Method and apparatus for information recommendation / determination, and system for information recommendation

By analyzing the historical behavior of users clicking, matching aggregated items and items, and calculating combination weights, the problem of difficulty in exploring user interests in the existing technology is solved, personalized information recommendation is achieved, and user experience is improved.

WO2024239929A1PCT designated stage expired Publication Date: 2024-11-28BEIJING WODONG TIANJUN INFORMATION TECH CO LTD +1
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2024/090617
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-24
Filing Date
2024-04-29
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Existing information recommendation systems are difficult to accurately tap users' interest in aggregated items and target items in them, resulting in poor recommendation results.

Method used

By analyzing the historical behavior sequence of the item clicks of the target user, query the first correlation information between different items and the second correlation information between the items and the aggregated items, match the aggregated items and items, calculate the combination weight, and filter out the user's sense of the user Polymer items of interest and their target items combinations are recommended.

Benefits of technology

It realizes personalized information recommendation, improves user experience, accurately explores and locates aggregated items and target items that users are interested in, and improves the accuracy of recommendations and user satisfaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024090617_28112024_PF_FP_ABST
    Figure CN2024090617_28112024_PF_FP_ABST
Patent Text Reader

Abstract

The present disclosure provides a method and apparatus for information recommendation / determination, and a system for information recommendation, and relates to the technical field of computers. The method comprises: using a historical article clicking action sequence of a target user to query first correlation information between different articles and second correlation information between articles and aggregated articles to obtain an article set and an aggregated article set that are related to the historical articles on which the target user has clicked and may interest the target user, then determining target articles of the aggregated articles in the aggregated article set for the target user by matching the aggregated articles in the aggregated article set with the articles in the article set, and recommending to the target user a combination of the aggregated articles in the aggregated article set and the target articles thereof. Therefore, combinations of aggregated articles in which each user is interested and target articles are accurately mined, positioned, and recommended, and different combinations of aggregated articles and target articles are recommended to users with different interests to achieve personalized information recommendation.
Need to check novelty before this filing date? Find Prior Art

Description

Information recommendation / determination method and device and information recommendation system

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is based on the application with CN application number 202310595489.0 and application date May 24, 2023, and claims its priority. The disclosed content of the CN application is hereby introduced as a whole into this application. Technical Field

[0003] The present disclosure relates to the field of computers, and in particular to an information recommendation / determination method and device, and an information recommendation system. Background Art

[0004] With the development of information technology, more and more users choose to browse and purchase items in e-commerce systems. Information recommendation systems in e-commerce systems can identify items of interest to users based on their browsing and purchasing behaviors, and recommend them to them, thereby improving user experience.

[0005] In e-commerce product displays, in addition to single items, there are also aggregated products consisting of multiple items. Aggregated products typically display an image of the target item as the main image, with a description of the aggregated product displayed below the main image. The target item is one of the items in the aggregated product and is the item that drives traffic to the aggregated product page.

[0006] Summary of the Invention

[0007] Some embodiments of the present disclosure provide an information recommendation method, including:

[0008] Using the target user's item click history sequence, query the first correlation information between different items to obtain a set of items related to the target user's historical clicked items;

[0009] Using the target user's item click history sequence, query the second correlation information between items and aggregated items to obtain a set of aggregated items related to the target user's historically clicked items;

[0010] Determining target items of the polymer items in the polymer item set for a target user by matching the polymer items in the polymer item set with items in the item set;

[0011] The combination of the polymer items in the polymer item set and the target items is recommended to the target user.

[0012] In some embodiments, determining the aggregated items in the aggregated item set as target items for the target user includes:

[0013] Matching the polymer items in the polymer item set with items in the item set according to the attribute information satisfied by the target item of the polymer item set in the matching rule;

[0014] An item in the item set that matches the polymer item in the polymer item set is determined as a target item of the polymer item for a target user.

[0015] In some embodiments, recommending a combination of a polymer item in the polymer item set and a target item thereof to a target user includes:

[0016] Calculating the weight of a combination of the aggregated items in the aggregated item set and the target item thereof, to represent the degree of interest of the target user in the combination;

[0017] Filter combinations based on their weights;

[0018] Recommend the selected combinations to target users.

[0019] In some embodiments, calculating the weight of a combination of a polymer item in the polymer item set and its target item includes:

[0020] For any combination of the polymer product in the polymer product set and its target item:

[0021] Obtaining a weight of the target item in the combination from the first correlation information, to represent the degree of interest of the target user in the target item in the combination;

[0022] Obtaining a weight of the aggregated items in the combination from the second correlation information, used to represent the degree of interest of the target user in the aggregated items in the combination;

[0023] The weight of the combination is calculated according to the weight of the target item in the combination and the weight of the aggregate items in the combination.

[0024] In some embodiments, further comprising:

[0025] Determine the item feature matrix and the aggregate item feature matrix based on the item and aggregate item click history sequence of each user;

[0026] Determining first correlation information between different items based on the item feature matrix;

[0027] Second correlation information between the items and the polymer items is determined based on the item feature matrix and the polymer item feature matrix.

[0028] In some embodiments, determining the item feature matrix and the aggregate item feature matrix includes: using each user's item and aggregate item click history behavior sequence to train the language representation model to obtain the item feature matrix and the aggregate item feature matrix, wherein: the item feature matrix includes the feature vector of each item, and the aggregate item feature matrix includes the feature vector of each aggregate item.

[0029] In some embodiments, training the language representation model includes:

[0030] Divide each user's item and aggregate item click history into multiple sessions. A session includes a sequence of objects clicked by the same user within a unit time. The objects include items and aggregate items. Each session serves as a training sample.

[0031] The language representation model is trained using the training samples, wherein the training parameters include an item feature matrix and a polymer item feature matrix of an encoding layer of the language representation model;

[0032] Output the item feature matrix and aggregate item feature matrix of the encoding layer of the trained language representation model.

[0033] In some embodiments, training the language representation model using each training sample includes:

[0034] Input any first training sample in the same training batch into the language representation model, and the output layer of the language representation model outputs the object probability distribution of the target position in the first training sample based on the first vector.

