Hypergraph-based collaborative filtering recommended

Abstract

An electronic device and method for performing hypergraph-based collaborative filtering recommendations are provided. 【Solution means】The electronic device receives a collaborative filtering graph corresponding to a user set and an item set. The electronic device determines a first user embedding set and a first item embedding set. The electronic device applies a semantic clustering model to determine a second user embedding set and a second item embedding set. The electronic device constructs a hypergraph to determine a third user embedding set and a third item embedding set. The electronic device determines a first contrast loss and a second contrast loss to determine a collaborative filtering score. The electronic device determines item recommendations for a user based on the determined collaborative filtering score. The electronic device renders the determined recommended items on a display device.

Description

[Technical Field]

[0001] [Incorporation by cross-referencing / citation with related applications] This application claims priority to U.S. Provisional Patent Application No. 63 / 365,540, filed on 31 May 2022, and also claims priority to U.S. Patent Application No. 18 / 319,096, filed with the U.S. Patent and Trademark Office on 17 May 2023, the contents of which are incorporated herein by reference in their entirety.

[0002] Various embodiments of this disclosure relate to recommendation systems. Specifically, various embodiments of this disclosure relate to electronic devices and methods for hypergraph-based collaborative filtering recommendations. [Background technology]

[0003] Advances in the field of recommendation systems have led to the development of various types of recommendation models that can provide users with personalized recommendations. Recommendation systems can be used in diverse fields such as media and entertainment, finance, e-commerce, retail, banking, and telecommunications. Typically, recommendation systems can recommend items (e.g., movies) related to a particular domain (e.g., a movie domain for over-the-top platforms) to a user based on parameters such as the user's personal details / profile, viewing history, movie consumption patterns (e.g., time spent watching each movie), and the genre of movies in the viewing history. Traditional recommendation systems sometimes ignore higher-order relationships between users and items. Therefore, traditional recommendation systems can be suboptimal and often make inaccurate recommendations. [Overview of the project] [Problems that the invention aims to solve]

[0004] Those skilled in the art will be able to see the limitations and disadvantages of conventional methods by comparing the described system with some aspects of the disclosure shown with reference to the drawings in the remainder of this application. [Means for solving the problem]

[0005] The present invention provides an electronic apparatus and method for hypergraph-based collaborative filtering recommendations, as illustrated and / or described in substantially relation to at least one figure and more fully provided in the claims.

[0006] These and other features and advantages of the disclosure can be understood by considering the following detailed description of the disclosure with reference to the accompanying drawings, which indicate the same elements throughout by the same reference numerals. [Brief explanation of the drawing]

[0007] [Figure 1] This block diagram shows an exemplary network environment for hypergraph-based collaborative filtering recommendations according to embodiments of the present disclosure. [Figure 2] This is a block diagram showing an exemplary electronic device according to an embodiment of the present disclosure, as shown in Figure 1. [Figure 3] This figure shows an exemplary scenario of a collaborative filtering graph according to an embodiment of the present disclosure. [Figure 4A] This figure shows an exemplary processing pipeline for hypergraph-based collaborative filtering recommendations according to embodiments of the present disclosure. [Figure 4B] This figure shows an exemplary processing pipeline for hypergraph-based collaborative filtering recommendations according to embodiments of the present disclosure. [Figure 5] This figure shows an exemplary scenario of a hypergraph embedding architecture according to embodiments of the present disclosure. [Figure 6] This figure shows an exemplary scenario of comparative learning according to the embodiments of the present disclosure. [Figure 7]This figure shows an exemplary scenario in which an item set is recommended to a user set according to an embodiment of the present disclosure. [Figure 8] This flowchart shows the operation of an exemplary method for hypergraph-based collaborative filtering recommendations according to embodiments of the present disclosure. [Modes for carrying out the invention]

[0008] Electronic devices and methods for hypergraph-based collaborative filtering recommendations can be found in embodiments described below. Exemplary embodiments of this disclosure can provide an electronic device capable of receiving a collaborative filtering graph corresponding to a user set and an item set associated with the user set. The collaborative filtering graph can correspond to user-item interaction data. The electronic device can determine a first user embedding set and a first item embedding set based on the received collaborative filtering graph. The electronic device can apply a semantic clustering model to each of the determined first user embedding set and the determined first item embedding set. The electronic device can determine a second user embedding set and a second item embedding set based on the application of the semantic clustering model. The electronic device can construct a hypergraph from the received collaborative filtering graph. The electronic device can determine a third user embedding set and a third item embedding set based on the constructed hypergraph. The electronic device can determine a first contrastive loss based on the determined second user embedding set and the determined third user embedding set. The electronic device can determine a second control loss based on the determined second and third item embedding sets. Furthermore, the electronic device can determine a collaborative filtering score based on the determined first and second control losses. Subsequently, the electronic device can determine item recommendations for the user based on the determined collaborative filtering score. The electronic device can then render the determined recommended items on a display device.

[0009] Typically, a recommendation system can recommend items related to a domain based on one or more parameters, such as personal details related to the target user (e.g., age, gender, and demographic information), item consumption history, item consumption patterns, and the similarity between the recommended items and items consumed by the target user. Some other typical recommendation systems can generate user embeddings based on features extracted from one or more parameters. A recommendation system can generate embeddings related to items (e.g., movies) based on domain characteristics (e.g., genre, length, cast, and studio). A recommendation system can compare item embeddings in the target user's item consumption history with items in the domain. A recommendation system can recommend domain items related to embeddings similar to those in the item consumption history.

[0010] Furthermore, some conventional recommendation systems for content recommendations on OTT platforms or streaming services can accept regular bipartite graphs as input. Such bipartite graphs can include a set of edges connecting pairs of nodes. Moreover, bipartite graphs can also provide only interdomain correlations (e.g., user-item correlations). Conventional recommendation systems can also simultaneously learn intradomain similarity (e.g., user-user correlations or item-item correlations). Generalizing such intradomain similarity can be difficult. Furthermore, since most users may not interact with all items in an item set, the data related to intradomain similarity can be sparse. Consequently, the distribution of edge types can become highly imbalanced. Therefore, the recommendation system may not be optimal.

[0011] To address the aforementioned issues, the disclosed electronic device can employ a hypergraph-based collaborative filtering framework for item recommendations. Here, the electronic device can determine a second user embedding set and a second item embedding set by applying a semantic clustering model to each of the determined first user embedding set and first item embedding set. The electronic device can obtain positive and negative samples based on the application of the semantic clustering model. Furthermore, the electronic device can construct a hypergraph from the received collaborative filtering graph. Using the constructed hypergraph, higher-order relationships between user sets and item sets can be explored. Furthermore, the electronic device can determine a third user embedding set and a third item embedding set based on the constructed hypergraph. The determined third user embedding set and the determined third item embedding set may include features related to the potential relationships between user sets and item sets captured in the constructed hypergraph. The electronic device can employ a contrastive framework to determine a first contrast loss and a second contrast loss in order to determine recommendations. In some embodiments, the electronic device can determine final user and item embeddings that take into account higher-order relationships captured in the constructed hypergraph, allowing unstructured but similar nodes (e.g., user sets and item sets) to be placed close together and dissimilar nodes to be placed far apart. The final user and item embeddings can maintain a balance between the higher-order view and the collaborative view of the interaction data inferred from the collaborative filtering graph. This balance between the final user and item embeddings can help achieve optimal results in downstream tasks such as recommendation systems, user clustering, community clustering, and classification tasks.

[0012] Figure 1 is a block diagram illustrating an exemplary network environment for hypergraph-based collaborative filtering recommendations according to an embodiment of the present disclosure. Figure 1 shows a network environment 100. The network environment 100 may include an electronic device 102, a server 104, a database 106, and a communication network 108. The electronic device 102 may include a semantic clustering model 110, a recommendation model 112, a graph neural network (GNN) model 114, a first hypergraph convolutional network (HGCN) model set 116A, and a second HGCN model set 116B. Figure 1 further shows a collaborative filtering graph 118 that can be stored in the database 106. Furthermore, a user 120 that may be associated with or capable of operating the electronic device 102 is also shown.

[0013] The electronic device 102 may include suitable logic, circuitry, interfaces, and / or code that can be configured to receive a collaborative filtering graph 118 corresponding to a user set and an item set associated with the user set. The electronic device 102 may receive the collaborative filtering graph 118 from a database 106 (which can store the collaborative filtering graph 118) via a server 104. The electronic device 102 may determine a first user embedding set and a first item embedding set based on the received collaborative filtering graph 118. The electronic device 102 may apply a semantic clustering model 110 to each of the determined first user embedding set data and the determined first item embedding set data. The electronic device 102 may determine a second user embedding set and a second item embedding set based on the application of the semantic clustering model 110. The electronic device 102 may construct a hypergraph from the received collaborative filtering graph 118. The electronic device 102 may determine a third user embedding set and a third item embedding set based on the constructed hypergraph. The electronic device 102 can determine a first control loss based on a determined second user embedding set and a determined third user embedding set. The electronic device 102 can determine a second control loss based on a determined second item embedding set and a determined third item embedding set. For example, the electronic device 102 can determine a first control loss based on a determined first user embedding set having spectral similarity and a local collaborative graph, a second user embedding set from a hypergraph, and a determined third item embedding set. The electronic device 102 can determine a second control loss based on a determined first item embedding set from a local collaborative graph grouped by semantic similarity, and a determined second item embedding set from a hypergraph.Furthermore, the electronic device 102 can determine a collaborative filtering score based on the determined first contrast loss and the determined second contrast loss. Then, the electronic device 102 can determine recommendations for items for a user (e.g., user 120) based on the determined collaborative filtering score. The electronic device 102 can render the determined recommended items on a display device.

