Information sorting method and device, electronic device and storage medium

Abstract

The present disclosure provides an information sorting method and device, an electronic device and a storage medium, which relate to the field of data processing technology, and in particular to intelligent search and information recommendation systems. The implementation scheme is: obtaining a list to be sorted; obtaining multiple first representation vectors corresponding to multiple information respectively; in response to the absence of a historical information list obtained in response to a previous information request, adding a randomly generated placeholder vector to the multiple first representation vectors; performing attention encoding processing on the multiple first representation vectors and the placeholder vectors to obtain multiple first fused representation vectors corresponding to the multiple first representation vectors and a second fused representation vector corresponding to the placeholder vector; and obtaining multiple first scores based on the multiple first fused representation vectors to determine the second order of the multiple information.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to an intelligent search and information recommendation system, and specifically to an information sorting method, device, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] Artificial intelligence is a discipline that studies how to use computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0003] Information sorting is one of the most important modules of the information recommendation system. High-quality information sorting has a very significant promoting effect on user activity and retention.

[0004] The methods described in this section are not necessarily methods that have been previously conceived or employed. Unless otherwise indicated, it should not be assumed that any method described in this section is considered to be prior art simply because it is included in this section. Similarly, unless otherwise indicated, the issues mentioned in this section should not be considered to have been recognized in any prior art. Summary of the invention

[0005] The present disclosure provides an information ranking method, device, electronic device, computer-readable storage medium, and computer program product for an information recommendation system.

[0006] According to one aspect of the present disclosure, a method for sorting information for an information recommendation system is provided, the method comprising: obtaining a list to be sorted, wherein the list to be sorted includes multiple information obtained in response to a current information request, and the multiple information is arranged in a first order; obtaining multiple first representation vectors corresponding to the multiple information respectively; in response to the absence of a historical information list obtained in response to a previous information request, adding a randomly generated placeholder vector to the multiple first representation vectors, wherein the dimension of the placeholder vector is the same as the dimension of each first representation vector in the multiple first representation vectors; performing attention encoding processing on the multiple first representation vectors and the placeholder vector to obtain multiple first fused representation vectors corresponding to the multiple first representation vectors respectively and a second fused representation vector corresponding to the placeholder vector, wherein the second fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to the information sorting of the information list obtained in response to a subsequent information request; and obtaining multiple first scores based on the multiple first fused representation vectors to determine the second order of the multiple information.

[0007] According to another aspect of the present disclosure, a training method for an information ranking model for an information recommendation system is provided, the method comprising: obtaining a sample data set, wherein each sample data in the sample data set includes an information list of a sample user and multiple labels corresponding to multiple information in the information list, wherein each of the multiple labels is used to indicate whether the corresponding information is clicked by the sample user; initializing multiple parameters of the information ranking model; and for each sample data, performing the following operations: obtaining multiple first representation vectors corresponding to multiple information respectively; performing attention encoding processing on the multiple first representation vectors to obtain multiple first fused representation vectors corresponding to the multiple first representation vectors respectively; obtaining multiple first scores based on the multiple first fused representation vectors; calculating a loss function based on the multiple first scores and the multiple labels corresponding to the sample data; and adjusting multiple parameters of the information ranking model based on the loss function.

[0008] According to another aspect of the present disclosure, there is provided an information sorting device for an information recommendation system, the device comprising: a first acquisition unit, configured to acquire a list to be sorted, wherein the list to be sorted comprises a plurality of information acquired in response to a current information request, and the plurality of information are arranged in a first order; a second acquisition unit, configured to acquire a plurality of first representation vectors corresponding to the plurality of information respectively; an adding unit, configured to add a randomly generated placeholder vector to the plurality of first representation vectors in response to the absence of a historical information list acquired in response to a previous information request, wherein the dimension of the placeholder vector is the same as the dimension of each first representation vector in the plurality of first representation vectors; a first encoding unit, configured to perform attention encoding processing on the plurality of first representation vectors and the placeholder vector to acquire a plurality of first fused representation vectors corresponding to the plurality of first representation vectors respectively and a second fused representation vector corresponding to the placeholder vector, wherein the second fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to the information sorting of the information list acquired in response to a subsequent information request; and a third acquisition unit, configured to acquire a plurality of first scores based on the plurality of first fused representation vectors to determine a second order of the plurality of information.

[0009] According to another aspect of the present disclosure, a training device for an information ranking model for an information recommendation system is provided, the device comprising: a sixth acquisition unit, configured to acquire a sample data set, wherein each sample data in the sample data set comprises an information list of a sample user and a plurality of labels corresponding to a plurality of information in the information list, wherein each of the plurality of labels is used to indicate whether the corresponding information is clicked by the sample user; an initialization unit, configured to initialize a plurality of parameters of the information ranking model; and an execution unit, configured to perform operations of the following subunits for each sample data, wherein the execution unit comprises: a first acquisition subunit, configured to acquire a plurality of first representation vectors corresponding to a plurality of information respectively; an encoding subunit, configured to perform attention encoding processing on the plurality of first representation vectors to acquire a plurality of first fused representation vectors corresponding to the plurality of first representation vectors respectively; a second acquisition subunit, configured to acquire a plurality of first scores based on the plurality of first fused representation vectors; a calculation subunit, configured to calculate a loss function based on the plurality of first scores and the plurality of labels corresponding to the sample data; and an adjustment subunit, configured to adjust a plurality of parameters of the information ranking model based on the loss function.