[0035] The first vector is obtained by concatenating the feature vectors of the items or polymer items corresponding to the positive labels in each second training sample, and the second training sample is other training samples in the same training batch except the first training sample.

[0036] In some embodiments, the language representation model includes: a BERT model, an ELMo model, a GTP model, an XLNet model, or a T5 model.

[0037] In some embodiments, the first correlation information between different items includes: a weight of the second item relative to the first item, which is used to represent the degree of interest of the user in the second item when the user has clicked on the first item.

[0038] In some embodiments, determining first correlation information between different items includes:

[0039] determining a weight of the second item relative to the first item based on an inner product between a feature vector of the first item and a feature vector of the second item in the item feature matrix;

[0040] or,

[0041] The weight of the second item relative to the first item is determined based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix and the posterior click rate of the second item.

[0042] In some embodiments, the second correlation information between the item and the aggregate item includes: a weight of the first aggregate item relative to the first item, used to represent the degree of interest of the user in the first aggregate item if the user has clicked on the first item.

[0043] In some embodiments, determining second correlation information between the item and the polymer item includes:

[0044] determining a weight of the first polymer item relative to the first item based on an inner product between a feature vector of the first item in the item feature matrix and a feature vector of the first polymer item in the polymer item feature matrix;

[0045] or,

[0046] The weight of the first aggregated item relative to the first item is determined based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix and the posterior click rate of the first aggregated item.

[0047] In some embodiments, the method further includes: before recommending to the target user, screening or sorting the combination of the polymer items in the polymer item set and the target items thereof according to one or more preset refinement rules or rearrangement rules.

[0048] In some embodiments, aggregated items include activities, stores, and aggregated pages.

[0049] Some embodiments of the present disclosure provide an information determination method, including:

[0050] Determine the item feature matrix and the aggregate item feature matrix based on the item and aggregate item click history sequence of each user;

[0051] Determining first correlation information between different items based on the item feature matrix;

[0052] Second correlation information between the items and the polymer items is determined based on the item feature matrix and the polymer item feature matrix.

[0053] Some embodiments of the present disclosure provide an information recommendation device, including:

[0054] Memory; and

[0055] A processor is coupled to the memory, and is configured to execute an information recommendation method based on instructions stored in the memory.

[0056] Some embodiments of the present disclosure provide an information determination device, including:

[0057] Memory; and

[0058] A processor is coupled to the memory, and the processor is configured to execute an information determination method based on instructions stored in the memory.

[0059] Some embodiments of the present disclosure provide a non-transitory computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the method of each embodiment are implemented.

[0060] Some embodiments of the present disclosure provide an information recommendation device, including:

[0061] An item set acquisition unit is configured to use a target user's item click history sequence to query first correlation information between different items, and obtain an item set related to the target user's historically clicked items;

[0062] The aggregated item set acquisition unit is configured to query second correlation information between items and aggregated items using the target user's item click history sequence, and obtain an aggregated item set related to the target user's historically clicked items;

[0063] a target item determination unit configured to determine a target item for a target user from the aggregated items in the aggregated item set by matching the aggregated items in the aggregated item set with the items in the item set;

[0064] The information recommendation unit is configured to recommend a combination of the polymer items in the polymer item set and the target items to the target user.

[0065] Some embodiments of the present disclosure provide an information determination device, including:

[0066] a feature determination unit configured to determine an item feature matrix and an aggregated item feature matrix based on a historical click behavior sequence of items and aggregated items of each user;

[0067] a first correlation information determining unit, configured to determine first correlation information between different items based on the item feature matrix;

[0068] The second correlation information determining unit is configured to determine second correlation information between the item and the polymer item based on the item feature matrix and the polymer item feature matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The following briefly introduces the drawings required for describing the embodiments or related technologies. The present disclosure can be more clearly understood based on the following detailed description with reference to the drawings.

[0070] Obviously, the drawings described below are only some embodiments of the present disclosure. A person skilled in the art can obtain other drawings based on these drawings without creative work.

[0071] FIG1 is a schematic diagram showing a polymer article and its target article according to some embodiments of the present disclosure.

[0072] FIG2 shows a schematic diagram of an information recommendation system according to some embodiments of the present disclosure.

[0073] FIG3 shows a schematic diagram of a BERT language representation model according to some embodiments of the present disclosure.

[0074] FIG4 shows a schematic diagram of an information determination method according to some embodiments of the present disclosure.

[0075] FIG5 shows a schematic diagram of an information recommendation method according to some embodiments of the present disclosure.

[0076] FIG6 shows a schematic diagram of an information determination apparatus according to some embodiments of the present disclosure.

[0077] FIG7 shows a schematic diagram of an information recommendation device according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0078] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present disclosure.

[0079] Unless otherwise specified, descriptions such as “first” and “second” in the present disclosure are used to distinguish different objects and are not used to indicate meanings such as size or time sequence.

[0080] Research has found that user interest in aggregated products is primarily reflected in two aspects: interest in the aggregated product itself and interest in the target items displayed within the aggregated product. Current challenges facing information recommendation services include recommending aggregated products of interest to users and selecting the most engaging items within the aggregated product as target items for traffic generation.

[0081] The disclosed embodiment utilizes the target user's item click history behavior sequence to query the first correlation information between different items and the second correlation information between items and aggregated items, respectively, to obtain a set of items and an aggregated item set that the target user may be interested in that are related to the target user's historical clicked items. Then, by matching the aggregated items in the aggregated item set with the items in the item set, the target items of the aggregated items in the aggregated item set are determined to be the target items of the target user, and a combination of the aggregated items in the aggregated item set and their target items is recommended to the target user. Thus, the aggregated items and target items of interest to each user are accurately mined and located, and personalized combinations of the aggregated items and target items of interest to each user are recommended, recommending different combinations of aggregated items and target items to users with different interests, achieving a personalized information recommendation effect and improving the user experience.