[0014] Examples of the electronic device 102 include, but are not limited to, a computer device, a smartphone, a cellular phone, a mobile phone, a gaming device, a mainframe machine, a server, a computer workstation, a machine learning device (e.g., corresponding to or hosting computing resources, memory resources, and networking resources), and / or a consumer electronics (CE) device.

[0015] Server 104 may include suitable logic, circuitry, interfaces, and / or code that can be configured to receive a collaborative filtering graph 118 from database 106 corresponding to user sets and item sets associated with user sets. Server 104 can determine a first user embedding set and a first item embedding set based on the received collaborative filtering graph 118. Server 104 can apply a semantic clustering model 110 to each of the determined first user embedding set data and the determined first item embedding set data. Server 104 can determine a second user embedding set and a second item embedding set based on the application of the semantic clustering model 110. Server 104 can construct a hypergraph from the received collaborative filtering graph 118. Server 104 can determine a third user embedding set and a third item embedding set based on the constructed hypergraph. Server 104 can determine a first control loss based on the determined second user embedding set and the determined third user embedding set. Server 104 can determine a second control loss based on the determined second and third item embedding sets. Server 104 can determine a collaborative filtering score based on the determined first and second control losses. Server 104 can determine item recommendations for users such as user 120 based on the determined collaborative filtering score. Server 104 can render the determined recommended items on a display device.

[0016] Server 104 can be implemented as a cloud server and can execute operations via, for example, web applications, cloud applications, HTTP requests, repository operations, and file transfers. Other implementation examples of server 104 include, but are not limited to, database servers, file servers, web servers, media servers, application servers, mainframe servers, machine learning servers (e.g., corresponding to or hosting computing resources, memory resources, and networking resources), or cloud computing servers.

[0017] In at least one embodiment, server 104 can be implemented as a plurality of distributed cloud-based resources by using multiple techniques well known to those skilled in the art. Those skilled in the art will understand that the scope of the present disclosure can also not be limited to the implementation of server 104 and electronic device 102 as two independent entities. In some embodiments, without departing from the scope of the present disclosure, the functions of server 104 can be wholly or at least partially incorporated into electronic device 102. In some embodiments, server 104 can host database 106. Alternatively, server 104 can be communicatively coupled to database 106 separately from database 106.

[0018] Database 106 may include suitable logic, interfaces, and / or code that can be configured to store the collaborative filtering graph 118. Database 106 may also store information related to user sets and item sets. Database 106 may derive data from relational or non-relational databases, or from comma-separated values ​​(csv) file sets in conventional or big data storage. Database 106 may be stored or cached in a device such as a server (e.g., server 104) or electronic device 102. The device storing database 106 may be configured to receive queries for the collaborative filtering graph 118 from the electronic device 102 or server 104. In response, the device of database 106 may be configured to retrieve the queried collaborative filtering graph 118 based on the received query and provide it to the electronic device 102 or server 104.

[0019] In some embodiments, the database 106 can be hosted on multiple servers stored in the same or different locations. The operation of the database 106 can be performed using hardware including a processor, a microprocessor (for example, one or more operations that perform or control the execution of), a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC). In some other cases, the database 106 can also be implemented using software.

[0020] The communication network 108 may include a communication medium that enables the electronic device 102 and the server 104 to communicate with each other. The communication network 108 may be either a wired or wireless connection. Examples of the communication network 108 include, but are not limited to, the internet, cloud networks, cellular or wireless mobile networks (such as Long-Term Evolution and 5th Generation (5G) New Radio (NR)), satellite communication systems (e.g., using low-Earth orbit satellites), Wireless Fidelity (Wi-Fi) networks, personal area networks (PANs), local area networks (LANs), or metropolitan area networks (MANs). Various devices within the network environment 100 may be configured to connect to the communication network 108 according to various wired and wireless communication protocols. Examples of such wired and wireless communication protocols include, but are not limited to, at least one of the following: Transmission Control Protocol and Internet Protocol (TCP / IP), User Datagram Protocol (UDP), Hypertext Transfer Protocol (HTTP), File Transfer Protocol (FTP), ZigBee, EDGE, IEEE 802.11, Light Fidelity (Li-Fi), 802.16, IEEE 802.11s, IEEE 802.11g, Multihop Communication, Wireless Access Point (AP), Device-to-Device Communication, Cellular Communication Protocol, and Bluetooth (BT) Communication Protocol.

[0021] The semantic clustering model 110 can be a machine learning (ML) model that clusters an input dataset into cluster sets, where each cluster may contain a subset of similar datasets. The semantic clustering model 110 of this disclosure can be applied to each of a determined first user embedding set and a determined first item embedding set. The semantic clustering model 110 can determine a second user embedding set and a second item embedding set from the determined first user embedding set and the determined first item embedding set, respectively. In some embodiments, the semantic clustering model 110 can correspond to a spectral clustering model configured for dimensionality reduction, where the dimensions of the determined second user embedding set and the determined second item embedding set (determined based on the application of the semantic clustering model 110) can be smaller than the dimensions of the determined first user embedding set and the determined first item embedding set, respectively.

[0022] The recommendation model 112 can be an ML model that can determine recommendations based on various criteria. For example, the recommendation model 112 can recommend one or more products to a customer based on the customer's purchase history, geographical location, and customer needs. The recommendation model 112 of this disclosure can determine item recommendations for user 120 based on a determined collaborative filtering score.

[0023] The GNN model 114 can be a deep learning model capable of constructing a graph based on an received dataset. The GNN model 114 can then process the constructed graph and perform inference based on the constructed graph. The GNN model 114 of this disclosure can be applied to an received collaborative filtering graph 118. The GNN model 114 can process the applied collaborative filtering graph 118 to determine a first user embedding set and a first item embedding set, respectively.

[0024] The first HGCN model set 116A can be an ML model that processes information related to the hypergraph and makes inference decisions based on the processing. The first HGCN model set 116A can be applied to a fourth user embedding set. Here, the fourth user embedding set can be determined based on a set of user-item correlations and a set of user-user correlations, and these correlations can be determined based on the constructed hypergraph. The first HGCN model set 116A can determine a third user embedding set based on the determined fourth user embedding set. The second HGCN model set 116B can be applied to a fourth user embedding set. Here, the fourth item embedding set can be determined based on a set of item-user correlations, and these correlations can be determined based on the constructed hypergraph. The second HGCN model set 116B can determine a third item embedding set based on the determined fourth item embedding set.

[0025] The GNN model 114, the first HGCN model set 116A, and the second HGCN model set 116B can be graphic neural network (GNN) models. A GNN model may include suitable logic, circuitry, interfaces, and / or code that can be configured to classify or analyze input graph data to produce output results for a specific real-time application. For example, a trained GNN model, such as GNN model 114, can recognize different nodes in the input graph data and the edges between each node in the input graph data. The edges may correspond to different connections or relationships between each node in the input graph data. Based on the recognized nodes and edges, the trained GNN model 114 can classify different nodes in the input graph data into different labels or classes. In one example, a particular node in the input graph data may include a set of associated features. The feature set may include, but is not limited to, media content type, media content length, media content genre, and the geographical location of user 120. Furthermore, each edge may connect to different nodes having similar feature sets. The electronic device 102 can be configured to encode a feature set using a GNN model to generate feature vectors. After encoding, information can be passed between a specific node and its neighbors connected via edges. Based on the information passed to the neighbors, a final vector can be generated for each node. Such a final vector can provide reliable and accurate information related to a particular node by including information related to the feature sets of that node and its neighbors. As a result, the GNN model can analyze information represented as input graph data. The GNN model can be implemented using hardware including a processor, a microprocessor (e.g., one or more operations that perform or control the execution of), a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC). In some other cases, the GNN model can be code, a program, or a set of software instructions.GNN models can also be implemented using a combination of hardware and software.

[0026] In some embodiments, a GNN model can have multiple classification layers, each successive layer which can use the output of the previous layer as input, to classify different nodes in the input graph data. Each classification layer can be associated with multiple edges, and each edge can be further associated with multiple weights. During training, the GNN model can be configured to filter or remove edges or nodes based on the input graph data to further provide the output results of the GNN model (i.e., the graph representation). Examples of GNN models include, but are not limited to, graph convolutional networks (GCNs), hypergraph convolutional networks (HGCNs), graph space-time networks with GCNs, recurrent neural networks (RNNs), deep Bayesian neural networks, and / or combinations of such networks.

[0027] In one embodiment, the semantic clustering model 110, the recommendation model 112, the GNN model 114, the first HGCN model set 116A, and the second HGCN model set 116B may be machine learning (ML) models. Each ML model can be trained to identify relationships between inputs, such as features in the training dataset, and output labels. Each ML model can be defined by hyperparameters such as the number of weights, cost function, input size, and number of layers. The parameters of each ML model can be adjusted to update the weights toward the global minimum of the cost function of the corresponding ML model. After training for multiple epochs on feature information in the training dataset, each ML model can be trained to output recommendations, predictions, information related to cluster sets, or classification results for a set of inputs. For example, the ML model associated with the recommendation model 112 can recommend items for user 120.

[0028] Each ML model may include electronic data that can be implemented as a software component of an application executable on, for example, an electronic device 102. Each ML model may rely on libraries, external scripts, or other logic / instructions for execution by a processing unit. Each ML model may include code and routines configured to enable a computer device, such as the electronic device 102, to perform one or more actions, such as making a recommendation. In addition to or instead of this, each ML model may also be implemented using hardware including a processor, microprocessor, field-programmable gate array (FPGA), or application-specific integrated circuit (ASIC). Alternatively, in some embodiments, a combination of hardware and software may be used to implement the ML model.