[0010] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above-mentioned information sorting method or the above-mentioned information sorting model training method.

[0011] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the above-mentioned information sorting method or the above-mentioned information sorting model training method.

[0012] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements the above-mentioned information sorting method or the above-mentioned information sorting model training method when executed by a processor.

[0013] According to one or more embodiments of the present disclosure, a fused representation vector for characterizing the global feature information of the current list can be obtained, and the sorting effect can be improved by applying the vector to the information sorting of the subsequent refresh list.

[0014] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0016] Figure 1 A schematic diagram showing an exemplary system in which the various methods described herein may be implemented according to an embodiment of the present disclosure;

[0017] Figure 2 A flowchart of an information ranking method for an information recommendation system according to an embodiment of the present disclosure is shown;

[0018] Figure 3 A schematic diagram of the structure of an information recommendation system according to an embodiment of the present disclosure is shown;

[0019] Figure 4 A flowchart of an information ranking method for an information recommendation system according to an embodiment of the present disclosure is shown;

[0020] Figure 5 A schematic diagram of the structure of an information sorting model according to an embodiment of the present disclosure is shown;

[0021] Figure 6 A schematic structural diagram of a Fastformer encoder according to an embodiment of the present disclosure is shown;

[0022] Figure 7A flowchart of a method for training an information ranking model for an information recommendation system according to an embodiment of the present disclosure is shown;

[0023] Figure 8 A structural block diagram of an information ranking device for an information recommendation system according to an embodiment of the present disclosure is shown;

[0024] Fig. 9 A structural block diagram of an information ranking device for an information recommendation system according to an embodiment of the present disclosure is shown;

[0025] Fig.10 A structural block diagram of a training device for an information ranking model of an information recommendation system according to an embodiment of the present disclosure is shown;

[0026] Fig.11 A structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0027] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0028] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, temporal relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0029] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0030] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0031] Figure 1 FIG. 1 is a schematic diagram of an exemplary system 100 in which various methods and apparatuses described herein may be implemented according to an embodiment of the present disclosure. Figure 1, the system 100 includes one or more client devices 101, 102, 103, 104, 105, and 106, a server 120, and one or more communication networks 110 coupling the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105, and 106 may be configured to execute one or more applications.

[0032] In an embodiment of the present disclosure, the server 120 may run one or more services or software applications that enable the method of information sorting to be performed.

[0033] In some embodiments, server 120 may also provide other services or software applications that may include non-virtualized environments and virtualized environments. In some embodiments, these services may be provided as web-based services or cloud services, such as provided to users of client devices 101, 102, 103, 104, 105, and / or 106 under a software as a service (SaaS) model.

[0034] exist Figure 1 In the configuration shown, the server 120 may include one or more components that implement the functions performed by the server 120. These components may include software components, hardware components, or a combination thereof that can be executed by one or more processors. Users operating client devices 101, 102, 103, 104, 105, and / or 106 may in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by these components. It should be understood that a variety of different system configurations are possible, which may differ from the system 100. Therefore, Figure 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

[0035] The user may use client devices 101, 102, 103, 104, 105 and / or 106 to send information requests. The client devices may provide an interface that enables the user of the client device to interact with the client device. The client device may also output information to the user via the interface. Figure 1 Only six client devices are depicted, but one skilled in the art will appreciate that the present disclosure may support any number of client devices.

[0036] Client devices 101, 102, 103, 104, 105 and / or 106 may include various types of computer devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, game systems, thin clients, various messaging devices, sensors or other sensing devices, etc. These computer devices may run various types and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as GOOGLE Chrome OS); or include various mobile operating systems, such as MICROSOFT Windows Mobile OS, iOS, Windows Phone, Android. Portable handheld devices may include cellular phones, smart phones, tablet computers, personal digital assistants (PDAs), etc. Wearable devices may include head-mounted displays (such as smart glasses) and other devices. Game systems may include various handheld game devices, Internet-enabled game devices, etc. Client devices are capable of executing various different applications, such as various Internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and may use various communication protocols.

[0037] The network 110 may be any type of network known to those skilled in the art that may support data communications using any of a variety of available protocols, including but not limited to TCP / IP, SNA, IPX, etc. By way of example only, the one or more networks 110 may be a local area network (LAN), an Ethernet-based network, a token ring, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a public switched telephone network (PSTN), an infrared network, a wireless network (e.g., Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0038] Server 120 may include one or more general purpose computers, dedicated server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and / or combination. Server 120 may include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that may be virtualized to maintain a server's virtual storage device). In various embodiments, server 120 may run one or more services or software applications that provide the functionality described below.

[0039] The computing units in the server 120 may run one or more operating systems including any of the above operating systems and any commercially available server operating systems. The server 120 may also run any of a variety of additional server applications and / or middle-tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0040] In some implementations, server 120 may include one or more applications to analyze and consolidate data feeds and / or event updates received from users of client devices 101, 102, 103, 104, 105, and / or 106. Server 120 may also include one or more applications to display data feeds and / or real-time events via one or more display devices of client devices 101, 102, 103, 104, 105, and / or 106.