[0082] Figure 1 shows a schematic diagram of a polymer product and its target item in some embodiments of the present disclosure. The target item is one of the multiple items included in the polymer product, and is an item that diverts traffic to the polymer product page. Therefore, it is also called a diversion item. As shown in Figure 1, the display format of the polymer product is usually: the picture of the target item is used as the main picture for display, and the descriptive information related to the polymer product is displayed below the main picture. After the user clicks on the picture of the target item, he enters the polymer product page and can continue to browse the various items on the polymer product page. For example, polymer products include but are not limited to activities, stores, aggregation pages, etc., which include a collection of multiple items.

[0083] Figure 2 illustrates a schematic diagram of an information recommendation system according to some embodiments of the present disclosure. As shown in Figure 2 , the information recommendation system 200 includes, for example, an identification acquisition module 210, a data acquisition module 220, an offline mining module 230, and an online recall module 240. Specifically, the information recommendation system 200 includes an information determination device 600 and an information recommendation device 700. The information determination device 600 includes, for example, the offline mining module 230 and may also include the identification acquisition module 210 and the data acquisition module 220. The offline mining module 230 may include, for example, a feature determination unit 231, a first correlation information determination unit 232, and a second correlation information determination unit 233. The information recommendation device 700 includes, for example, an online recall module 240 and may also include the identification acquisition module 210 and the data acquisition module 220. The online recall module 240 may include, for example, an item set acquisition unit 241, an aggregated item set acquisition unit 242, a target item determination unit 243, and an information recommendation unit 244.

[0084] The following describes each module and unit in FIG2 .

[0085] The identifier acquisition module 210 is configured to obtain the identifier of a user browsing information in the e-commerce system (used to uniquely identify the user) and transmit this identifier to the data acquisition module 220. When the identifier acquisition module 210 is applied to the online recall module 240, in an information flow scenario, the obtained user identifier is transmitted to the data acquisition module 220 to trigger the online recall module 240 to make information recommendations. When the identifier acquisition module 210 is applied to the offline mining module 230, in either an information flow or search scenario, the obtained user identifier is transmitted to the data acquisition module 220, so that the offline mining module 230 can determine first correlation information between different items and second correlation information between items and aggregated items based on the user's historical behavior.

[0086] The data collection module 220 is configured to collect the user's click history behavior data on the e-commerce system based on the user's identification and update it regularly. The click history behavior data includes, for example, item click history behavior sequences and aggregated item click history behavior sequences.

[0087] The offline mining module 230 is configured to mine first correlation information between different items and second correlation information between items and aggregated items based on the historical click behavior sequences of items and aggregated items of each user collected by the data collection module 220, and can also transmit the mined information to the network disk for use by the online recall module 240.

[0088] The offline mining module 230, for example, includes a feature determination unit 231, a first correlation information determination unit 232, and a second correlation information determination unit 233. The first correlation information determination unit 232 and the second correlation information determination unit 233 can also be combined into a single unit, referred to as the correlation information determination unit. The feature determination unit 231 is configured to determine an item feature matrix and an aggregated item feature matrix based on the item and aggregated item click history of each user. The first correlation information determination unit 232 is configured to determine first correlation information between different items based on the item feature matrix. For example, the first correlation information between different items may include the weight of the second item relative to the first item, which indicates the user's level of interest in the second item after clicking on the first item. The first correlation information may be in the form of a vocabulary. Each record in the first correlation information vocabulary may be in the form of, for example, <first item, second item, weight of the second item relative to the first item>. The second correlation information determination unit 233 is configured to determine second correlation information between the items based on the item feature matrix and the aggregated item feature matrix. The second correlation information between the item and the aggregated item includes: a weight of the first aggregated item relative to the first item, which indicates the user's interest in the first aggregated item after clicking on the first item. The second correlation information may be in the form of a vocabulary. Each record in the second correlation information vocabulary may be in the form of <first item, first aggregated item, weight of the first aggregated item relative to the first item>.

[0089] The following is a detailed description of each unit in the offline mining module 230.

[0090] In some embodiments, feature determination unit 231 is configured to train a language representation model using individual users' item and aggregated item click history sequences to generate an item feature matrix and an aggregated item feature matrix. The item feature matrix includes a feature vector for each item, and the aggregated item feature matrix includes a feature vector for each aggregated item. In this case, feature determination unit 231 is also referred to as a pre-training unit.

[0091] The following is a detailed description of the feature determination unit 231 from three aspects: data preprocessing, model training, and feature generation.

[0092] The data preprocessing part is explained in detail below.

[0093] The historical click behavior sequence of each user's items and aggregated items is divided into multiple sessions. The same session includes the sequence of objects clicked by the same user within a unit time. The objects include items and aggregated items. Each session is used as a training sample.

[0094] For example, a user's click history of items and aggregated items in the past 10 days is selected as training data. The click history of items and aggregated items in the 10 days is divided into multiple sessions, for example, with 10 minutes as the boundary. The same session includes items or aggregated items clicked by the same user within 10 minutes, and the session length (the number of items or aggregated items clicked by the user in one session) is approximately between 4 and 16. If the click time of two clicked objects is greater than 10 minutes apart, the two clicked objects are divided into two sessions. Furthermore, the same session can include items or aggregated items clicked by the same user on the same access device within a unit time, thereby distinguishing the click behavior of the same user ID on different access devices.

[0095] The model training part is explained in detail below.

[0096] The language representation model is used to compress the semantics of language units (such as characters and words) in natural language into a low-dimensional, compact vector space. The language representation model is generally obtained by pre-training. The language representation model is trained using the various training samples obtained above. The language representation model includes an encoding layer, a hidden layer, and an output layer. The training parameters of the language representation model include: the item feature matrix and the aggregate item feature matrix of the encoding layer, and may also include the position vector matrix of the encoding layer, and may also include linear transformation parameters such as weights and biases of the hidden layer, and may also include linear transformation parameters such as weights and biases of the output layer.

[0097] Language representation models include, but are not limited to, BERT (Bidirectional Encoder Representation from Transformer), ELMo (Embeddings from Language Models), GTP (Generative Pre-Training), XLNet (a generalized autoregressive language model for learning unsupervised representations of text sequences), and T5 (Transfer Text-to-Text Transformer).

[0098] The following uses the BERT model as an example to illustrate the language representation model and its training process.