[0029] The collaborative filtering graph 118 can provide a compact representation of the interaction between the user set and the item set. The user set and the item set can be represented by the user node set and the item node set, respectively. Each edge of the collaborative filtering graph 118 can provide an interaction between pairs of nodes. Thus, the collaborative filtering graph 118 can be a bipartite graph. Further details regarding the collaborative filtering graph 118 are shown in Figure 3.

[0030] The electronic device 102 can receive a collaborative filtering graph 118 corresponding to a user set and an item set associated with the user set during operation. For example, the database 106 can store the collaborative filtering graph 118. The electronic device 102 can request the collaborative filtering graph 118 from the database 106 and receive the requested collaborative filtering graph 118 from the database 106 via the server 104. The collaborative filtering graph 118 can be a bipartite graph that can represent various interactions between user sets and item sets. User sets and item sets can be represented as nodes in the collaborative filtering graph 118. Each edge of the collaborative filtering graph 118 can represent an interaction between a pair of nodes in the collaborative filtering graph 118. For example, user "A" may have bookmarked item "B". Therefore, the collaborative filtering graph 118 may include an edge between user "A" and item "B" that represents user "A" bookmarking item "B". Further details regarding the collaborative filtering graph 118 are explained in Figure 3, for example.

[0031] The electronic device 102 can determine a first user embedding set and a first item embedding set based on the received collaborative filtering graph 118. It can be understood that the embeddings can correspond to vector representations of features related to entities. Each user embedding in the first user embedding set can provide features related to a subset of items from an item set that is thought to have been viewed or selected by the user associated with the corresponding user embedding. Each item embedding in the first item embedding set can correspond to features related to a subset of users from a user set that is thought to have viewed or selected the items associated with the corresponding item embedding.

[0032] In the neighborhood aggregation phase, a collaborative filtering graph 118 can be used to generate a first user embedding set and a first item embedding set having multiple "k" hops. Furthermore, local collaborative signals can be techniques for addressing user-item interactions in a way that makes the hypergraph signals appear like global signals. In one embodiment, the process of generating the first user embedding set and the first item embedding set having multiple "k" hops described above can be iteratively performed at odd and even hops, respectively. For example, at the first hop, user embeddings associated with user "U1" can be represented by vectors such as items "I1" and "I2". At the third hop, further items can be added from the collaborative filtering graph 118 to the user embeddings associated with user "U1". Similarly, at even hops, each item can be associated with multiple users (users who are thought to have had some interaction with that item). Thus, the first hop aggregation can include user "U1" represented as a vector in terms of directly connected items such as "I1" and "I2". The second hop can be useful in representing items as vectors from the perspective of users who can connect to certain items directly or indirectly. For example, user "U1" can be represented on the first hop as an aggregate of items "I1" and "I2" that user "U1" can directly interact with. However, if item "I1" is also connected to user "U2", and user "U2" is connected to item "I5", then an indirect connection can also exist between user "U1" and item "I5". The above relationships can be aggregated on the third hop. Further details regarding the determination of the first user embedding set and the first item embedding set are explained in Figure 4A, for example.

[0033] The electronic device 102 can apply the semantic clustering model 110 to each of the determined first user embedding set data and the determined first item embedding set data. Based on the application of the semantic clustering model 110, the semantic views of the user set and item set can be determined. User subsets and item subsets that can be directly connected to each other are considered similar and can be grouped together to form a cluster. Further details regarding the application of the semantic clustering model are explained, for example, in Figure 4A.

[0034] The electronic device 102 can determine a second user embedding set and a second item embedding set based on the application of the semantic clustering model 110. The second user embedding set and the second item embedding set can be extracted from the semantic views of the user set and the item set. Further details regarding the determination of the second user embedding set and the second item embedding set are explained in Figure 4A, for example.

[0035] The electronic device 102 can construct a hypergraph from the received collaborative filtering graph 118. The hypergraph can be a graph that represents the higher-order relationships between user sets and item sets related to the collaborative filtering graph 118 using hyperedges. In an OTT platform, users are not always directly or indirectly connected to each other through item nodes. Collaborative filtering graphs can be prone to information loss. To mitigate the above problem, a third user embedding set and a third item embedding set can be determined from the constructed hypergraph. Further details regarding the construction of the hypergraph are explained in Figure 5, for example.

[0036] The electronic device 102 can determine a third user embedding set and a third item embedding set based on the constructed hypergraph. The third user embedding set and the third item embedding set thus determined may include information relating to higher-order relationships between user sets and item sets. Furthermore, the third user embedding set and the third item embedding set may also include features relating to potential relationships between user sets and item sets captured in the constructed hypergraph. Further details regarding the determination of the third user embedding set and the third item embedding set are shown, for example, in Figure 5.

[0037] The electronic device 102 can determine a first contrast loss based on a determined second user embedding set and a determined third user embedding set. The first contrast loss can be a variation of nearest-neighbor contrastive learning (NNCLR) of the visual representation, which can be determined based on the determined second user embedding set and the determined third user embedding set. Further details regarding the determination of the first contrast loss are explained in Figure 4B, for example.

[0038] The electronic device 102 can determine a second control loss based on the determined second item embedding set and the determined third item embedding set. The second control loss may be a variation of NNCLR that can be determined based on the determined second item embedding set and the determined third item embedding set. Further details regarding the determination of the second control loss are explained in Figure 4B, for example.

[0039] The electronic device 102 can determine a collaborative filtering score based on the determined first control loss and the determined second control loss. The collaborative filtering score can provide a score set for each user's item set in the user set. The score set can be used as the basis for recommendation decisions for the user set. Further details regarding the determination of the collaborative filtering score are explained in Figure 4B, for example.

[0040] The electronic device 102 can determine item recommendations for user 120 based on the determined collaborative filtering score. For each user, the item most likely to be associated with the highest score can be selected as a recommendation. For example, for user 120, the score set may be "0.78", "0.67", and "0.82". Therefore, the item associated with the score "0.82" can be determined as a recommendation for user 120. Further details regarding the determination of item recommendations are explained in Figure 4B, for example.

[0041] The electronic device 102 can render the determined recommended item on the display device. In one example, the determined recommended item may be an action movie that can be displayed as a recommendation on the display device. As a result, user 120 can select this action movie and then play it. Further details regarding the rendering of the determined recommended item are explained in Figure 4B, for example.

[0042] The electronic device 102 can employ contrast learning with positive and negative pair formation from hypergraph embedding, GCN collaborative structural embedding, and spectral cluster-based semantic embedding. Forming positive pairs with a third user embedding set and a third item embedding set using the semantic clustering model 110 can help retain similarity information for better learning. The electronic device 102 can be used to provide personalized recommendations on over-the-top (OTT) platforms and e-commerce platforms, etc. Here, the electronic device 102 can further process the recommendation task as a link prediction task or an edge prediction task for each item in the item set.

[0043] Figure 2 is a block diagram showing an exemplary electronic device of Figure 1 according to an embodiment of the present disclosure. The description of Figure 2 will be made in relation to the elements of Figure 1. Figure 2 shows an exemplary electronic device 102. The electronic device 102 may include a circuit 202, a memory 204, an input / output (I / O) device 206, a network interface 208, a semantic clustering model 110, a recommendation model 112, a GNN model 114, a first HGCN model set 116A, and a second HGCN model set 116B. The memory 204 may store a collaborative filtering graph 118. The input / output (I / O) device 206 may include a display device 210.

[0044] Circuit 202 may include preferred logic, circuits, and / or interfaces that can be configured to execute program instructions related to different operations performed by the electronic device 102. These operations may include receiving a co-filtering graph, applying a GNN model, determining a first embedding, applying a semantic clustering model, determining a second embedding, building a hypergraph, determining a third embedding, determining a first control loss, determining a second control loss, determining a co-filtering score, determining a recommendation, and rendering a recommendation. Circuit 202 may include one or more processing units that can be implemented as independent processors. In some embodiments, one or more processing units may be implemented as an integrated processor or group of processors that collectively execute the functions of one or more processing units. Circuit 202 may be implemented based on several processor technologies well known in the art. Implementation examples of circuit 202 may include x86-based processors, graphics processing units (GPUs), reduced instruction set computing (RISC) processors, application-specific integrated circuit (ASIC) processors, composite instruction set computing (CISC) processors, microcontrollers, central processing units (CPUs), and / or other control circuits.

[0045] Memory 204 may include preferred logic, circuits, interfaces and / or code that can be configured to store one or more instructions executed by circuit 202. The one or more instructions stored in memory 204 may be configured to perform different operations of circuit 202 (and / or electronic device 102). Memory 204 may be further configured to store a cooperative filtering graph 118. In some embodiments, memory 204 may also store user embeddings and item embeddings. Examples of implementations of memory 204 include, but are not limited to, random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), hard disk drives (HDDs), solid-state drives (SSDs), CPU caches, and / or secure digital (SD) cards.

[0046] The I / O device 206 may include suitable logic, circuitry, interfaces, and / or code that can be configured to receive inputs and provide outputs based on the received inputs. For example, the I / O device 206 may receive a first user input indicating a request to generate item recommendations for user 120. The I / O device 206 may be further configured to display or render the recommended items. The I / O device 206 may include a display device 210. Examples of the I / O device 206 include, but are not limited to, a display (e.g., a touch screen), a keyboard, a mouse, a joystick, a microphone, or a speaker. Further examples of the I / O device 206 include braille I / O devices such as braille keyboards and braille readers.

[0047] The network interface 208 may include suitable logic, circuitry, interfaces, and / or code that can be configured to facilitate communication between the electronic device 102 and the server 104 via the communication network 108. The network interface 208 may be implemented to support wired or wireless communication between the electronic device 102 and the communication network 108 using a variety of known techniques. The network interface 208 may include, but is not limited to, an antenna, a radio frequency (RF) transceiver, one or more amplifiers, a tuner, one or more oscillators, a digital signal processor, a coder-decoder (CODEC) chipset, a subscriber ID module (SIM) card, or a local buffer circuit.

