corresponding to the interaction information to be processed can be used as the original feature vector corresponding to the interaction information to be processed.

In some embodiments, the step of determining, for each original feature vector, an enhanced feature vector corresponding to the original feature vector in S203 may include the steps of:

for each original feature vector, determining a domain information vector corresponding to the original feature vector according to the original feature vector; and determining an enhanced feature vector corresponding to the original feature vector according to the original feature vector and the intra-domain information vector corresponding to the original feature vector.

In this embodiment, after the original feature vector corresponding to each piece of interaction information to be processed is obtained, the domain information vector corresponding to the original feature vector may be determined according to the original feature vector first for the original feature vector corresponding to each piece of interaction information to be processed, where the domain information vector corresponding to the original feature vector may be understood as a feature vector capable of reflecting the feature information of each dimension and the relationship between the feature information of each dimension in the original feature vector. Then, an enhanced feature vector corresponding to the original feature vector may be determined according to the original feature vector and the intra-domain information vector corresponding to the original feature vector.

In one implementation, the method is applied to a trained interaction result prediction model, wherein the interaction result prediction model comprises a feature domain modeling module. The feature domain modeling module may be a Deep Neural Network (DNN) or a Bert model, and the feature domain modeling module may be configured to perform the step corresponding to S203.

As an example, the feature domain modeling module may determine, first, a domain information vector corresponding to the original feature vector according to the original feature vector. Specifically, the feature domain modeling module may determine, according to feature information of each dimension in the original feature vector, a relationship between feature information of each dimension in the original feature vector; that is, the feature domain modeling module determines the feature information corresponding to each dimension in the original feature vector, and then the feature domain modeling module may extract the relationship between the feature information of each dimension in the original feature vector. Then, the feature domain modeling module may determine a domain information vector corresponding to the original feature vector according to a relationship between feature information of each dimension in the original feature vector, for example, the relationship between the original feature vector and feature information of each dimension in the original feature vector may be ordered according to a preset sequence ordering manner, so as to obtain the domain information vector corresponding to the original feature vector.

Then, the feature domain modeling module may determine an enhanced feature vector corresponding to the original feature vector according to the original feature vector and the intra-domain information vector corresponding to the original feature vector. Specifically, the feature domain modeling module may perform fusion processing on the original feature vector and the intra-domain information vector corresponding to the original feature vector to obtain an enhanced feature vector corresponding to the original feature vector, for example, may splice, add, multiply in alignment, or further perform fusion processing on the original feature vector and the intra-domain information vector corresponding to the original feature vectorAnd complicated operations such as a gating mechanism and the like can obtain the enhanced feature vector corresponding to the original feature vector. For example, the enhanced feature vector corresponding to the original feature vector may be determined by the following formula:

wherein e _i Is the original feature vector; e' _i The intra-domain information vector corresponding to the original feature vector; />

An enhanced feature vector corresponding to the original feature vector; the F () function is a function that operates on two vectors, and may be, for example, a concat (i.e., a concatenation) function, an element-add (i.e., an addition operation by element), a join function, a para-multiply, or a more complex operation such as a gating mechanism.

In some embodiments, the method may be applied to a trained interaction result prediction model, where the interaction result prediction model may include a result prediction module. In one implementation, the outcome prediction module may be an FM model, a DNN model, or a multi-headed attention mechanism model, which may be used to perform the step of S204. Specifically, the step of obtaining the interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector in S204 may include the following steps:

s204a: and the result prediction module performs cross fusion processing on the enhancement feature vectors corresponding to the original feature vectors to obtain fusion enhancement feature vectors.

After the result prediction module obtains the enhancement feature vectors corresponding to the original feature vectors, the result prediction module carries out cross fusion processing on the enhancement feature vectors corresponding to the original feature vectors to obtain fusion enhancement feature vectors. In one implementation, the manner of cross-fusion processing may include one of the following: linear weighting, self-attention (attention mechanism) processing, etc.