[0099] As shown in Figure 3, a BERT model, the Bert4Rec model, mainly includes three sub-networks: encoding layer, self-attention layer, and output layer.

[0100] The BERT model training process involves replacing some words in each training text at the encoding layer. The resulting text is then vectorized and fed into the self-attention layer. The self-attention mechanism learns the underlying semantic information in the text, and the output layer predicts the labels of the replaced words. The BERT model is continuously trained using each training text until it can accurately predict the labels of the replaced words.

[0101] Specifically in this embodiment, as mentioned above, each training sample is a session, and the same session includes a sequence of objects clicked by the same user within a unit of time, and the objects include items and aggregated items. Therefore, the training process of the BERT model in this embodiment includes: for each session, at the encoding layer, randomly replacing an object (item or aggregated item) in the session, replacing the replaced object with a mask, performing position encoding and feature encoding on the processed session, and then inputting it into the self-attention layer. The self-attention mechanism learns the potential semantic information in the session, and predicts the mask position at the output layer. Using each training session, the BERT model is continuously trained until the BERT model can accurately predict the mask position of the replaced object. The accuracy of the BERT model's prediction (projection) meets the preset conditions, and the preset conditions can be, for example, that the prediction accuracy reaches a certain threshold. As a result, the trained BERT model can predict the object that the user may click at the current position based on the object clicked previously and / or the object clicked later by the user.

[0102] The following describes in detail the training process of the BERT model in this embodiment with reference to the BERT language representation model (BERT model for short) shown in FIG3 .

[0103] Coding layer

[0104] The encoding layer generates a semantic vector for any clicked object in the click object sequence of the session.

[0105] First, in each session’s click object sequence, one of the objects (items or aggregated items) is randomly replaced, and the replaced object v t Use mask instead.

[0106] Then, the encoding layer generates a specific position vector for each clicked object (item or aggregated object) in the click object sequence of the session to represent the click time interval between the clicked objects in the click object sequence of the session. In some embodiments, for any clicked object v in the session, i , by querying the position vector matrix P, we can get v i The corresponding position vector P i For example, for v t, by querying the position vector matrix P, we can get v t The corresponding position vector P t The position vector matrix P is one of the parameters that need to be trained. Its value can be randomly initialized before training. The position vector matrix P can be set using the tf.Variable function of the TensorFlow framework.

[0107] Then, the encoding layer generates a feature vector for each clicked object (item or aggregated item) in the clicked object sequence of the session. In some embodiments, for any clicked object v in the session, i , if v i is an item, by querying the item feature matrix V S , we get v i The corresponding eigenvector V i ; If v i It is a polymer product. By querying the polymer product feature matrix V A , we get v i The corresponding eigenvector V i Among them, the item feature matrix V S and polymer product characteristic matrix V A It is a parameter that needs to be trained, and its value can be randomly initialized before training. S and polymer product characteristic matrix V A It can be set using the tf.Variable function of the Tensorflow framework.

[0108] Finally, for any clicked object v in the session i , v i The corresponding position vector P i and v i The corresponding eigenvector V i Add them together to get v i The semantic vector E at the encoding layer i .

[0109] Self-Attention Layer

[0110] The self-attention layer captures the correlation between different objects in the click object sequence of the same session through the self-attention mechanism, which includes the correlation between different items (first correlation information) and the correlation between items and aggregated items (second correlation information).

[0111] In order to fully extract the semantic information between different objects, the self-attention layer usually includes multiple layers, and each layer structure can be defined as Trm lEach self-attention layer contains two sub-modules: a multi-head self-attention layer and a feed-forward neural network layer. After each sub-module, a residual connection and layer normalization module can be connected.

[0112] (a) Multi-head self-attention layer

[0113] The input vector of the current self-attention layer is It is the output vector of the previous self-attention layer. It maps the input vector to multiple independent feature subspaces through a linear transformation mechanism, and then captures the semantic information between different dimensions through multiple self-attention heads.

[0114] First, for the i-th self-attention head, the input vector is mapped to three independent subspaces through three different linear transformation matrices. That is, the input vector is multiplied by three different linear transformation matrices to obtain three linearly transformed output vectors.

[0115] That is, assuming: are 3 linear transformation matrices, take Represents the output vector after linear transformation.

[0116] Then, a dot product attention calculation is performed on the output feature vectors Q, K, and V. The attention calculation method includes, for example, performing a dot product operation on the feature vector Q and the transpose of the feature vector K, calculating the quotient of the dimension of the input vector and the number of self-attention heads, calculating the square root of the quotient, performing a softmax operation on the quotient of the dot product operation result and the square root result, obtaining the attention weights between different objects in the click object sequence of the session, performing a dot product operation on the attention weights and the feature vector V, and obtaining the attention between different objects in the click object sequence of the session as the output vector of the i-th self-attention head.

[0117] The formula for calculating the output vector of the i-th self-attention head is as follows:

[0118] Where: softmax(*) represents the attention weights between different objects in the click object sequence of the session, and d represents the input vector H l The dimension of h represents the number of self-attention heads.

[0119] Multiple self-attention heads in the self-attention layer are calculated in parallel, and each self-attention head is based on the input vector H l The corresponding output vector head is obtained i .

[0120] Finally, the output vectors of multiple self-attention heads in the same layer are concatenated and linearly transformed using the linear transformation matrix to obtain the output of the multi-head self-attention layer of the layer. The formula is expressed as: MultiHead l (H l ,H l ,H l )=Concat(head1,head2,…,head h )W O

[0121] Among them, head i Represents the output vector of the i-th attention head, where i ranges from 1,…,h, and h represents the number of attention heads. Concat() represents concatenating the output vectors of each self-attention head by dimension. W O Represents the linear transformation matrix, H l Represents the input vector of the lth self-attention layer, MultiHead l Represents the output of the l-th multi-head self-attention layer.

[0122] (b) Feedforward neural network layer

[0123] To ensure the nonlinearity of the network structure and the correlation between different dimensions of the feature vectors, a feed-forward neural network is added after the multi-head attention layer to perform a linear transformation on the output of the aforementioned multi-head attention layer, wherein: linear transformation parameters such as weights and biases are used for linear transformation; this linear transformation can be performed multiple times, for example, twice.