[0048] The network interface 208 can be configured to communicate wirelessly with networks such as the Internet, intranet, wireless network, cellular telephone network, wireless local area network (LAN), or metropolitan area network (MAN). Wireless communication can be configured to use one or more of several communication standards, protocols, and technologies, such as Global System for Mobile Communications (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (W-CDMA), Long-Term Evolution (LTE), 5th Generation (5G) New Radio (NR), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Bluetooth, Wireless Fidelity (WiFi) (such as IEEE 802.11a, IEEE 802.11b, IEEE 802.11g, or IEEE 802.11n), Voice over Internet Protocol (VoIP), Light Fidelity (Li-Fi), Worldwide Interoperability for Microwave Access (Wi-MAX), and protocols for email, instant messaging, and short message service (SMS).

[0049] The display device 210 may include suitable logic, circuitry, and interfaces that can be configured to display or render determined recommended items. The display device 210 may be a touch screen that allows a user (e.g., user 120) to provide user input via the display device 210. The touch screen may be at least one of resistive touch screens, capacitive touch screens, or thermal touch screens. The display device 210 may be implemented through several known technologies, including, but is not limited to, liquid crystal display (LCD) displays, light-emitting diode (LED) displays, plasma displays, or organic LED (OLED) display technologies, or at least one of other display devices. According to one embodiment, the display device 210 may mean a display screen for a head-mounted display (HMD), smart glasses, see-through display, projection-based display, electrochromic display, or transparent display. Various operations of the circuit 202 for implementing hypergraph-based collaborative filtering recommendations are further described, for example, in Figures 4A and 4B.

[0050] Figure 3 shows an exemplary scenario of a collaborative filtering graph according to an embodiment of the present disclosure. The description of Figure 3 will be made in relation to the elements of Figures 1 and 2. Figure 3 shows an exemplary scenario 300. Scenario 300 may include a user set and an item set. The user set may include a first user 302A, a second user 302B, and a third user 302C. The item set may include a first item 304A, a second item 304B, and a third item 304C. A series of operations related to scenario 300 will be described here.

[0051] In Scenario 300 of Figure 3, the item set, such as the first item 304A, the second item 304B, and the third item 304C, can be different multimedia content such as sitcoms, news reports, and digital games. First, the first user 302A, the second user 302B, and the third user 302C can be registered on the OTT platform. Each user in the user set can view one or more items from the item set and rate each of the viewed items on a scale from "1" to "5". A rating of "1" may mean that the user does not like the rated item at all, and a rating of "5" may mean that the user likes the rated item very much.

[0052] Referring to Figure 3, the first user 302A can interact with the first item 304A and give it a rating of "5" as shown by edge 306A. The second user 302B can interact with the first item 304A and the second item 304B as shown by edges 306B and 306C, respectively. Furthermore, the second user 302B can rate the first item 304A as "5" and the second item 304B as "2". That is, it can be concluded that the second user 302B prefers the first item 304A to the second item 304B. The third user 302C can interact with the first item 304A, the second item 304B, and the third item 304C as shown by edges 306D, 306E, and 306F, respectively. Furthermore, the third user 302C can rate the first item 304A, the second item 304B, and the third item 304C as "5", "5", and "5", respectively. In other words, it can be considered that the third user 302C equally likes the first item 304A, the second item 304B, and the third item 304C.

[0053] Furthermore, Scenario 300 in Figure 3 is for illustrative purposes only and should not be interpreted as limiting the scope of this disclosure.

[0054] Figures 4A and 4B show exemplary processing pipelines for hypergraph-based collaborative filtering recommendations according to embodiments of the present disclosure. The description of Figures 4A and 4B will be made in relation to the elements of Figures 1, 2 and 3. Figures 4A and 4B show exemplary processing pipeline 400 that demonstrates exemplary operations 402–424 for performing hypergraph-based collaborative filtering recommendations. Exemplary operations 402–424 can be performed by a computer system such as, for example, the electronic device 102 in Figure 1 or the circuit 202 in Figure 2. Figures 4A and 4B further include a collaborative filtering graph 118, a GNN model 114, a first user embedding set 406A, a first item embedding set 406B, a second user embedding set 410A, a second item embedding set 410B, a third user embedding set 414A, and a third item embedding set 414B.

[0055] In 402, a collaborative filtering graph receiving operation can be performed. Circuit 202 can be configured to receive a collaborative filtering graph 118 corresponding to a user set and an item set associated with the user set. Here, the item set may include different multimedia content such as sitcoms, news reports, and digital games that may be associated with the user set. The item set may also include various items such as clothing, electronic devices, game devices, and books that can be sold on an e-commerce application or website. It can be understood that different types of interactions may exist between the user set (e.g., the first user 302A, the second user 302B, and the third user 302C in Figure 3) and the item set (e.g., the first item 304A, the second item 304B, and the third item 304C in Figure 3). For example, different types of interactions may include selecting an item, adding an item to a digital cart, adding an item to a wishlist on an e-commerce app, watching a video, bookmarking a video, or "liking" a video, or rating a video on an OTT platform. If there is only one type of interaction between the user set and the item set, the interaction between the user set and the item set can be represented as a graph called a bipartite graph. On the other hand, if the interactions between the user set and the item set are of different types, the graph thus formed can be heterogeneous in nature and can form a multiplex bipartite graph. The collaborative filtering graph 118 can be a bipartite graph or a multiplex bipartite graph formed based on the interaction between the user set and the item set. Further details regarding collaborative filtering graphs are shown, for example, in Figure 3.

[0056] In 404, the operation of applying the GNN model 114 to the received collaborative filtering graph 118 can be performed. Circuit 202 can be configured to apply the GNN model 114 to the received collaborative filtering graph 118. The GNN model 114 can process the received collaborative filtering graph 118 to derive information related to each user and each item. In one embodiment, the GNN model 114 may be a graph convolutional network (GCN) model.

[0057] In 406, the determination operation of the first user embedding set 406A and the first item embedding set 406B can be performed. Circuit 202 can be configured to determine the first user embedding set 406A and the first item embedding set 406B. Here, the first user embedding set 406A and the first item embedding set 406B can be determined based on the application of the GNN model 114. Embeddings can correspond to vector representations of features associated with entities. For example, each of the first user embedding set 406A can correspond to features associated with an item subset from an item set that the corresponding user is thought to have viewed or selected. Each item embedding in the first item embedding set 406B can correspond to features associated with a user subset from a user set that is thought to have viewed or selected the user subset.

[0058] Referring to Figure 3, the third user 302C may have rated the first item 304A, the second item 304B, and the third item 304C as "5", "5", and "5", respectively. Therefore, the user embedding of the third user 302C can include the identification numbers of the items that the third user 302C was able to rate as "5". That is, the user embedding of the third user 302C can include the identification numbers of the first item 304A, the second item 304B, and the third item 304C. Furthermore, the user embedding of the third user 302C can include identification numbers such as item type, genre, video length, and language related to the first item 304A, the second item 304B, and the third item 304C. Similarly, for each rating provided by the first user 302A and the second user 302B, user embeddings related to the first user 302A and the second user 302B can also be determined. Furthermore, referring to Figure 3, it can be seen that the third item 304C is rated as "5" only by the third user 302C. Therefore, the item embedding for the third item 304C can include information such as the name, identification, and geographical location of the third user 302C. Similarly, for each rating provided by the first user 302A, the second user 302B, and the third user 302C, the item embeddings associated with the first item 304A and the second item 304B can be determined. In this way, the first user embedding set 406A and the first item embedding set 406B can be determined.

[0059] Referring again to Figure 4, at 408, the semantic clustering model application operation can be performed. Circuit 202 can be configured to apply the semantic clustering model 110 to each of the determined first user embedding set 406A and the determined first item embedding set 406B.

[0060] In one embodiment, the semantic clustering model 110 may correspond to a spectral clustering model configured for dimensionality reduction of the first user embedding set 406A and the first item embedding set 406B, respectively. A spectral clustering model can be understood as a clustering mechanism that can perform dimensionality reduction of an input dataset by leveraging spectra, such as eigenvalues ​​of the similarity matrix of the input dataset, before clustering the input dataset to a lower dimension. The input dataset of this disclosure may include the first user embedding set 406A and the first item embedding set 406B, respectively.

[0061] Spectral clustering algorithms associated with spectral clustering models can project an input dataset onto an "Rn" matrix, which may be necessary for clustering into "k" clusters. To construct an affinity matrix based on the projected input dataset, a Gaussian kernel matrix "K" or an adjacency matrix "A" can be created. It can be understood that spectral clustering algorithms can use a Gaussian kernel function to measure similarity. The adjacency matrix "A" can be a representation of the projected input dataset, where a set of rows in the adjacency matrix "A" represents a first user set, and a set of columns in the adjacency matrix "A" represents a first item set. Each entry in the adjacency matrix "A" can provide information about the interaction between a user and an item. In one example, the entry in the first row and first column of the adjacency matrix "A" may be "1". Thus, the first user associated with the first row may have viewed or selected the first item associated with the first column of the adjacency matrix "A". Furthermore, in some examples, the entries in the first row and second column of the adjacency matrix "A" can be "0". Thus, the first user associated with the first row may not have viewed or selected the second item associated with the second column of the adjacency matrix "A". An affinity matrix can be constructed based on the created Gaussian kernel matrix "K" or adjacency matrix "A". The affinity matrix can also be called the similarity matrix and provides information related to how similar pairs of entities are to each other. If the entry associated with a pair of entities in the affinity matrix is ​​"0", the corresponding pair of entities may not be similar. If the entry associated with a pair of entities is "1", the corresponding pair of entities may be similar. In other words, each entry in the affinity matrix can correspond to the edge weight associated with a pair of entities. Based on the constructed affinity matrix, a graph Laplacian matrix "L" can be created.The graph Laplacian matrix "L" can be understood as being obtained based on the difference between the adjacency matrix "A" and the degree matrix. Once the graph Laplacian matrix "L" is determined, the eigenvalue problem can be corrected. The advantage of using the graph Laplacian matrix "L" is that we can determine how well clusters are connected to each other based on the smallest eigenvalue of the graph Laplacian matrix "L". This is particularly useful because a low value can mean that the clusters are weakly connected, and different clusters can have weak connections. A k-dimensional subspace can be established based on the selection of k eigenvectors that can correspond to k minimum (or maximum) eigenvalues. Then, clusters can be created within the k-dimensional subspace using a k-means clustering algorithm. Further details on spectral clustering are shown, for example, in Figure 6.