[0041] In some embodiments, the server 120 may be a server of a distributed system, or a server combined with a blockchain. The server 120 may also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. A cloud server is a host product in a cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services.

[0042] The system 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as audio files and video files. The databases 130 may reside in various locations. For example, the database used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection. The databases 130 may be of different types. In some embodiments, the databases used by the server 120 may be, for example, relational databases. One or more of these databases may store, update, and retrieve data to and from the databases in response to commands.

[0043] In some embodiments, one or more of the databases 130 may also be used by applications to store application data. The databases used by the applications may be different types of databases, such as a key-value store, an object store, or a conventional store backed by a file system.

[0044] Figure 1 The system 100 may be configured and operated in various ways to enable the application of various methods and apparatuses described in the present disclosure.

[0045] At present, in the information sorting system, the methods for sorting information lists mainly include generative methods and list optimization methods. These two methods can sort the information in the list based on the global information of the information list obtained by the current refresh, but neither can model the information list obtained by the user's historical refresh, resulting in the inability to fully utilize the information and the difficulty in further optimizing the sorting effect.

[0046] According to the embodiments of the present disclosure, Figure 2 As shown, a method for sorting information for an information recommendation system is provided, comprising: step S201, obtaining a list to be sorted, wherein the list to be sorted includes multiple information obtained in response to a current information request, and the multiple information is arranged in a first order; step S202, obtaining multiple first representation vectors corresponding to the multiple information respectively; step S203, in response to the absence of a historical information list obtained in response to a previous information request, adding a randomly generated placeholder vector to the multiple first representation vectors, wherein the dimension of the placeholder vector is the same as the dimension of each first representation vector in the multiple first representation vectors; step S204, performing attention encoding processing on the multiple first representation vectors and the placeholder vector to obtain multiple first fused representation vectors corresponding to the multiple first representation vectors respectively and a second fused representation vector corresponding to the placeholder vector, wherein the second fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to the information sorting of the information list obtained in response to the subsequent information request; and step S205, based on the multiple first fused representation vectors, obtaining multiple first scores to determine the second order of the multiple information.

[0047] According to the embodiments of the present disclosure, a fused representation vector for characterizing the global feature information of the current list can be obtained, and then by applying the vector to the information sorting of the subsequent refresh list, full use of the information in the historical information list can be achieved, thereby optimizing the sorting effect.

[0048] Figure 3 A structural diagram of the information recommendation system is given. The current information recommendation mainly includes four information processing stages: information recall stage, information rough sorting stage, information fine sorting stage and information re-sorting stage, which correspond to the information recall module 310, information rough sorting module 320, information fine sorting module 330 and information re-sorting module 340 of the information recommendation system 300, respectively. The information sorting method given in the embodiment of the present disclosure can be applied to the information rough sorting stage, information fine sorting stage and information re-sorting stage, and is not limited here.

[0049] In some embodiments, the obtained list to be sorted can be obtained through the information precise sorting model in the information recommendation system. Among them, the information precise sorting model is the main component of the information precise sorting module 330, which can perform prediction calculations based on user characteristics, information characteristics, etc., so as to obtain a more accurate information sorting result. The list to be sorted output by the information precise sorting model, in addition to including multiple information arranged in the first order after precise sorting, also includes the user's behavioral feature information and the precise sorting hidden layer representation vector corresponding to each information output by the hidden layer of the information precise sorting model, wherein the precise sorting hidden layer representation vector includes the interaction information between each information and the user characteristics.

[0050] In some embodiments, each of the plurality of first representation vectors is obtained by concatenating the information representation vector of the corresponding information and at least one of the following representation vectors: a user feature representation vector, a position representation vector of the corresponding information, and a representation vector of the corresponding information obtained by the information refined ranking model and fused with user information. Thus, by introducing user behavior features, resource location information, and resource representation fused with user information obtained from the refined ranking model, the information used for list reordering is further enriched, and the effect of list reordering is optimized.

[0051] The information representation vector of the corresponding information can be obtained by inputting each information in the information list into the pre-trained language model.

[0052] The user feature representation vector can be obtained by inputting the user behavior feature data into the unsupervised semantic representation model. User behavior feature data is generally high-dimensional sparse data, which includes the user's search behavior, web browsing behavior, and user click behavior. By transforming it through the unsupervised semantic representation model, it is possible to obtain user behavior feature information while reducing the dimension of the data and the amount of subsequent calculations.

[0053] The position representation vector of the information can be obtained by applying a method similar to the position embedding in BERT or the position encoding of Transformer. For example, the position representation vector PE can be obtained by applying the following formula:

[0054] PE(t,2i)=sin(t / 10000 2i / d )

[0055] PE(t,2i+1)=cos(t / 10000 2i / d )

[0056] Among them, t is the arrangement position of the corresponding information in the information list, 2i and 2i+1 respectively represent the even bits and odd bits in the position representation vector PE, and d is the dimension of the position representation vector.

[0057] The representation vector of the fused user information obtained by the information precise ranking model is also the refined hidden layer representation vector obtained by the information precise ranking model, which contains the interaction information between each information and user features.

[0058] By correspondingly concatenating the information representation vector, position representation vector, sorted hidden layer representation vector and user feature representation vector of the above corresponding information, the first representation vector of the corresponding information can be obtained. Thus, by obtaining multiple first representation vectors corresponding to multiple information in the information list, the feature information contained in the information list can be modeled more completely.