[0124] For example, the feedforward neural network uses the linear transformation parameters in the first layer of the feedforward neural network to perform the first linear transformation on the output of the connected multi-head attention layer. The result of the first linear transformation is acted upon by a nonlinear activation function, and the linear transformation parameters in the second layer of the feedforward neural network are used to perform the second linear transformation on the result of the above action to produce the output of the feedforward neural network. The calculation formula is as follows: FFN l (x) = CELU (xW (1) +b (1) )W (2) +b (2)

[0125] Among them, x represents the output of the multi-head self-attention layer, FFN l (x) represents the output of the feedforward neural network, W (1) ,b (1) Represents the linear transformation parameters in the first layer of the feedforward neural network, which are weights and biases, W (2) ,b (2)Represents the linear transformation parameters in the second layer of feedforward neural network, which are weight and bias respectively, and GELU represents the nonlinear activation function in the first layer of feedforward neural network.

[0126] (c) Residual connection and layer normalization module

[0127] In order to accelerate the convergence of the model and ensure the stability of data distribution during model training, the residual connection and layer normalization modules are added after the multi-head self-attention layer and the feedforward neural network layer respectively. This module performs residual connection and normalizes the residual connection result, and finally obtains the output Hl of the l-th layer self-attention layer. C1 , the calculation process is as follows: f out (x)=LayerNorm(x+SunLayer(x))

[0128] Where x represents the input of the multi-head self-attention layer or the feedforward neural network layer, SubLayer(x) represents the output of the multi-head self-attention layer or the feedforward neural network layer, x+SubLayer(x) represents the residual connection, LayerNorm(*) represents layer normalization, that is, normalizing all neurons in an intermediate layer, and f out (x) represents the output of the residual connection and layer normalization module, where: When x represents the input of the feedforward neural network layer, f out (x) is the output H of the lth self-attention layer l+1 .

[0129] Output layer

[0130] The output layer is used to predict the mask position. The output layer uses a nonlinear feedforward neural network and is normalized by the softmax function to obtain the object probability distribution of the target position t in the click object sequence of the session. For example, the linear transformation parameters of the output layer are used to linearly transform the output of the last self-attention layer, and then the nonlinear activation function in the output layer is used to process the linear transformation result. The processed result is multiplied with the learnable item feature matrix and the aggregate item feature matrix in the encoding layer. The sum of the multiplication result and the bias term is normalized using the softmax function to obtain the object probability distribution of the target position t in the click object sequence of the session. The calculation formula is:

[0131] in, Represents the output of the last layer (layer L) of the self-attention layer, W P ,b P Represents the linear transformation parameter of the output layer, CELU represents the nonlinear activation function in the output layer, V S 、VA Represent the learnable item feature matrix and aggregate item feature matrix in the encoding layer, respectively, b O represents the bias term, and P(v) represents the probability distribution.

[0132] If the number of object types involved in the conversation is relatively large, for example, an e-commerce system includes hundreds of millions of items and aggregated items, considering that directly performing multiple classification on hundreds of millions of objects will bring huge computational overhead, it will seriously reduce the accuracy of model estimation. In order to improve the efficiency of model training, ensure that the model can run stably on large-scale training sets (such as hundreds of millions), and prevent the model from exceeding the memory limit during training, this embodiment inputs any first training sample in the same training batch into the language representation model, and the output layer of the language representation model is based on the first vector Output the object probability distribution of the target position in the first training sample, where: the first vector The second training samples are obtained by concatenating the feature vectors of the items or aggregate items corresponding to the positive labels in each second training sample, and the second training samples are other training samples in the same training batch except the first training samples.

[0133] That is, the linear transformation parameters of the output layer are used to linearly transform the output of the last self-attention layer, and then the nonlinear activation function in the output layer is used to process the linear transformation result, and the processed result is combined with the first vector Multiply, and use the softmax function to normalize the sum of the multiplication result and the bias term to obtain the object probability distribution of the target position t in the click object sequence of the session. The calculation formula is improved to:

[0134] Following the above method, the BERT model is trained continuously using each session as training samples until it can accurately predict the mask position of the replaced object. Then, the feature generation part is performed.

[0135] The feature generation part is explained in detail below.

[0136] After the language representation model is trained, the item feature matrix and aggregated item feature matrix of the encoding layer of the trained language representation model are output and saved to an offline file. The item feature matrix records the mapping relationship between each item and its feature vector, and the aggregated item feature matrix records the mapping relationship between each aggregated item and its feature vector.

[0137] The following describes the correlation information determination units (the first correlation information determination unit 232 and the second correlation information determination unit 233 ) in detail.

[0138] The first correlation information determination unit 232 is configured to determine first correlation information between different items based on the item feature matrix. The first correlation information between different items includes the weight of the second item relative to the first item, which is used to represent the user's level of interest in the second item after clicking on the first item. Two exemplary methods for determining the first correlation information are described below. The first method is to determine the weight of the second item relative to the first item based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix. For example, this inner product is used as the weight of the second item relative to the first item. The second method is to determine the weight of the second item relative to the first item based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix and the posterior click-through rate of the second item. For example, the product of the inner product and the posterior click-through rate of the second item is used as the weight of the second item relative to the first item. Furthermore, as previously described, one of the records of the first correlation information in the vocabulary form may be, for example, <first item, second item, weight of the second item relative to the first item>.

[0139] The second correlation information determination unit 233 is configured to determine second correlation information between the items and the aggregated items based on the item feature matrix and the aggregated item feature matrix. The second correlation information between the items and the aggregated items includes: a weight of the first aggregated item relative to the first item, which is used to represent the user's level of interest in the first aggregated item after clicking on the first item. Two exemplary methods for determining the second correlation information are described below. The first method is to determine the weight of the first aggregated item relative to the first item based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix. For example, this inner product is used as the weight of the first aggregated item relative to the first item. The second method is to determine the weight of the first aggregated item relative to the first item based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix, as well as the posterior click-through rate of the first aggregated item. For example, the product of the inner product and the posterior click-through rate of the first aggregated item is used as the weight of the first aggregated item relative to the first item. In addition, as mentioned above, one of the records of the second correlation information in the vocabulary form is, for example, <first item, first aggregate item, weight of the first aggregate item relative to the first item>.