[0062] At 410, a second embedding determination operation can be performed. Circuit 202 can be configured to determine a second user embedding set 410A and a second item embedding set 410B based on the application of the semantic clustering model 110. A cluster set can be determined based on the application of the semantic clustering model 110. The second user embedding set 410A and the second item embedding set 410B can be extracted from this cluster set. The determination of the second user embedding set and the second item embedding set is further explained, for example, in Figure 6.

[0063] In 412, a hypergraph construction operation can be performed. Circuit 202 can be configured to construct a hypergraph from the received collaborative filtering graph 118. The hypergraph can be a graph that can represent higher-order relationships between user sets and item sets related to the collaborative filtering graph 118 by using hyperedges. The regular edges of the graph can indicate interactions between pairs of nodes, and thus it can be understood that information between one node type and the latent representation of a node type having other node types can be ignored. In one example, the received collaborative filtering graph 118 may indicate that user "A" is thought to like movie "X". Such information can be captured in the embedding space using, for example, a first user embedding set 406A and a first item embedding set 406B. However, due to the nature of the received collaborative filtering graph 118 and the scarcity of information, the embedding space may not contain information related to other items that user "A" is thought not to have interacted with. For example, user "A" is thought to have interacted with movie "X" and not with other movies. Therefore, hypergraphs can use a special type of edge called a hyperedge, which can connect multiple nodes within an "n-dimensional" space. Further details about hypergraphs are shown, for example, in Figure 5.

[0064] In 414, a third embedding determination operation can be performed. Circuit 202 can be configured to determine a third user embedding set 414A and a third item embedding set 414B based on the constructed hypergraph. Further details regarding the determination of the third user embedding set 414A and the third item embedding set 414B are shown, for example, in Figure 5.

[0065] At 416, the first counter loss determination operation can be performed. Circuit 202 can be configured to determine the first counter loss based on the determined second user embedding set 410A and the determined third user embedding set 414A. The first counter loss can be a variation of NNCLR, where the nearest neighbor operator can be replaced with a cluster of similar nodes of each type, and similar user hypergraph embeddings can be used instead of an augmented view. NNCLR can be obtained according to the following equation (1): TIFF0007888107000001.tif17150(1) Here, "τ" can be the SoftMax temperature. "X ui " can be a third user embed related to user "i", "Z ui*,j " is the most similar user "i" obtained from cluster "j". * This can be the second embedding of ".

[0066] At 418, a second control loss determination operation can be performed. Circuit 202 can be configured to determine the second control loss based on the determined second item embedding set 410B and the determined third item embedding set 414B. The second control loss can be similar to the first control loss and can be determined according to the following equation (2): TIFF0007888107000002.tif17150(2) Here, "τ" can be the SoftMax temperature. "X ui " can be a third item embed related to item "i", "Z vi*,j " is the most similar item "i" obtained from cluster "j". * This can be the second embedding of ".

[0067] At 420, a collaborative filtering score determination operation can be performed. Circuit 202 can be configured to determine a collaborative filtering score based on a determined first control loss and a determined second control loss. The collaborative filtering score can provide a score set for each user's item set in the user set. The score set can follow the user set's preferences, past interactions, and selections.

[0068] In one embodiment, the circuit 202 may be further configured to determine a fifth user embedding set based on a first control loss and a third user embedding set. The circuit 202 may be further configured to determine a fifth item embedding set based on a second control loss and a third item embedding set. The fifth user embedding set may provide a vector representation of features associated with the user set. The fifth item embedding set may provide a vector representation of features associated with the item set.

[0069] In one embodiment, the circuit 202 may be further configured to determine the final user embeddings based on a determined fifth set of user embeddings. The circuit 202 may be further configured to determine the final item embeddings based on a determined fifth set of item embeddings. Here, the determination of the collaborative filtering score may be based on the determined final user embeddings and the determined final item embeddings.

[0070] In one example, a user set can interact with an item set by bookmarking items, partially viewing items, and fully viewing items. Thus, a determined fifth user embedding set may include a fifth user embedding related to bookmarking a subset of items from each user's item set, a fifth user embedding related to partial viewing of a subset of items from the item set, and a fifth user embedding related to full viewing of a subset of items from the item set. Similarly, a determined fifth item embedding set may include, per item, a fifth item embedding related to bookmarking the corresponding item by a user subset from the user set, a fifth user embedding related to partial viewing of the corresponding item by a user subset from the user set, and a fifth user embedding related to full viewing of the corresponding item by a user subset from the user set. The final user embedding for a user, such as user 120, can be determined based on a combination of the corresponding user's determined fifth user embeddings. That is, the final user embedding for a corresponding user can be determined by combining the fifth user embedding related to bookmarking, the fifth user embedding related to partial viewing, and the fifth user embedding related to full viewing for a user, such as user 120. Similarly, the final item embedding of an item can be determined based on a combination of the determined fifth item embeddings for the corresponding item. That is, the final item embedding can be determined by combining the fifth item embedding related to bookmarking the corresponding item, the fifth item embedding related to partial viewing, and the fifth item embedding related to complete viewing. In one embodiment, the final user embedding and the final item embedding can be applied to a graph neural network (GNN) model or a natural language processing (NLP) model to generate recommendation probabilities for the item set.

[0071] In one embodiment, each of the determined final user embeddings and determined final item embeddings can correspond to at least one linkage of a collaborative view, a hypergraph view, or a semantic view. The collaborative view of each of the determined final user embeddings and determined final item embeddings can be associated with the first user embedding set 406A and the first item embedding set 406B, respectively. The hypergraph view can also be called a higher-order view. The hypergraph view of each of the determined final user embeddings and determined final item embeddings can be associated with the second user embedding set 410A and the second item embedding set 410B, respectively. The semantic view of each of the determined final user embeddings and determined final item embeddings can be associated with the third user embedding set 414A and the third item embedding set 414B, respectively. The determined final user embeddings can be associated with the first user embedding set 406A, the second user embedding set 410A, and the third user embedding set 414A. Similarly, the determined final item embeddings can be associated with the first item embedding set 406B, the second item embedding set 410B, and the third item embedding set 414B. Thus, each of the determined final user embeddings and the determined final item embeddings can correspond to at least one linkage of a collaborative view, a hypergraph view, or a semantic view.

[0072] In 422, a recommendation decision operation can be performed. Circuit 202 can be configured to determine an item recommendation for user 120 based on the determined collaborative filtering score. In one embodiment, the collaborative filtering score may provide a score set for each user's item set in the user set. For each user, the item that may be associated with the highest score can be selected as a recommendation. In one example, the user set includes users "A", "B", and "C", and the item set includes items "X", "Y", and "Z". For user "A", the score set may include "0.1", "0.5", and "0.7" associated with items "X", "Y", and "Z", respectively. In such an example, since item "Z" has the highest score for user "A", item "Z" can be determined as a recommendation for user "A".

[0073] In 424, the rendering operation of the recommended item can be performed. Circuit 202 can be configured to render the determined recommended item on the display device 210. In one example, the determined recommended item may be the movie "X". The recommended movie "X" can be displayed on the display device 210 to notify the user 120 associated with the electronic device 102. Subsequently, the movie "X" can be played based on user input related to the selection of movie "X" from the user 120.

[0074] Figure 5 shows an exemplary scenario of an architecture for hypergraph embedding according to embodiments of the present disclosure. The description of Figure 5 is made in relation to the elements of Figures 1, 2, 3, 4A, and 4B. Figure 5 shows an exemplary scenario 500. Scenario 500 may include a hypergraph 502, a fourth user embedding 504A, a fourth user embedding 504B, a first hypergraph convolutional network (HGCN) model 506A, a first HGCN model 506B, a third user embedding 508A, a third user embedding 508B, a fourth item embedding 510A, a fourth item embedding 510B, a second HGCN model 512A, a second HGCN model 512B, a third item embedding 514A, and a third item embedding 514B. A series of operations related to scenario 500 are described here.

[0075] In scenario 500, a hypergraph 502 can be constructed based on the received collaborative filtering graph (for example, collaborative filtering graph 118 in Figure 4A). In one embodiment, the constructed hypergraph 502 may correspond to a multipartite graph having homogenous edges. The constructed hypergraph 502 can be a multipartite graph because it can represent multiple types of interactions between the user set and the item set. Furthermore, the constructed hypergraph 502 may be formed such that each hyperedge represents one type of interaction.

[0076] In one embodiment, the first edge type in the hypergraph 502 can correspond to an interaction between a first user and a first subset of items associated with the first user. The second edge type in the hypergraph can correspond to an interaction between a second subset of users and a second item associated with each of the second subset of users. For example, the first hyper-edge type can be formed to indicate a subset of items that the first user is considered to have evaluated as "1". Another first hyper-edge type can be formed to indicate a subset of items that the first user is considered to have evaluated as "2". The second hyper-edge can be formed to indicate a subset of users that the first item is considered to have been evaluated as "1" by. Another second hyper-edge can be formed to indicate a subset of users that the first item is considered to have been evaluated as "2" by.