[0059] According to some embodiments, Figure 4 As shown, a method for sorting information for an information recommendation system is also provided, which also includes: step S406, in response to the existence of a historical information list, obtaining a third fused representation vector corresponding to the historical information list, wherein the third fused representation vector is used to represent the global feature information of the historical information list; step S407, performing attention encoding processing on multiple first representation vectors and the third fused representation vector to obtain multiple fourth fused representation vectors corresponding to the multiple first representation vectors respectively; and step S408, based on the multiple fourth fused representation vectors, obtaining multiple second scores to determine the third order of the multiple information. The operations of steps S401-S405 are similar to the operations of the above steps S201-S205, and are not described in detail here.

[0060] Therefore, by representing the historically refreshed resource list as a third fusion information representation vector and introducing the historical list information into the current sorting process through the vector, the historically refreshed information can be fully utilized in the sorting of the currently refreshed list, thereby optimizing the sorting effect.

[0061] According to some embodiments, performing attention encoding processing on multiple first representation vectors and a third fused representation vector also includes: based on the attention encoding processing, obtaining a fifth fused representation vector corresponding to the third fused representation vector, wherein the fifth fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to the information sorting of the information list obtained in response to a subsequent information request.

[0062] In order to introduce the information in the historically updated information list into the sorting process of the current information list, it is first necessary to obtain the characteristic information representation vector of the historically updated information list.

[0063] In some embodiments, in response to the absence of a historical information list, a placeholder vector may be added to the multiple first representation vectors, and the placeholder vector may be a randomly generated vector with the same dimension as the first representation vector. By simultaneously inputting the placeholder vector and the multiple first representation vectors into the attention encoding model, the output second fused representation vector corresponding to the placeholder vector may represent the global feature information of the current information list.

[0064] In some embodiments, in response to the existence of a historical information list, a third fused representation vector corresponding to the historical information list can be added to multiple first representation vectors and simultaneously input into the attention encoding model. The output fifth fused representation vector corresponding to the third fused representation vector can represent the global feature information of the current information list.

[0065] Through attention encoding processing, the global feature information of the current information list is fused into a fused representation vector, that is, the second fused representation vector or the fifth fused representation vector obtained by the above method. When the current information list is used as a historical information list in the information sorting process of the subsequent refreshed information list, its corresponding second fused representation vector or the fifth fused representation vector is the third fused representation vector used in the information sorting of the subsequent list. Therefore, by fusion-representing the global feature information of the historical information list into a vector, the efficiency of sorting the current information list can be guaranteed while fully acquiring the feature information of the historical information list.

[0066] According to some embodiments, obtaining a third fused representation vector corresponding to a historical information list includes: in response to multiple first information lists obtained respectively in response to multiple preceding information requests, obtaining multiple sixth fused representation vectors corresponding to the multiple first information lists respectively, wherein each of the multiple sixth fused representation vectors is obtained by performing attention encoding processing on the corresponding first information list to represent the global feature information of the corresponding first information list; and performing weighted summation on the multiple sixth fused representation vectors to obtain a third fused representation vector.

[0067] When there are multiple historical information lists, corresponding weights can be set for the multiple sixth fusion representation vectors corresponding to the multiple historical information lists, and then a third fusion representation vector integrating the global feature information of the multiple historical information lists is obtained by weighted summation. The sixth fusion representation vector is the second fusion representation vector or the fifth fusion representation vector obtained during the information sorting process of the corresponding historical information list and used to represent the global feature information of the information list.

[0068] In one example, it can be considered that the closer the historical information list is to the current update, the more important its global feature information is, so a higher weight should be set for it. Then the final fusion representation vector can be obtained by the following formula:

[0069]

[0070] Among them, g t represents the fusion representation vector of the previous t historical updates; h i is the fusion representation vector of the historical information list corresponding to the i-th historical refresh, where i∈[1, t]; a is the weight coefficient, a∈(0, 1). It can be understood that the specific value of the weight coefficient a can be determined according to the actual situation and is not limited here.

[0071] Therefore, by performing weighted summation on the fusion representation vectors corresponding to each historical information list, the influence of each historical refresh on the current refresh can be taken into account while obtaining the fusion information of the historical information list, thereby further enriching the fusion information of the historical refresh and optimizing the sorting effect.

[0072] like Figure 5 The figure shows a schematic diagram of the information sorting model of the present disclosure. Taking information re-sorting as an example, when the list to be sorted does not have a historical information list, the e0 shown in the figure is a placeholder vector; when the list to be sorted has a historical information list, the e0 shown in the figure is a fusion representation vector of the historical information list. [e1, e2, ..., e N-1 , e N ] are the multiple first representation vectors, where N is the number of information included in the information list. N-1 , e N ] are simultaneously input into the attention encoder, and multiple fusion representation vectors [y0, y1, y2, ..., y N-1 ,y N ], where the fused representation vector y0 is the fused representation vector used to represent the global feature information of the current list.

[0073] The fusion represents the vector [y1, y2, ..., y N-1 ,y N ] represent the fusion representation vector of each information in the list after integrating the list global information and the history list information. To obtain the first score of each information, it is necessary to convert [y1, y2, ..., y N-1 ,y N ] are input into the feed-forward network FFN, and each vector is converted into a one-dimensional intermediate value x i , i∈[1,N]. Among them, the feed-forward network FFN can be a fully connected layer.