It should be noted that, the purpose of performing the cross fusion processing on the enhancement feature vectors corresponding to the original feature vectors is to mine the relationship between the enhancement feature vectors corresponding to the different original feature vectors. For example, a girl aged 20-30 may prefer cosmetics, and a girl aged 20-30 may prefer playing games, but these information are not presented with only one age or sex (i.e., the enhanced feature vector corresponding to the original feature vector alone), so that the result prediction module is required to perform cross-fusion processing on the enhanced feature vectors corresponding to the original feature vectors to obtain fused enhanced feature vectors.

S204b: and the result prediction module determines an interaction result corresponding to the interaction information to be processed according to the fusion enhancement feature vector.

After the fusion enhancement feature vector is obtained, the result prediction module can determine an interaction result corresponding to the interaction information to be processed according to the fusion enhancement feature vector. In one implementation, the result prediction module may output a plurality of predicted interaction results and probability values corresponding to each predicted interaction result according to the fusion enhanced feature vector; it should be noted that, the probability value corresponding to the predicted interactive result may reflect the probability that the predicted interactive result will actually occur, and it may be understood that the larger the probability value corresponding to the predicted interactive result, the larger the probability that the predicted interactive result will actually occur, otherwise, the smaller the probability value corresponding to the predicted interactive result, the smaller the probability that the predicted interactive result will actually occur. Then, the predicted interaction result with the largest probability value can be used as the interaction result corresponding to the interaction information to be processed.

For example, the output of a plurality of predicted interaction results and the probability value corresponding to each predicted interaction result by the result prediction module according to the fusion enhanced feature vector is specifically: the predicted interactive result a, the user purchases goods, and the corresponding probability value thereof are 0.9, and the predicted interactive result b, the user does not purchase goods, and the corresponding probability value thereof are 0.1. Because the probability value corresponding to the predicted interactive result a 'user purchases goods' is the largest, the predicted interactive result a 'user purchases goods' can be used for determining the interactive result corresponding to the to-be-processed interactive information.

Any combination of the above-mentioned optional solutions may be adopted to form an optional embodiment of the present disclosure, which is not described herein in detail.

The following are device embodiments of the present disclosure that may be used to perform method embodiments of the present disclosure. For details not disclosed in the embodiments of the apparatus of the present disclosure, please refer to the embodiments of the method of the present disclosure.

Fig. 3 is a schematic diagram of an apparatus for determining an interaction result provided by an embodiment of the disclosure. As shown in fig. 3, the device for determining the interaction result includes:

an information acquisition unit 301, configured to acquire interaction information to be processed;

a first determining unit 302, configured to determine, according to the interaction information to be processed, a plurality of original feature vectors corresponding to the interaction information to be processed;

a second determining unit 303, configured to determine, for each original feature vector, an enhanced feature vector corresponding to the original feature vector;

and the result determining unit 304 is configured to obtain an interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector.

Optionally, the to-be-processed interaction information includes interaction user information, interaction object information and interaction behavior information.

Optionally, if the type of the interaction information to be processed is a discrete feature type, the first determining unit 302 is specifically configured to:

Mapping the interaction information to be processed into a serial number corresponding to the interaction information to be processed according to a preset dictionary;

coding the serial number corresponding to the interaction information to be processed to obtain a sparse vector;

mapping the sparse vector to a dense vector;

and taking the dense vector as an original feature vector corresponding to the interaction information to be processed.

Optionally, if the type of the interaction information to be processed is a numerical feature type, the first determining unit 302 is specifically configured to:

normalizing the interaction information to be processed to obtain a normalized numerical feature vector corresponding to the interaction information to be processed;

and taking the normalized numerical value feature vector corresponding to the interaction information to be processed as an original feature vector corresponding to the interaction information to be processed.