[0077] Note that the homogeneous hypergraph constructed based on the first hyper-edge type can be defined according to the following equation (3), TIFF0007888107000003.tif11157 and TIFF0007888107000004.tif6150(3) Here, "G U,base " can be a homogeneous graph, "U" can be a set of users, and "E U,i " can be a set of the first hyper-edge types.

[0078] Note that the homogeneous hypergraph constructed based on the second hyper-edge type can be defined according to the following equation (4), TIFF0007888107000005.tif16157 and TIFF0007888107000006.tif6150(4) Here, "G U,base " can be a homogeneous graph, "I" can be a set of items, and "EI,j This can be the second hyper-edge type set.

[0079] Furthermore, the hypergraph 502 can use an incidence matrix "H" for the user set "U". The incidence matrix for user set "U" can be determined according to the following equation (5): TIFF0007888107000007.tif9150 and TIFF0007888107000008.tif6150(5) Here, "E U,i " can be the first hyperedge typeset, and "i" can represent the constructed hypergraph. Similarly, the connection matrix of itemset "I" is "H I,j It can be defined as "(i,e)".

[0080] In one embodiment, the circuit 202 can be further configured to determine user-item correlation sets, item-user correlation sets, and user-user correlation sets based on the constructed hypergraph 502. User-item correlation sets can be determined based on a first edge type and can represent relationships between users and items. For example, a first user-item correlation can provide information related to a set of items that a first user is thought to have fully viewed. A second user-item correlation can provide information related to a set of items that a first user is thought to have selected as a base. Item-user correlation sets can be determined based on a second edge type and can provide information related to relationships between items and users. For example, a first item-user correlation can represent a set of users that are thought to have fully viewed a first item. A second item-user correlation can provide information related to a set of users that are thought to have selected a first item as a base. User-user correlation sets can be determined based on the first and second edge types and can provide information related to potential relationships between users. For example, a first user may have fully viewed the movie "X". Similarly, the second user can also watch the movie "X" in its entirety. Here, a relationship can exist between the first user and the second user. To capture the aforementioned relationship, a user-user correlation can be determined.

[0081] The circuit 202 can be further configured to determine a fourth user embedding set (e.g., a fourth user embedding 504A) based on the determined user-item correlation set and user-user correlation set. The fourth user embedding set may include one or more user embeddings for each user. Each of the one or more user embeddings associated with a user may correspond to one interaction type. For example, referring to Figure 5, a first interaction type may relate to fully viewing one or more items, and a second interaction type may relate to partially viewing one or more items. The fourth user embedding 504A can be formed based on user-item correlations and user-user correlation sets corresponding to a first interaction type associated with a first user. The fourth user embedding 504B can be formed based on user-item correlations and user-user correlation sets corresponding to a second interaction type associated with a first user.

[0082] Once a fourth user embedding set is determined, the circuit 202 can be further configured to apply the first HGCN model set (e.g., the first HGCN model set 116A in Figure 1) to the determined fourth user embedding set (e.g., the fourth user embedding 504A). An HGCN model from the first HGCN model set can be applied to each of the fourth user embedding sets. The first HGCN model set (e.g., the first HGCN model set 116A in Figure 1) can be an ML model that processes information related to the hypergraph 502 and makes inference decisions based on the processing.

[0083] The convolution operator associated with the first HGCN model set for the constructed hypergraph 502 (e.g., the first HGCN model set 116A in Figure 1) can be determined according to the following equation (6): TIFF0007888107000009.tif6150(6) Here, "σ" can be a non-linear activation function, "X" can be a feature matrix, and "P" can be a learnable weight matrix. Furthermore, "HWH T " can be used to measure the pair relationship between nodes within the same homogeneous hypergraph, and "W" can be a weight matrix on which weights can be assigned to all hyperedges.

[0084] Normalized versions of the symmetric and asymmetric convolution operators can be defined according to the following equations (7) and (8). TIFF0007888107000010.tif9150(7) and TIFF0007888107000011.tif6150(8) Referring to equation (7), "I" can be the identity matrix and "D" can be the node degree matrix of a simple graph. Referring to equation (8), "σ" can be the nonlinear activation function. TIFF0007888107000012.tif6150 can be a feature of layer "l" and "W U ∈ "R|V|×|V|" can be the identity matrix, and "P" can represent a learnable filter matrix, and "D l " and "D l+1 These can be the dimensions of sheave "l" and sheave "l+1", respectively.

[0085] For example, referring to Figure 5, the first HGCN model 506A can be applied to the fourth user embedding 504A, and the first HGCN model 506B can be applied to the fourth user embedding 504B. The third user embedding set can be determined based on the application of the first HGCN model set (e.g., the first HGCN model set 116A in Figure 1). For example, referring to Figure 5, the third user embedding 508A for the first user can be determined based on the application of the first HGCN model 506A. The third user embedding 508B for the first user can be determined based on the application of the first HGCN model 506B. Similarly, the fourth user embedding for each user in the user set can be determined for each interaction type.

[0086] The circuit 202 can be further configured to determine a fourth item embedding set (e.g., a fourth item embedding 510A) based on the determined item-user correlation set. The fourth item embedding set may include one or more item embeddings for each item. Each of the one or more item embeddings associated with an item may correspond to one interaction type. For example, referring to Figure 5, the fourth item embedding 510A can be formed based on an item-user correlation corresponding to a first interaction type associated with the first item. The fourth item embedding 510B can be formed based on an item-user correlation corresponding to a second interaction type associated with the first item.

[0087] Once the fourth item embedding set is determined, the circuit 202 can be further configured to apply the second HGCN model set (e.g., the second HGCN model set 116B) to the determined fourth item embedding set. An HGCN model can be applied to each fourth item embedding. The second HGCN model set (e.g., the second HGCN model set 116B in Figure 1) can be an ML model that processes information related to the hypergraph 502 and determines inferences based on the processing. For example, referring to Figure 5, the second HGCN model 512A can be applied to the fourth item embedding 510A, and the second HGCN model 512B can be applied to the fourth item embedding 510B. A third item embedding set can be determined based on the application of the second HGCN model set (e.g., the second HGCN model set 116B in Figure 1). For example, referring to Figure 5, based on the application of the second HGCN model 512A, a third item embedding 514A of the first item related to fully viewing the first item can be determined. Based on the application of the second HGCN model 512B, a third item embedding 514B of the first item related to selecting the first item as a base can be determined. Similarly, a fourth item embedding of each item in the itemset can be determined for each interaction type.

[0088] Furthermore, Scenario 500 in Figure 5 is for illustrative purposes only and should not be interpreted as limiting the scope of this disclosure.

[0089] Figure 6 shows an exemplary scenario of comparative learning according to embodiments of the present disclosure. The description of Figure 6 is made in relation to the elements of Figures 1, 2, 3, 4A, 4B and 5. Figure 6 shows an exemplary scenario 600. Scenario 600 may include a collaborative filtering graph 602, a graph convolutional network (GCN) 604, a semantic cluster 606 of user nodes and item nodes, a second user embedding 608, a hypergraph embedding block 610, and a third user embedding 612. A series of operations related to scenario 600 are described here. Figure 6 illustrates comparative learning for user embeddings. However, scenario 600 in Figure 6 can also be applied to comparative learning for item embeddings without departing from the scope of the present disclosure.

[0090] The self-monitoring approach commonly used in the field of computer vision can be understood as involving a process to determine the most discriminative representation of embeddings. In one example, the discriminative representation of embeddings can be determined by augmentation of the same object in an image for a given set of different views. In another example, the discriminative representation of embeddings can be obtained by using similar objects and comparing an object with other dissimilar objects. The contrast learning approach described above can be extended to recommendation systems. Here, different augmentations of user-item interactions can be used. Different augmentations can be obtained based on node deletion, edge deletion, and node duplication, etc. Views of augmented node embeddings in a minibatch of interactions can form positive pairs, and the remaining embeddings from the minibatch can form negative pairs.

[0091] For example, referring to Figure 6, GCN604 can be applied to the collaborative filtering graph 602. GCN604 can be a generalized convolutional neural network that can employ a semi-supervised learning approach to the graph. Based on the application of GCN604 to the collaborative filtering graph 602, a first user embedding set and a first item embedding set can be obtained. Furthermore, based on the application of GCN604 to the collaborative filtering graph 602, semantic clusters of user nodes and item nodes 606 can be obtained. Subsequently, a second user embedding 608 can be obtained based on the semantic clusters of user nodes and item nodes 606. The second user embedding 608 can be associated with similar users determined from the semantic clusters of user nodes and item nodes 606. Furthermore, the collaborative filtering graph 602 can be applied to the hypergraph embedding block 610. The hypergraph embedding block 610 may include a first set of HGCN models (e.g., the first HGCN model 506A and the first HGCN model 506B in Figure 5) and a second set of HGCN models (e.g., the second HGCN model 512A and the second HGCN model 512B in Figure 5). A third user embedding 612 can be obtained based on the application of the collaborative filtering graph 602 to the hypergraph embedding block 610. The second user embedding 608 and the third user embedding 612 may be a positive pair of embeddings that can be used for controlled learning purposes. Positive pairs of embeddings can be used for controlled learning purposes. Furthermore, negative samples may be samples that are not considered to be part of the cluster to which user "U1" belongs.

[0092] Furthermore, Scenario 600 in Figure 6 is for illustrative purposes only and should not be interpreted as limiting the scope of this disclosure.