[0074] Then, the value x can be i They are input into the Softmax layer for normalization, and finally the final score score(i), i∈[1, N] of each information is obtained. The normalization process in the Softmax layer can be expressed by the following formula:

[0075]

[0076] Among them, x i and x j are the intermediate values ​​of the i-th and j-th information, respectively, where i∈[1,N], j∈[1,N], and w i and w j x i and x j The weight coefficient of .

[0077] The final score score(i) is the score corresponding to the i-th information in the list to be sorted. Based on this score, multiple information in the list to be sorted can be re-sorted.

[0078] According to some embodiments, the attention encoding process is performed by a Fastformer encoder.

[0079] The Fastformer encoder is an encoder based on the additive attention mechanism. Its structural diagram is as follows: Figure 6 As shown. Its input list is [e0, e1, e2, ..., e N-1 , e N ], where [e1, e2, ..., e N-1 , e N ] are the multiple first representation vectors, where N is the number of information included in the information list; when the list to be sorted does not have a historical information list, e0 shown in the figure is a placeholder vector; when the list to be sorted has a historical information list, e0 shown in the figure is the fusion representation vector of the historical information list.

[0080] First, the input list is input into the query vector transformation layer, key vector transformation layer and value vector transformation layer for simple linear transformation, and the corresponding query vectors q are obtained respectively. i , key vector k i and the value vector v i , where i∈[0,N]. The corresponding linear transformations can be expressed as follows:

[0081]

[0082]

[0083]

[0084] Among them, W Q , W K , W V Represent the weight coefficients of the three linear transformations respectively.

[0085] To obtain the global query vector q through the additive attention mechanism, we first need to obtain the i-th query vector q i The attention weight α i , α i It can be obtained by the following formula:

[0086]

[0087] Among them, w q represents a learnable parameter vector, d represents the dimension of the query vector, and q i and q j They represent the i-th and j-th query vectors respectively, where i∈[0,N], j∈[0,N].

[0088] The global query vector q can be obtained by the following formula:

[0089]

[0090] In order to efficiently interact with the key vectors while maintaining the differences between different key vectors, the encoder uses a point-by-point inner product approach to calculate the global query vector q and the key vector k. i Interact, and the resulting interaction vector p i It can be expressed by the following formula:

[0091] p i =q*k i

[0092] Then, to obtain the global key vector k based on the additive attention mechanism, we first need to obtain the i-th interaction vector p i The attention weight β i , β i It can be obtained by the following formula:

[0093]

[0094] Among them, w k represents a learnable parameter vector, d represents the dimension of the interaction vector, and p i and p j They represent the i-th and j-th interaction vectors respectively, where i∈[0,N], j∈[0,N].

[0095] The global key vector k can be obtained by the following formula:

[0096]

[0097] Similarly, the value vector v can be obtained by pointwise inner product i Interact with the global key vector k, and the resulting interaction vector u i It can be expressed by the following formula:

[0098] u i =k*v i

[0099] To better extract the interaction vector u i The hidden information can be used to i Further linear transformation is performed to obtain the output vector r i , the corresponding linear transformation can be expressed as follows:

[0100]

[0101] Among them, w u is the weight coefficient of the linear transformation.

[0102] Finally, by outputting the vector r i With the query vector q i Add them separately to get the final output y of the Fastformer encoder i :

[0103] y i =r i +q i

[0104] Compared with the conventional Transformer encoder, the Fastformer encoder can reduce the computational complexity from the original O(N d-dimensional representation vectors) to O(N d-dimensional representation vectors) by applying the additive attention mechanism. 2 ×d) is reduced to a value close to the linear complexity of O(N×d), thereby reducing the complexity of encoding and improving the encoding efficiency while ensuring the effective capture of global context information, thereby improving the efficiency of information sorting.

[0105] According to some embodiments, the attention encoding process is performed by cascading multiple Fastformer encoders, and the multiple Fastformer encoders share parameters. Thus, by applying multiple Fastformer encoders, the global information of the information list can be further extracted; at the same time, through parameter sharing, the overfitting problem can be prevented while reducing the parameter scale and improving the modeling efficiency.

[0106] According to some embodiments, Figure 7 As shown, a training method for an information ranking model for an information recommendation system is also provided, the method comprising: step S701, obtaining a sample data set, wherein each sample data in the sample data set comprises an information list of a sample user and multiple labels corresponding to multiple information in the information list, wherein each of the multiple labels is used to indicate whether the corresponding information is clicked by the sample user; step S702, initializing multiple parameters of the information ranking model; and for each sample data, performing the following operations: step S703, obtaining multiple first representation vectors corresponding to multiple information respectively; step S704, performing attention encoding processing on the multiple first representation vectors to obtain multiple first fused representation vectors corresponding to the multiple first representation vectors respectively; step S705, obtaining multiple first scores based on the multiple first fused representation vectors; step S706, calculating a loss function based on the multiple first scores and the multiple labels corresponding to the sample data; and step S707, adjusting multiple parameters of the information ranking model based on the loss function.