Optionally, the second determining unit 303 is specifically configured to:

Optionally, the method is applied to a trained interaction result pre-estimation model, wherein the interaction result pre-estimation model comprises a feature domain modeling module;

the second determining unit 303 is specifically configured to:

the feature domain modeling module determines the relation among the feature information of each dimension in the original feature vector according to the feature information of each dimension in the original feature vector;

and the feature domain modeling module determines a domain information vector corresponding to the original feature vector according to the relation between the feature information of each dimension in the original feature vector.

Optionally, the second determining unit 303 is specifically configured to:

and the feature domain modeling module performs fusion processing on the original feature vector and the intra-domain information vector corresponding to the original feature vector to obtain an enhanced feature vector corresponding to the original feature vector.

Optionally, the method is applied to a trained interaction result prediction model, wherein the interaction result prediction model comprises a result prediction module;

the result determining unit 304 is configured to:

the result prediction module carries out cross fusion processing on the enhancement feature vectors corresponding to the original feature vectors to obtain fusion enhancement feature vectors;

And the result prediction module determines an interaction result corresponding to the interaction information to be processed according to the fusion enhancement feature vector.

Compared with the prior art, the embodiment of the disclosure has the beneficial effects that: the embodiment of the disclosure provides a device for determining an interaction result, which comprises: the information acquisition unit is used for acquiring the interaction information to be processed; the first determining unit is used for determining a plurality of original feature vectors corresponding to the interaction information to be processed according to the interaction information to be processed; a second determining unit, configured to determine, for each original feature vector, an enhanced feature vector corresponding to the original feature vector; and the result determining unit is used for obtaining the interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector. It can be seen that, in this embodiment, after the original feature vector corresponding to each piece of interaction information to be processed is obtained, the enhancement feature vector corresponding to each original feature vector can be determined first, that is, in this embodiment, the information modeling in the feature domain can be performed on the original feature vector corresponding to each piece of interaction information to be processed, so as to obtain the enhancement feature vector corresponding to the original feature vector, so that the enhancement feature vector corresponding to the original feature vector can fully embody the accuracy and richness of the feature domain semantics of the original feature vector; therefore, in the process of determining the interaction result corresponding to the interaction information to be processed according to the enhancement feature vectors respectively corresponding to the original feature vectors, the similarity in the feature domain of the original feature vector corresponding to the interaction information to be processed can be fully considered. In this way, in the process of determining the interaction result corresponding to the interaction information to be processed in this embodiment, not only the similarity in the feature domain of the original feature vectors corresponding to each respective interaction information to be processed is considered, but also the relevance between the enhanced feature vectors corresponding to each respective original feature vector is considered, so that the accuracy of the determined interaction result corresponding to the interaction information to be processed can be improved, the accuracy of CTR estimation in the recommendation scene can be effectively improved, and further the user experience can be improved, and the efficiency of the whole recommendation system can be improved.

It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic of each process, and should not constitute any limitation on the implementation process of the embodiments of the disclosure.

Fig. 4 is a schematic diagram of a computer device 4 provided by an embodiment of the present disclosure. As shown in fig. 4, the computer device 4 of this embodiment includes: a processor 401, a memory 402 and a computer program 403 stored in the memory 402 and executable on the processor 401. The steps of the various method embodiments described above are implemented by processor 401 when executing computer program 403. Alternatively, the processor 401 may implement the functions of the modules/modules in the above-described device embodiments when executing the computer program 403.

Illustratively, the computer program 403 may be partitioned into one or more modules/modules, which are stored in the memory 402 and executed by the processor 401 to complete the present disclosure. One or more of the modules/modules may be a series of computer program instruction segments capable of performing particular functions to describe the execution of the computer program 403 in the computer device 4.

The computer device 4 may be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The computer device 4 may include, but is not limited to, a processor 401 and a memory 402. It will be appreciated by those skilled in the art that fig. 4 is merely an example of computer device 4 and is not intended to limit computer device 4, and may include more or fewer components than shown, or may combine certain components, or different components, e.g., a computer device may also include an input-output device, a network access device, a bus, etc.