[0093] Figure 7 shows an exemplary scenario for recommending an item set to a user set according to an embodiment of the present disclosure. The description of Figure 7 will be made in relation to the elements of Figures 1, 2, 3, 4A, 4B, 5, and 6. Figure 7 shows an exemplary scenario 700. Scenario 700 may include a hyperedge 702, a first user 704A, a second user 704B, a third user 704C, a first news channel 706, a final user embedding 708, a final item embedding 710, and a recommended item set 712. The recommended item set 712 may include a second news channel 712A, a third news channel 712B, and a fourth news channel 712C. A series of operations related to scenario 700 will be described here.

[0094] In scenario 700 of Figure 7, the second user 704B and the third user 704C may have an active interest in the first news channel 706. For example, the second user 704B and the third user 704C may have watched the first news channel 706. The first user 704A may not have watched the first news channel 706. On the other hand, the first user 704A and the third user 704C may have also watched news channels (not shown) similar to the first news channel 706. Therefore, a potential relationship may exist between the first user 704A and the first news channel 706. Furthermore, a potential relationship may also exist between the first user 704A and the second user 704B. Therefore, the first user 704A, the second user 704B, and the third user 704C may form a hyperedge such as the hyperedge 702 together with the first news channel 706. A hypergraph can be constructed by forming multiple hyperedges similar to hyperedge 702. Based on the constructed hypergraph, a third user embedding set can be determined. The third user embedding set (not shown) may include a third user embedding associated with the first user 704A, a third user embedding associated with the second user 704B, and a third user embedding associated with the third user 704C. The third user embedding associated with the first user 704A, the third user embedding associated with the second user 704B, and the third user embedding associated with the third user 704C may be similar to each other. Based on the determined third user embedding set, the final user embedding 708 and the final item embedding 710 can be obtained. For example, as shown in Figure 7, the final user embedding 708 may be "0.87", "0.79", and "0.77" for the first user 704A, the second user 704B, and the third user 704C, respectively. The final user embedding 708 can correspond to the collaborative filtering scores associated with users 704A, 704B, and 704C.Furthermore, the final item embedding 710 can be "0.95" for the first item, "0.89" for the second item, and "0.87" for the third item, and these can be recommended to the first user 704A, the second user 704B, and the third user 704C, respectively. The final item embedding 710 can correspond to the collaborative filtering scores associated with the first, second, and third items. For example, based on the final user embedding 708 and the final item embedding 710, the second news channel 712A can be recommended to the first user 704A, the third news channel 712B to the second user 704B, and the fourth news channel 712C to the third user 704C. Note that the second news channel 712A, the third news channel 712B, and the fourth news channel 712C can be similar to each other.

[0095] Furthermore, Scenario 700 in Figure 7 is for illustrative purposes only and should not be interpreted as limiting the scope of this disclosure.

[0096] Figure 8 is a flowchart illustrating the operation of an exemplary method for recommending hypergraph-based collaborative filtering according to embodiments of the present disclosure. The description of Figure 8 is made in relation to the elements of Figures 1, 2, 3, 4A, 4B, 5, 6, and 7. Figure 8 shows flowchart 800. Flowchart 800 may include operations 802-824 and can be implemented by the electronic device 102 in Figure 1 or the circuit 202 in Figure 2. Flowchart 800 can start from 802 and proceed to 804.

[0097] In 804, a collaborative filtering graph 118 corresponding to the user set and the item set associated with the user set can be received. Circuit 202 can be configured to receive a collaborative filtering graph 118 corresponding to the user set and the item set associated with the user set. Further details regarding the collaborative filtering graph 118 are explained in Figure 3, for example.

[0098] In 806, a first user embedding set 406A and a first item embedding set 406B can be determined based on the received collaborative filtering graph 118. Circuit 202 can be configured to determine the first user embedding set 406A and the first item embedding set 406B based on the received collaborative filtering graph 118. Further details regarding the first user embedding set 406A and the first item embedding set 406B are described in Figure 4A, for example.

[0099] In 808, the semantic clustering model 110 can be applied to each of the determined first user embedding set 406A and the determined first item embedding set 406B. Circuit 202 can be configured to apply the semantic clustering model 110 to each of the determined first user embedding set 406A and the determined first item embedding set 406B. Further details regarding the application of the semantic clustering model 110 are described in Figure 4A, for example.

[0100] In 810, a second user embedding set 410A and a second item embedding set 410B can be determined based on the application of the semantic clustering model 110. Circuit 202 can be configured to determine the second user embedding set 410A and the second item embedding set 410B based on the application of the semantic clustering model 110. Further details regarding the second user embedding set 410A and the second item embedding set 410B are described in Figure 4A, for example.

[0101] In 812, a hypergraph (such as the hypergraph 502 in Figure 5) can be constructed from the received co-filtering graph 118. Circuit 202 can be configured to construct a hypergraph (such as the hypergraph 502 in Figure 5) from the received co-filtering graph 118. Further details regarding hypergraph 502 are explained in Figure 5, for example.

[0102] In 814, a third user embedding set 414A and a third item embedding set 414B can be determined based on the constructed hypergraph. Circuit 202 can be configured to determine the third user embedding set 414A and the third item embedding set 414B based on the constructed hypergraph. Further details regarding the third user embedding set 414A and the third item embedding set 414B are described in Figure 4B, for example.

[0103] In 816, the first reference loss can be determined based on the determined second user embedding set 410A and the determined third user embedding set 414A. Circuit 202 can be configured to determine the first reference loss based on the determined second user embedding set 410A and the determined third user embedding set 414A. Further details regarding the first reference loss are described in Figure 4B, for example.

[0104] In 818, a second control loss can be determined based on the determined second item embedding set 410B and the determined third item embedding set 414B. Circuit 202 can be configured to determine the second control loss based on the determined second item embedding set 410B and the determined third item embedding set 414B. Further details regarding the second control loss are explained in Figure 4B, for example.

[0105] In 820, the cooperative filtering score can be determined based on at least the determined first control loss and the determined second control loss. Circuit 202 can be configured to determine the cooperative filtering score based on at least the determined first control loss and the determined second control loss. Further details regarding the cooperative filtering score are explained in Figure 4B, for example.

[0106] In 822, item recommendations for user 120 can be determined based on the determined collaborative filtering score. Circuit 202 can be configured to determine item recommendations for user 120 based on the determined collaborative filtering score. Further details regarding item recommendations are explained, for example, in Figure 4B.

[0107] In step 824, the determined recommended items can be rendered on the display device 210. Circuit 202 can be configured to render the determined recommended items on the display device 210. Further details regarding the rendering of the determined recommended items are explained in Figure 4B, for example. Control can then proceed to termination.

[0108] While flowchart 800 is shown as discrete operations such as 804, 806, 808, 810, 812, 814, 816, 818, 820, 822, and 824, the disclosure is not limited in this way. Accordingly, in some embodiments, such discrete operations can be further divided into further operations, combined into fewer operations, or deleted, depending on the implementation, without impairing the essence of the disclosed embodiments.

[0109] Various embodiments of this disclosure can provide a non-temporary computer-readable medium and / or storage medium storing computer-executable instructions that can be executed by a machine and / or computer to operate an electronic device (e.g., electronic device 102 in Figure 1). Such instructions can cause the electronic device 102 to perform an operation which may include receiving a collaborative filtering graph (e.g., collaborative filtering graph 118) corresponding to a user set and an item set associated with the user set. The operation may further include determining a first user embedding set (e.g., first user embedding set 406A) and a first item embedding set (e.g., first item embedding set 406B) based on the received collaborative filtering graph 118. The operation may further include applying a semantic clustering model (e.g., semantic clustering model 110) to each of the determined first user embedding set 406A and the determined first item embedding set 406B. The operation may further include determining a second user embedding set (e.g., second user embedding set 410A) and a second item embedding set (e.g., second item embedding set 410B) based on the application of the semantic clustering model 110. The operation may further include constructing a hypergraph (such as hypergraph 502 in Figure 5) from the received collaborative filtering graph 118. The operation may further include determining a third user embedding set (e.g., third user embedding set 414A) and a third item embedding set (e.g., third item embedding set 414B) based on the constructed hypergraph 502. The operation may further include determining a first control loss based on the determined second user embedding set 410A and the determined third user embedding set 414A. The operation may further include determining a second control loss based on the determined second item embedding set 410B and the determined third item embedding set 414B.The operation may further include determining a collaborative filtering score based on a determined first control loss and a determined second control loss. The operation may further include determining recommendations for items for a user, such as user 120, based on the determined collaborative filtering score. The operation may further include rendering the determined recommended items on a display device (such as display device 210).

[0110] Exemplary embodiments of this disclosure can provide an electronic device (such as the electronic device 102 in Figure 1) that includes a circuit (such as circuit 202). Circuit 202 can be configured to receive a collaborative filtering graph 118 corresponding to a user set and an item set associated with the user set. Circuit 202 can be configured to determine a first user embedding set 406A and a first item embedding set 406B based on the received collaborative filtering graph 118. Circuit 202 can be configured to apply a semantic clustering model 110 to each of the determined first user embedding set 406A and the determined first item embedding set 406B. Circuit 202 can be configured to determine a second user embedding set 410A and a second item embedding set 410B based on the application of the semantic clustering model 110. Circuit 202 can be configured to construct a hypergraph (such as the hypergraph 502 in Figure 5) from the received collaborative filtering graph 118. Circuit 202 can be configured to determine a third user embedding set 414A and a third item embedding set 414B based on the constructed hypergraph. Circuit 202 can be configured to determine a first control loss based on the determined second user embedding set 410A and the determined third user embedding set 414A. Circuit 202 can be configured to determine a second control loss based on the determined second item embedding set 410B and the determined third item embedding set 414B. Circuit 202 can be configured to determine a collaborative filtering score based on the determined first control loss and the determined second control loss. Circuit 202 can be configured to determine item recommendations for user 120 based on the determined collaborative filtering score. Circuit 202 can be configured to render the determined recommended items on the display device 210.