[0107] In some embodiments, the information list obtained can be obtained through the information precise ranking model in the information recommendation system. Among them, the information precise ranking model is the main component of the information precise ranking module, which can perform prediction calculations based on user characteristics, information characteristics, etc., so as to obtain a more accurate information ranking result. The information list output by the information precise ranking model, in addition to including multiple information arranged in the first order after precise ranking, also includes the behavioral feature information of the sample user and the precise ranking hidden layer representation vector corresponding to each information output by the hidden layer of the information precise ranking model, wherein the precise ranking hidden layer representation vector includes the interaction information between each information and the sample user characteristics.

[0108] In some embodiments, each sample data in the sample data set also includes behavioral characteristic information of the sample user, and each first representation vector of the multiple first representation vectors is obtained by concatenating the information representation vector of the corresponding information and at least one of the following representation vectors: a sample user feature representation vector, a position representation vector of the corresponding information, and a representation vector of the corresponding information obtained by fusion of the sample user information based on the information precise sorting model.

[0109] The information representation vector of the corresponding information can be obtained by inputting each information in the information list into the pre-trained language model.

[0110] The sample user feature representation vector can be obtained by inputting the sample user behavior feature data into the unsupervised semantic representation model. User behavior feature data is generally high-dimensional sparse data, which includes user search behavior, web browsing behavior, and user click behavior. By transforming it through the unsupervised semantic representation model, it is possible to obtain user behavior feature information while reducing the dimension of the data and the amount of subsequent calculations.

[0111] The position representation vector of the information can be obtained by applying a method similar to the position embedding in BERT or the position encoding of Transformer. For example, the position representation vector PE can be obtained by applying the following formula:

[0112] PE(t,2i)=sin(t / 10000 2i / d )

[0113] PE(t,2i+1)=cos(t / 10000 2i / d )

[0114] Among them, t is the arrangement position of the corresponding information in the information list, 2i and 2i+1 respectively represent the even bits and odd bits in the position representation vector PE, and d is the dimension of the position representation vector.

[0115] The representation vector of the fused user information obtained by the information precise ranking model is also the refined hidden layer representation vector obtained by the information precise ranking model, which contains the interaction information between each information and the sample user features.

[0116] According to some embodiments, the attention encoding process is performed by a Fastformer encoder.

[0117] According to some embodiments, the attention encoding process is performed by cascading multiple Fastformer encoders, and the multiple Fastformer encoders share parameters.

[0118] During the training process, a loss function may be calculated based on the multiple first scores and the multiple labels corresponding to the sample data. In one example, the loss function may be expressed as the following formula:

[0119]

[0120] Among them, x i represents the i-th output value of the information sorting model, m is the number of samples, K is the number of information in the information list, and θ is the variable in the loss function.

[0121] According to some embodiments, Figure 8As shown, an information sorting device 800 for an information recommendation system is also provided, including: a first acquisition unit 810, configured to acquire a list to be sorted, wherein the list to be sorted includes multiple information acquired in response to a current information request, and the multiple information is arranged in a first order; a second acquisition unit 820, configured to acquire multiple first representation vectors corresponding to the multiple information respectively; an adding unit 830, configured to add a randomly generated placeholder vector to the multiple first representation vectors in response to the absence of a historical information list acquired in response to a previous information request, wherein the dimension of the placeholder vector is the same as the dimension of each first representation vector in the multiple first representation vectors; a first encoding unit 840, configured to perform attention encoding processing on the multiple first representation vectors and the placeholder vector to acquire multiple first fused representation vectors corresponding to the multiple first representation vectors respectively and a second fused representation vector corresponding to the placeholder vector, wherein the second fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to the information sorting of the information list acquired in response to the subsequent information request; and a third acquisition unit 850, configured to acquire multiple first scores based on the multiple first fused representation vectors to determine the second order of the multiple information. The operations of unit 810 to unit 850 are similar to the operations of step S201 to step S205 in the above-mentioned information ranking method for the information recommendation system, and are not described in detail here.

[0122] According to some embodiments, Fig. 9 As shown, an information ranking device 900 for an information recommendation system is also provided, further comprising: a fourth acquisition unit 960, configured to acquire a third fused representation vector corresponding to the historical information list in response to the existence of the historical information list, wherein the third fused representation vector is used to represent the global feature information of the historical information list; a second encoding unit 970, configured to perform attention encoding processing on multiple first representation vectors and the third fused representation vector to acquire multiple fourth fused representation vectors corresponding to the multiple first representation vectors respectively; and a fifth acquisition unit 980, configured to acquire multiple second scores based on the multiple fourth fused representation vectors to determine the third order of the multiple information. Among them, the operations of units 910-950 are similar to those of units 810-850 in the information ranking device 800 for the information recommendation system, and the operations of units 960-980 are similar to those of steps S406-S408 in the information ranking method for the information recommendation system, which will not be repeated here.