The processor 401 may be a central processing module (Central Processing Unit, CPU) or other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 402 may be an internal storage module of the computer device 4, for example, a hard disk or a memory of the computer device 4. The memory 402 may also be an external storage device of the computer device 4, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card) or the like, which are provided on the computer device 4. Further, the memory 402 may also include both internal memory modules of the computer device 4 and external memory devices. The memory 402 is used to store computer programs and other programs and data required by the computer device. The memory 402 may also be used to temporarily store data that has been output or is to be output.

It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of each functional module and module is illustrated, and in practical application, the above-described functional allocation may be performed by different functional modules and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional modules or modules to perform all or part of the above-described functions. The functional modules and the modules in the embodiment can be integrated in one processing module, or each module can exist alone physically, or two or more modules can be integrated in one module, and the integrated modules can be realized in a form of hardware or a form of a software functional module. In addition, the specific names of the functional modules and the modules are only for distinguishing from each other, and are not used for limiting the protection scope of the present disclosure. The modules in the above system, and the specific working process of the modules may refer to the corresponding process in the foregoing method embodiment, which is not described herein again.

In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.

Those of ordinary skill in the art will appreciate that the various illustrative modules and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

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

The modules illustrated as separate components may or may not be physically separate, and components shown as modules may or may not be physical modules, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional module in each embodiment of the present disclosure may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules.

The integrated modules/modules may be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product. Based on such understanding, the present disclosure may implement all or part of the flow of the method of the above-described embodiments, or may be implemented by a computer program to instruct related hardware, and the computer program may be stored in a computer readable storage medium, where the computer program, when executed by a processor, may implement the steps of the method embodiments described above. The computer program may comprise computer program code, which may be in source code form, object code form, executable file or in some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the content of the computer readable medium can be appropriately increased or decreased according to the requirements of the jurisdiction's jurisdiction and the patent practice, for example, in some jurisdictions, the computer readable medium does not include electrical carrier signals and telecommunication signals according to the jurisdiction and the patent practice.

The above embodiments are merely for illustrating the technical solution of the present disclosure, and are not limiting thereof; although the present disclosure has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the disclosure, and are intended to be included in the scope of the present disclosure.

Claims

1. A method for determining an interaction result, the method comprising:

acquiring interaction information to be processed;

2. The method of claim 1, wherein the interaction information to be processed includes interaction user information, interaction object information, and interaction behavior information.

3. The method according to claim 1 or 2, wherein if the type of the interaction information to be processed is a discrete feature type, the determining, according to the interaction information to be processed, a plurality of original feature vectors corresponding to the interaction information to be processed includes:

mapping the sparse vector to a dense vector;

4. The method according to claim 1 or 2, wherein if the type of the interaction information to be processed is a numerical feature type, the determining, according to the interaction information to be processed, a plurality of original feature vectors corresponding to the interaction information to be processed includes:

5. The method according to claim 1 or 2, wherein for each original feature vector, determining an enhanced feature vector corresponding to the original feature vector comprises:

6. The method of claim 5, wherein the method is applied to a trained interaction result prediction model, wherein the interaction result prediction model comprises a feature domain modeling module;

the determining, according to the original feature vector, the intra-domain information vector corresponding to the original feature vector includes:

7. The method of claim 6, wherein the determining the enhanced feature vector corresponding to the original feature vector from the original feature vector and the intra-domain information vector corresponding to the original feature vector comprises:

8. The method according to claim 1 or 2, wherein the method is applied to a trained interaction result prediction model, wherein the interaction result prediction model comprises a result prediction module;

the step of obtaining the interaction result corresponding to the interaction information to be processed according to the enhanced feature vector corresponding to each original feature vector comprises the following steps:

9. An apparatus for determining an interaction result, the apparatus comprising:

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1 to 8 when the computer program is executed.

11. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method according to any one of claims 1 to 8.