[0111] In one embodiment, the circuit 202 can be further configured to apply a GNN model (e.g., GNN model 114) to the received collaborative filtering graph 118, and based on the application of the GNN model 114, each of the first user embedding set 406A and the first item embedding set 406B can be further determined.

[0112] In one embodiment, the circuit 202 can be further configured to determine user-item correlation sets, item-user correlation sets, and user-user correlation sets based on the constructed hypergraph 502. The circuit 202 can be further configured to determine a fourth user embedding set based on the determined user-item correlation sets and user-user correlation sets. The circuit 202 can be further configured to apply a first HGCN model set (e.g., a first HGCN model set 116A) to the determined fourth user embedding set. The circuit 202 can be further configured to determine a third user embedding set 414A based on the application of the first HGCN model set 116A. The circuit 202 can be further configured to determine a fourth item embedding set based on the determined item-user correlation sets. The circuit 202 can be further configured to apply a second HGCN model set (e.g., a second HGCN model set 116B) to the determined fourth item embedding set. Circuit 202 can be further configured to determine a third item embedding set 414B based on the application of a second HGCN model set 116B.

[0113] In one embodiment, the semantic clustering model 110 can correspond to a spectral clustering model configured for dimensionality reduction.

[0114] In one embodiment, the circuit 202 may be further configured to determine a fifth user embedding set based on a first control loss and a third user embedding set 414A. The circuit 202 may be further configured to determine a fifth item embedding set based on a second control loss and a third item embedding set 414B.

[0115] In one embodiment, the circuit 202 may be further configured to determine the final user embeddings based on a determined fifth set of user embeddings. The circuit 202 may be further configured to determine the final item embeddings based on a determined fifth set of item embeddings, and the determination of the collaborative filtering score may further be based on the determined final user embeddings and the determined final item embeddings.

[0116] In one embodiment, each of the determined final user embeddings and the determined final item embeddings can correspond to at least one concatenation of a collaborative view, a hypergraph view, or a semantic view.

[0117] In one embodiment, the constructed hypergraph 502 can correspond to a multi-bipartite graph having homogeneous edges.

[0118] In one embodiment, a first edge type in the hypergraph 502 may correspond to an interaction between a first user and a first item subset associated with the first user, and a second edge type in the hypergraph 502 may correspond to an interaction between a second user subset and a second item associated with each of the second user subsets.

[0119] This disclosure includes all features that enable the implementation of the methods described herein and may also be located within a computer program product that, when loaded into a computer system, can perform these methods. In this context, a computer program means any expression in any language, code, or notation of an instruction set intended to be executed directly, or after either a) conversion into another language, code, or notation, or b) reproduction in a different form of content, on a system having information processing capabilities.

[0120] While this disclosure has been described with reference to several embodiments, those skilled in the art will understand that various modifications can be made and equivalents can be substituted without departing from the scope of this disclosure. Furthermore, many modifications can be made without departing from the scope of this disclosure to suit specific circumstances or content to the teachings of this disclosure. Accordingly, this disclosure is not limited to the specific embodiments disclosed, but is intended to include all embodiments that fall within the scope of the appended claims. [Explanation of symbols]

[0121] 100 Network Environment 102 Electronic equipment 104 Servers 106 Databases 108 Communication Networks 110 Semantic Clustering Models 112 Recommended Models 114 Graph Neural Network (GNN) Models 116A First Hypergraph Convolutional Network (HGCN) Model Set 116B Second HGCN Model Set 118 Collaborative Filtering Graph 120 users

Claims

1. An electronic device, The system receives a collaborative filtering graph corresponding to a user set and an item set associated with the user set. Based on the received collaborative filtering graph, the first user embedding set and the first item embedding set are determined. A semantic clustering model is applied to each of the first user embedding set and the first item embedding set determined above. Based on the application of the semantic clustering model described above, a second user embedding set and a second item embedding set are determined. A hypergraph is constructed from the aforementioned received collaborative filtering graph. Based on the constructed hypergraph, a third user embedding set and a third item embedding set are determined. Based on the determined second user embedding set and the determined third user embedding set, the first control loss is determined. A second control loss is determined based on the determined second item embedding set and the determined third item embedding set. Based on the first control loss and the second control loss determined above, the cooperative filtering score is determined. Based on the aforementioned determined collaborative filtering score, the recommendation of items for the user is determined. The determined recommended items are rendered on the display device. An electronic device characterized by having a circuit configured in such a way.

2. The aforementioned circuit is Further configured to apply a graph neural network model to the received collaborative filtering graph, Each of the first user embedding set and the first item embedding set is further determined based on the application of the graph neural network model. The electronic device according to claim 1.

3. The aforementioned circuit is Based on the constructed hypergraph, user-item correlation sets, item-user correlation sets, and user-user correlation sets are determined. A fourth user embedding set is determined based on the user-item correlation set and the user-user correlation set. The first hypergraph convolutional network (HGCN) model set is applied to the fourth user embedding set determined above. The third user embedding set is determined based on the application of the first HGCN model set, Based on the determined item-user correlation set, a fourth item embedding set is determined. The second HGCN model set is applied to the fourth item embedding set determined above. The third item embedding set is determined based on the application of the second HGCN model set. The electronic device according to claim 1, further configured as follows.

4. The aforementioned semantic clustering model corresponds to a spectral clustering model configured for dimensionality reduction. The electronic device according to claim 1.

5. The aforementioned circuit is A fifth user embedding set is determined based on the first control loss and the third user embedding set. A fifth item embedding set is determined based on the second control loss and the third item embedding set. The electronic device according to claim 1, further configured as follows.

6. The aforementioned circuit is Based on the fifth set of user embeddings determined above, the final user embeddings are determined. The final item embedding is determined based on the fifth set of item embeddings determined above. It is further configured in this way, The determination of the collaborative filtering score is further based on the determined final user embedding and the determined final item embedding. The electronic device according to claim 5.

7. The constructed hypergraph corresponds to a multi-bipartite graph having homogeneous edges. The electronic device according to claim 1.

8. The first edge type in the hypergraph corresponds to the interaction between a first user and a first item subset associated with the first user. The second edge type in the hypergraph corresponds to the interaction between a second user subset and a second item associated with each of the second user subsets. The electronic device according to claim 1.

9. In electronic devices, Receiving a collaborative filtering graph corresponding to a user set and an item set associated with the user set, Based on the received collaborative filtering graph, a first user embedding set and a first item embedding set are determined, Applying a semantic clustering model to each of the determined first user embedding set and the determined first item embedding set, Determining a second user embedding set and a second item embedding set based on the application of the semantic clustering model, Constructing a hypergraph from the aforementioned received collaborative filtering graph, Based on the constructed hypergraph, a third user embedding set and a third item embedding set are determined, The first control loss is determined based on the second user embedding set and the third user embedding set determined above, The second control loss is determined based on the determined second item embedding set and the determined third item embedding set, The cooperative filtering score is determined based on the first control loss and the second control loss determined above. Based on the aforementioned determined collaborative filtering score, the recommendation of items for the user is determined, The determined recommended items are rendered on the display device, A method characterized by including the following.

10. This further includes applying a graph neural network model to the received collaborative filtering graph, Each of the first user embedding set and the first item embedding set is further determined based on the application of the graph neural network model. The method according to claim 9.

11. Based on the constructed hypergraph, determine the user-item correlation set, the item-user correlation set, and the user-user correlation set. A fourth user embedding set is determined based on the user-item correlation set and the user-user correlation set. Applying the first hypergraph convolutional network (HGCN) model set to the fourth user embedding set determined above, Determining the third user embedding set based on the application of the first HGCN model set, Based on the determined item-user correlation set, a fourth item embedding set is determined, Applying the second HGCN model set to the fourth item embedding set determined above, Determining the third item embedding set based on the application of the second HGCN model set, The method according to claim 9, further comprising:

12. The aforementioned semantic clustering model corresponds to a spectral clustering model configured for dimensionality reduction. The method according to claim 9.

13. A fifth user embedding set is determined based on the first control loss and the third user embedding set. The fifth item embedding set is determined based on the second control loss and the third item embedding set, The method according to claim 9, further comprising:

14. The final user embedding is determined based on the fifth user embedding set determined above, The final item embedding is determined based on the fifth set of item embeddings determined above, It further includes, The determination of the collaborative filtering score is further based on the determined final user embedding and the determined final item embedding. The method according to claim 13.

15. The constructed hypergraph corresponds to a multi-bipartite graph having homogeneous edges. The method according to claim 9.

16. The first edge type in the hypergraph corresponds to the interaction between a first user and a first item subset associated with the first user. The second edge type in the hypergraph corresponds to the interaction between a second user subset and a second item associated with each of the second user subsets. The method according to claim 9.

17. A non-temporary computer-readable medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by an electronic device, Receiving a collaborative filtering graph corresponding to a user set and an item set associated with the user set, Based on the received collaborative filtering graph, a first user embedding set and a first item embedding set are determined, Applying a semantic clustering model to each of the determined first user embedding set and the determined first item embedding set, Determining a second user embedding set and a second item embedding set based on the application of the semantic clustering model, Constructing a hypergraph from the aforementioned received collaborative filtering graph, Based on the constructed hypergraph, a third user embedding set and a third item embedding set are determined, The first control loss is determined based on the second user embedding set and the third user embedding set determined above, The second control loss is determined based on the determined second item embedding set and the determined third item embedding set, The cooperative filtering score is determined based on the first control loss and the second control loss determined above. Based on the aforementioned determined collaborative filtering score, the recommendation of items for the user is determined, The determined recommended items are rendered on the display device, A non-temporary computer-readable medium characterized by causing the electronic device to perform an operation including the above.

18. The constructed hypergraph corresponds to a multi-bipartite graph having homogeneous edges. The non-temporary computer-readable medium according to claim 17.

Images