[0123] According to some embodiments, Fig.10As shown, a training device 1000 for an information ranking model of an information recommendation system is provided, comprising: a sixth acquisition unit 1010, configured to acquire a sample data set, wherein each sample data in the sample data set comprises an information list of a sample user and a plurality of labels corresponding to a plurality of information in the information list, wherein each label in the plurality of labels is used to indicate whether the corresponding information is clicked by the sample user; an initialization unit 1020, configured to initialize a plurality of parameters of the information ranking model; and an execution unit 1030, configured to perform the operations of the following subunits for each sample data, wherein the execution unit 1030 comprises The method includes: a first acquisition subunit 1031, configured to acquire multiple first representation vectors corresponding to multiple information respectively; an encoding subunit 1032, configured to perform attention encoding processing on the multiple first representation vectors to acquire multiple first fused representation vectors corresponding to the multiple first representation vectors respectively; a second acquisition subunit 1033, configured to acquire multiple first scores based on the multiple first fused representation vectors; a calculation subunit 1034, configured to calculate a loss function based on the multiple first scores and multiple labels corresponding to the sample data; and an adjustment subunit 1035, configured to adjust multiple parameters of the information ranking model based on the loss function. The operations of units 1010-1030 and subunits 1031-1035 are similar to the operations of steps S701-S707 in the training method of the information ranking model for the information recommendation system, and will not be repeated here.

[0124] According to an embodiment of the present disclosure, an electronic device, a readable storage medium and a computer program product are also provided.

[0125] refer to Fig.11 , a block diagram of an electronic device 1100 that can be used as a server or client of the present disclosure will now be described, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0126] like Fig.11As shown, the electronic device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the electronic device 1100 can also be stored. The computing unit 1101, the ROM 1102, and the RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0127] Multiple components in the electronic device 1100 are connected to the I / O interface 1105, including: an input unit 1106, an output unit 1107, a storage unit 1108, and a communication unit 1109. The input unit 1106 can be any type of device that can input information to the electronic device 1100. The input unit 1106 can receive input digital or character information and generate key signal input related to user settings and / or function control of the electronic device, and can include but is not limited to a mouse, a keyboard, a touch screen, a trackpad, a trackball, a joystick, a microphone, and / or a remote control. The output unit 1107 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 1108 can include but is not limited to a disk and an optical disk. The communication unit 1109 allows the electronic device 1100 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as Bluetooth TM devices, 802.11 devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0128] The computing unit 1101 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1101 performs the various methods and processes described above, such as an information ranking method for an information recommendation system. For example, in some embodiments, the information ranking method for an information recommendation system may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the computing unit 1101, one or more steps of the information ranking method for the information recommendation system described above may be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to execute the information ranking method for the information recommendation system in any other appropriate manner (eg, by means of firmware).

[0129] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0130] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0131] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0132] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0133] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0134] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0135] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0136] Although the embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above-mentioned methods, systems and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but only by the claims after authorization and their equivalent scope. Various elements in the embodiments or examples can be omitted or replaced by their equivalent elements. In addition, each step can be performed in an order different from that described in the present disclosure. Further, the various elements in the embodiments or examples can be combined in various ways. It is important that with the evolution of technology, many elements described herein can be replaced by equivalent elements that appear after the present disclosure.

Claims

1. An information ranking method for an information recommendation system, the method comprising: Acquire a list to be sorted, wherein the list to be sorted includes a plurality of information acquired in response to a current information request, and the plurality of information is arranged in a first order; Acquire a plurality of first representation vectors respectively corresponding to the plurality of information; In response to the absence of the history information list obtained in response to the preceding information request, adding a randomly generated placeholder vector to the plurality of first representation vectors, wherein a dimension of the placeholder vector is the same as a dimension of each first representation vector in the plurality of first representation vectors; Performing attention encoding processing on the multiple first representation vectors and the placeholder vector to obtain multiple first fused representation vectors corresponding to the multiple first representation vectors respectively and a second fused representation vector corresponding to the placeholder vector, wherein the second fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to information sorting of the information list obtained in response to the subsequent information request; and Based on the plurality of first fused representation vectors, a plurality of first scores are obtained to determine a second order of the plurality of information.

2. The method according to claim 1, further comprising: In response to the existence of the history information list, obtaining a third fused representation vector corresponding to the history information list, wherein the third fused representation vector is used to represent the global feature information of the history information list; Performing the attention encoding process on the plurality of first representation vectors and the third fused representation vector to obtain a plurality of fourth fused representation vectors respectively corresponding to the plurality of first representation vectors; and Based on the plurality of fourth fused representation vectors, a plurality of second scores are obtained to determine a third order of the plurality of information.

3. The method according to claim 2, wherein: Performing the attention encoding process on the multiple first representation vectors and the third fused representation vector also includes: Based on the attention encoding processing, a fifth fused representation vector corresponding to the third fused representation vector is obtained, wherein the fifth fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to the information sorting of the information list obtained in response to a subsequent information request.

4. The method according to claim 3, wherein: The acquiring of a third fused representation vector corresponding to the historical information list comprises: In response to a plurality of first information lists respectively obtained in response to a plurality of preceding information requests, a plurality of sixth fused representation vectors respectively corresponding to the plurality of first information lists are obtained, wherein each sixth fused representation vector in the plurality of sixth fused representation vectors is obtained by performing the attention encoding process on the corresponding first information list to represent the global feature information of the corresponding first information list; and Perform weighted summation on the multiple sixth fused representation vectors to obtain the third fused representation vector.

5. The method according to any one of claims 1 to 4, wherein: The attention encoding process is performed by a Fastformer encoder.

6. The method according to any one of claims 1 to 4, wherein: The attention encoding process is performed by cascading multiple Fastformer encoders, and the multiple Fastformer encoders share parameters.