[0140] The inner product between two vectors mentioned above is calculated by multiplying the first vector by the second vector, taking the modulus, and then dividing the result by the product of the modulus of the first vector and the modulus of the second vector. The resulting quotient is used as the inner product between the first and second vectors. The inner product between two vectors can also be calculated using the Faiss tool.

[0141] In some embodiments, the first correlation information and the second correlation information can also be filtered. For example, according to a set weight threshold, the first correlation information and the second correlation information with weights higher than the threshold are filtered out, thereby filtering out items or aggregate items with relatively high user interest.

[0142] The online recall module 240 and its involved item set acquisition unit 241 , aggregated item set acquisition unit 242 , target item determination unit 243 , and information recommendation unit 244 are described in detail below.

[0143] The online recall module 240 accurately locates the aggregate items and combinations of target items that each target user is interested in based on the first correlation information between different items and the second correlation information between items and aggregate items generated by the offline mining module 230, and makes personalized combination recommendations for each target user.

[0144] Item collection acquisition unit 241 is configured to use the target user's item click history sequence to query first correlation information between different items, obtain an item collection related to the target user's historically clicked items, and further obtain the weights of the items in the item collection relative to the target user's historically clicked items. This allows the target user to retrieve an item collection that may be of interest to the target user and determine the user's level of interest in the retrieved items.

[0145] That is, the items historically clicked by the target user are used as the first items in the first correlation information to query the first correlation information, and the second items in the queried first correlation information and the weight of the second items relative to the first item are used as the set of items related to the items historically clicked by the target user and the weight of each item in the item set.

[0146] For example, the set of items that the target user may be interested in is {sku1,sku2,…,sku m}, the target users’ interest in each recalled item is: {sku1_weight,sku2_weight,…,sku m _weight}.

[0147] Aggregate item set acquisition unit 242 is configured to use the target user's item click history sequence to query second correlation information between items and aggregated items, thereby obtaining an aggregated item set related to the target user's historically clicked items. It can also obtain the weights of the aggregated items in the aggregated item set relative to the target user's historically clicked items. This allows the target user to retrieve aggregated item sets that may be of interest to the target user and determine the user's level of interest in the retrieved aggregated items.

[0148] That is, the items historically clicked by the target user are used as the first items in the second correlation information, and the second correlation information is queried. The first aggregated items in the queried second correlation information and the weight of the first aggregated items relative to the first item are used as the aggregated item set related to the items historically clicked by the target user, and the weight of each aggregated item in the aggregated item set.

[0149] For example, the set of aggregated items (such as activities) that the target user may be interested in is {activity1,activity2,…,activity n}, the target users’ interest in each of the recalled polymer products is: {activity1_weight,activity2_weight,…,activity m _weight}.

[0150] The target item determination unit 243 is configured to determine the target item of the aggregated items in the aggregated item set for the target user by matching the aggregated items in the aggregated item set with the items in the item set.

[0151] In some embodiments, the target item determination unit 243 is configured to match the polymer items in the polymer item set with the items in the item set according to the attribute information satisfied by the target items of the polymer items set in the matching rules; and determine the items in the item set that match the polymer items in the polymer item set as the target items of the polymer items for the target user.

[0152] The attribute information that the target item of the polymer items set by the matching rule satisfies includes, but is not limited to, one or more of the store information, category information, brand information, etc. to which the target item should belong.

[0153] In some embodiments, before recommending a combination of a polymer item and its target item in the polymer item set to a target user, the weights of the combinations of the polymer item and its target item in the polymer item set are calculated to represent the target user's interest in the combination; combinations are screened based on the weights of the combinations; and the screened combinations are then recommended to the target user. This process can be performed by the target item determination unit 243, the information recommendation unit 244, or another separate unit. Thus, the polymer item and its target item combination that the target user is more interested in is recommended to the target user.

[0154] In some embodiments, calculating the weight of a combination of the polymer items in the polymer item set and their target items includes: for any combination of the polymer items in the polymer item set and their target items: obtaining the weight of the target item in the combination from the first correlation information, which is used to characterize the degree of interest of the target user in the target item in the combination; obtaining the weight of the polymer items in the combination from the second correlation information, which is used to characterize the degree of interest of the target user in the polymer items in the combination; and calculating the weight of the combination based on the weight of the target item in the combination and the weight of the polymer items in the combination.

[0155] Among them, obtaining the weight of the target item in the combination from the first correlation information, for example, includes: querying the record that the second item is the target item in the combination from the first correlation information, and the weight of the target item relative to the first item in the record is the weight of the target item.

[0156] Among them, obtaining the weight of the polymer item in the combination from the second correlation information, for example, includes: querying the record from the second correlation information that the first polymer item is a polymer item in the combination, and the weight of the polymer item relative to the first item in the record is the weight of the polymer item.

[0157] Calculating the weight of the combination based on the weight of the target item in the combination and the weight of the aggregated items in the combination, for example, includes multiplying the weight of the target item in the combination by the weight of the aggregated items in the combination to determine the weight of the combination. This allows for a more accurate determination of the user's overall interest in the aggregated items by taking both the user's interest in the target item and the aggregated items into consideration.

[0158] The information recommendation unit 244 is configured to recommend aggregated items and combinations of target items in the aggregated item set to the target user. This recommends aggregated items and combinations of target items that the target user is interested in. This recommends different aggregated items and combinations of target items to users with different interests, achieving personalized information recommendation and enhancing user experience.

[0159] In some embodiments, the information recommendation unit 244 is configured to filter or sort the combinations of the polymer items and the target items in the polymer item set according to one or more of the preset refinement rules or re-ranking rules before recommending the items to the target user, and then recommend the filtered or sorted results to the target user.

[0160] For example, the precise ranking rule is: score the combination of the aggregate product and its target product according to the predicted click-through rate (pCTR). The higher the predicted click-through rate, the higher the score, the higher the ranking, and the combination can be selected for recommendation and display first.