7. The method according to any one of claims 1 to 4, wherein: The to-be-sorted list is obtained through an information precise sorting model in the information recommendation system.

8. The method according to claim 7, wherein: Each of the multiple first representation vectors is obtained by concatenating the information representation vector of the corresponding information and at least one of the following representation vectors: a user feature representation vector, a position representation vector of the corresponding information, and a representation vector of the corresponding information obtained by fusion user information based on the information precise sorting model.

9. A training method for an information ranking model for an information recommendation system, wherein: The information sorting model is used to implement the information sorting method according to any one of claims 1 to 8, and the training method comprises: Acquire a sample data set, wherein each sample data in the sample data set includes an information list of a sample user and a plurality of tags respectively corresponding to a plurality of information in the information list, wherein each tag in the plurality of tags is used to indicate whether the corresponding information is clicked by the sample user; Initializing multiple parameters of the information sorting model; and For each sample data, perform the following operations: Acquire a plurality of first representation vectors respectively corresponding to the plurality of information; Performing attention encoding processing on the multiple first representation vectors to obtain multiple first fused representation vectors respectively corresponding to the multiple first representation vectors; Based on the multiple first fusion representation vectors, obtaining multiple first scores; Calculating a loss function based on the multiple first scores and the multiple labels corresponding to the sample data; and Multiple parameters of the information ranking model are adjusted based on the loss function.

10. The training method according to claim 9, wherein: The attention encoding process is performed by a Fastformer encoder.

11. The training method according to claim 9, wherein: The attention encoding process is performed by cascading multiple Fastformer encoders, and the multiple Fastformer encoders share parameters.

12. The training method according to any one of claims 9 to 11, wherein: The information list is obtained through an information precise ranking model in the information recommendation system.

13. The training method according to claim 12, wherein: Each sample data in the sample data set also includes behavioral characteristic information of the sample user, and each first representation vector in the multiple first representation vectors is obtained by concatenating the information representation vector of the corresponding information and at least one of the following representation vectors: a sample user feature representation vector, a position representation vector of the corresponding information, and a representation vector of the corresponding information obtained by fusion of sample user information based on the information precise sorting model.

14. An information ranking device for an information recommendation system, the device comprising: A first acquisition unit is configured to acquire a to-be-sorted list, wherein the to-be-sorted list includes a plurality of information acquired in response to a current information request, and the plurality of information is arranged in a first order; A second acquisition unit is configured to acquire a plurality of first representation vectors respectively corresponding to the plurality of information; an adding unit, configured to, in response to the absence of a list of historical information acquired in response to the request for the preceding information, add a randomly generated placeholder vector to the plurality of first representation vectors, wherein a dimension of the placeholder vector is the same as a dimension of each first representation vector in the plurality of first representation vectors; a first encoding unit configured to perform attention encoding processing on the plurality of first representation vectors and the placeholder vector to obtain a plurality of first fused representation vectors respectively corresponding to the plurality of first representation vectors and a second fused representation vector corresponding to the placeholder vector, wherein the second fused representation vector is used to represent the global feature information of the list to be sorted, so as to be applied to information sorting of the information list obtained in response to the subsequent information request; and The third acquisition unit is configured to acquire a plurality of first scores based on the plurality of first fusion representation vectors to determine a second order of the plurality of information.

15. The apparatus according to claim 14, further comprising: a fourth acquisition unit, configured to acquire, in response to the existence of the history information list, a third fused representation vector corresponding to the history information list, wherein the third fused representation vector is used to represent the global feature information of the history information list; A second encoding unit is configured to perform the attention encoding process on the plurality of first representation vectors and the third fused representation vector to obtain a plurality of fourth fused representation vectors respectively corresponding to the plurality of first representation vectors; and The fifth acquisition unit is configured to acquire a plurality of second scores based on the plurality of fourth fused representation vectors to determine a third order of the plurality of information.

16. A training device for an information ranking model for an information recommendation system, wherein: The information sorting model is used to implement the information sorting method according to any one of claims 1 to 8, and the device comprises: a sixth acquisition unit, configured to acquire a sample data set, wherein each sample data in the sample data set includes an information list of a sample user and a plurality of tags respectively corresponding to a plurality of information in the information list, wherein each tag in the plurality of tags is used to indicate whether the corresponding information is clicked by the sample user; an initialization unit, configured to initialize multiple parameters of the information sorting model; and The execution unit is configured to execute the operations of the following sub-units for each sample data, wherein the execution unit includes: A first acquisition subunit is configured to acquire a plurality of first representation vectors respectively corresponding to the plurality of information; an encoding subunit, configured to perform attention encoding processing on the plurality of first representation vectors to obtain a plurality of first fused representation vectors respectively corresponding to the plurality of first representation vectors; A second acquisition subunit is configured to acquire a plurality of first scores based on the plurality of first fused representation vectors; a calculation subunit, configured to calculate a loss function based on the plurality of first scores and the plurality of labels corresponding to the sample data; and An adjustment subunit is configured to adjust multiple parameters of the information ranking model based on the loss function.

17. An electronic device comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8 or 9-13.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-8 or 9-13.

19. A computer program product comprising a computer program, wherein: The computer program implements the method of any one of claims 1-8 or 9-13 when executed by a processor.

Images