[0161] For another example, the precise ranking rule is to score the combination of polymer items and their target items according to the predicted revenue. The predicted revenue is determined based on the predicted click-through rate and the cost per click. For example, the product of the predicted click-through rate and the cost per click is used as the predicted revenue. The higher the predicted revenue, the higher the score, the higher the ranking, and the combination can be selected for recommendation and display first.

[0162] For example, the rearrangement rule is to score the combination of aggregated items and their target items according to the rearrangement score. The higher the score, the higher the ranking, and the combination can be selected for recommendation and display. The rearrangement score is determined based on at least one of the novelty score and the window rule score. This balances accuracy and richness. k Probability distribution P(w k |S) and the i-th item s in the set S i Product word w k Probability distribution P(w k |s i ), and calculates a novelty score based on the Kullback-Leibler distance. Diversity constraints can be used as window rules to limit the frequency of occurrence of objects of the same category, brand, store, or material within the same window. A window rule can include one or more diversity constraints, each with different weights, which can be set based on importance. The sum of the weights of each diversity constraint satisfied by an object serves as the window rule score for that object.

[0163] The refined sorting rules or rearrangement rules for screening or sorting in the embodiments of the present disclosure are not limited to the above examples.

[0164] FIG4 is a schematic diagram of an information determination method according to some embodiments of the present disclosure, wherein the determined information includes first correlation information between different items and second correlation information between an item and a polymer item.

[0165] As shown in FIG4 , the information determination method includes the following steps.

[0166] In step 410, an item feature matrix and an aggregated item feature matrix are determined based on the item and aggregated item click history behavior sequences of each user.

[0167] In some embodiments, the language representation model is trained using the historical click behavior sequences of each user's items and aggregated items to obtain an item feature matrix and an aggregated item feature matrix, wherein: the item feature matrix includes a feature vector for each item, and the aggregated item feature matrix includes a feature vector for each aggregated item.

[0168] In some embodiments, training a language representation model using the item and aggregate item click history sequences of each user includes:

[0169] Divide each user's item and aggregate item click history into multiple sessions. A session includes a sequence of objects clicked by the same user within a unit time. The objects include items and aggregate items. Each session serves as a training sample.

[0170] The language representation model is trained using the training samples, wherein the training parameters include an item feature matrix and a polymer item feature matrix of an encoding layer of the language representation model;

[0171] Output the item feature matrix and aggregate item feature matrix of the encoding layer of the trained language representation model.

[0172] In some embodiments, training a language representation model using various training samples includes: inputting any first training sample within the same training batch into the language representation model, and the output layer of the language representation model outputs the object probability distribution of the target position in the first training sample based on the first vector, wherein: the first vector is obtained by splicing the feature vectors of the items or polymer items corresponding to the positive labels in each second training sample, and the second training samples are other training samples in the same training batch except the first training samples.

[0173] This improves model training efficiency, ensures that the model can run stably on large-scale training sets (such as hundreds of millions), and prevents errors such as exceeding memory limits during model training.

[0174] In step 420, first correlation information between different items is determined based on the item feature matrix.

[0175] The first correlation information between different items includes: the weight of the second item relative to the first item, which is used to represent the degree of interest of the user in the second item when the user has clicked on the first item.

[0176] In some embodiments, the weight of the second item relative to the first item is determined based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix. For example, the inner product is used as the weight of the second item relative to the first item.

[0177] In some embodiments, the weight of the second item relative to the first item is determined based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix and the posterior click-through rate of the second item. For example, the weight of the second item relative to the first item is determined by multiplying the inner product by the posterior click-through rate of the second item.

[0178] In step 430, second correlation information between the items and the polymer items is determined based on the item feature matrix and the polymer item feature matrix.

[0179] The second correlation information between the item and the aggregated item includes: a weight of the first aggregated item relative to the first item, which is used to represent the degree of interest of the user in the first aggregated item when the user has clicked on the first item.

[0180] In some embodiments, the weight of the first polymer item relative to the first item is determined based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first polymer item in the polymer item feature matrix. For example, the inner product is used as the weight of the first polymer item relative to the first item.

[0181] In some embodiments, the weight of the first aggregated item relative to the first item is determined based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix, as well as the posterior click-through rate of the first aggregated item. For example, the product of the inner product and the posterior click-through rate of the first aggregated item is used as the weight of the first aggregated item relative to the first item.

[0182] Thus, based on each user's click history of items and aggregated items, we can mine primary correlation information between different items and secondary correlation information between items and aggregated items, providing support for personalized information recommendations. Furthermore, we optimize the training process to improve model training efficiency, ensure the model can run stably on large-scale training sets (e.g., hundreds of millions of items), and prevent errors such as exceeding memory limits during training.

[0183] FIG5 is a schematic diagram of an information recommendation method according to some embodiments of the present disclosure, wherein the recommended information includes a combination of a polymer item and a target item.

[0184] As shown in FIG5 , the information recommendation method includes the following steps.

[0185] In step 510, the target user's item click history is used to query first correlation information between different items, obtaining a set of items related to the target user's historically clicked items. The weights of the items in the set relative to the target user's historically clicked items can also be obtained. Thus, a set of items that the target user may be interested in is retrieved, and the user's level of interest in the recalled items is determined.

[0186] That is, the items historically clicked by the target user are used as the first items in the first correlation information to query the first correlation information, and the second items in the queried first correlation information and the weight of the second items relative to the first item are used as the set of items related to the items historically clicked by the target user and the weight of each item in the item set.

[0187] In step 520, the target user's item click history is used to query the second correlation information between the items and the aggregated items, thereby obtaining a set of aggregated items related to the target user's historically clicked items. The weights of the aggregated items in the aggregated item set relative to the target user's historically clicked items can also be obtained. Thus, a set of aggregated items that the target user may be interested in is retrieved, and the user's level of interest in the recalled aggregated items is determined.

[0188] That is, the items historically clicked by the target user are used as the first items in the second correlation information, and the second correlation information is queried. The first aggregated items in the queried second correlation information and the weight of the first aggregated items relative to the first item are used as the aggregated item set related to the items historically clicked by the target user, and the weight of each aggregated item in the aggregated item set.

[0189] In step 530 , target items of the aggregated items in the aggregated item set for the target user are determined by matching the aggregated items in the aggregated item set with the items in the item set.

[0190] In some embodiments, the polymer items in the polymer item set are matched with the items in the item set according to the attribute information satisfied by the target items of the polymer items set according to the matching rules; and the items in the item set that match the polymer items in the polymer item set are determined as the target items of the polymer items for the target user.

[0191] The attribute information that the target item of the polymer items set by the matching rule satisfies includes, but is not limited to, one or more of the store information, category information, brand information, etc. to which the target item should belong.

[0192] In step 540 , a combination of the aggregate items in the aggregate item set and the target items is recommended to the target user.

[0193] In some embodiments, weights of combinations of polymer items and target items in the polymer item set are calculated to represent the target user's interest in the combinations; combinations are screened based on the weights of the combinations; and the screened combinations are recommended to the target user. Thus, combinations of polymer items and target items that are of greater interest to the target user are recommended to the target user.

[0194] Calculating the weight of the combination of the polymer item in the polymer item set and the target item includes:

[0195] For any combination of the polymer product in the polymer product set and its target item:

[0196] Obtaining a weight of the target item in the combination from the first correlation information, to represent the degree of interest of the target user in the target item in the combination;

[0197] Obtaining a weight of the aggregated items in the combination from the second correlation information, used to represent the degree of interest of the target user in the aggregated items in the combination;

[0198] The weight of the combination is calculated based on the weight of the target item in the combination and the weight of the aggregated items in the combination. For example, the weight of the combination is determined by multiplying the weight of the target item in the combination by the weight of the aggregated items in the combination. This allows for a more accurate determination of the user's overall interest in the aggregated items by taking into account both the user's interest in the aggregated items and the user's interest in the target item.

[0199] In some embodiments, before recommending to a target user, the combination of the polymer items in the polymer item set and the target item is screened or sorted according to one or more of the preset refinement rules or rearrangement rules. The refinement rules and rearrangement rules are described above in detail and will not be described in detail here.

[0200] This allows us to accurately identify and locate the aggregated items and target items that each user is interested in, and then recommend personalized combinations of these items. This allows us to recommend different combinations of aggregated items and target items to users with different interests, achieving personalized information recommendations and improving the user experience. Furthermore, by simultaneously considering the user's interest in the aggregated items and the target items, we can more accurately determine the user's overall interest in the aggregated items.

[0201] FIG6 shows a schematic diagram of an information determination apparatus according to some embodiments of the present disclosure.

[0202] As shown in FIG6 , the information determination apparatus 600 of this embodiment includes a memory 610 and a processor 620 coupled to the memory 610 . The processor 620 is configured to execute the information determination method of any of the aforementioned embodiments based on instructions stored in the memory 610 .

[0203] The apparatus 600 may further include an input / output interface 630 , a network interface 640 , a storage interface 650 , etc. These interfaces 630 , 640 , 650 , the memory 610 , and the processor 620 may be connected via a bus 660 , for example.

[0204] FIG7 shows a schematic diagram of an information recommendation device according to some embodiments of the present disclosure.

[0205] As shown in FIG7 , the information recommendation apparatus 700 of this embodiment includes a memory 710 and a processor 720 coupled to the memory 710 . The processor 720 is configured to execute the information recommendation method of any of the aforementioned embodiments based on instructions stored in the memory 710 .

[0206] The apparatus 700 may further include an input / output interface 730 , a network interface 740 , a storage interface 750 , etc. These interfaces 730 , 740 , 750 , the memory 710 , and the processor 720 may be connected, for example, via a bus 760 .

[0207] The memories 610 and 710 may include, for example, system memory, fixed non-volatile storage media, etc. The system memory may store, for example, an operating system, application programs, a boot loader, and other programs.

[0208] The processors 620 and 720 may be implemented using a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, or discrete hardware components such as discrete gates or transistors.

[0209] Among them, input and output interfaces 630 and 730 provide connection interfaces for input and output devices such as displays, mice, keyboards, and touch screens. Network interfaces 640 and 740 provide connection interfaces for various networked devices. Storage interfaces 650 and 750 provide connection interfaces for external storage devices such as SD cards and USB flash drives. Buses 660 and 760 can use any of a variety of bus structures. For example, bus structures include but are not limited to Industry Standard Architecture (ISA) buses, Micro Channel Architecture (MCA) buses, and Peripheral Component Interconnect (PCI) buses.

[0210] The embodiments of the present disclosure provide a non-transitory computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the program implements the steps of the information recommendation method and / or the information determination method.

[0211] (1) An information recommendation method, comprising:

[0212] Using the target user's item click history sequence, query the first correlation information between different items to obtain a set of items related to the target user's historical clicked items;

[0213] Using the target user's item click history sequence, query the second correlation information between items and aggregated items to obtain a set of aggregated items related to the target user's historically clicked items;

[0214] Determining target items of the polymer items in the polymer item set for a target user by matching the polymer items in the polymer item set with items in the item set;

[0215] The combination of the polymer items in the polymer item set and the target items is recommended to the target user.

[0216] (2) According to (1), determining that the target items for the target user in the set of aggregated items include:

[0217] Matching the polymer items in the polymer item set with items in the item set according to the attribute information satisfied by the target item of the polymer item set in the matching rule;

[0218] An item in the item set that matches the polymer item in the polymer item set is determined as a target item of the polymer item for a target user.

[0219] (3) According to (1) or (2), recommending to a target user a combination of the polymer item in the polymer item set and the target item thereof includes:

[0220] Calculating the weight of a combination of the aggregated items in the aggregated item set and the target item thereof, to represent the degree of interest of the target user in the combination;

[0221] Filter combinations based on their weights;

[0222] Recommend the selected combinations to target users.

[0223] (4) According to (3), calculating the weight of the combination of the polymer product in the polymer product set and the target product includes:

[0224] For any combination of the polymer product in the polymer product set and its target item:

[0225] Obtaining a weight of the target item in the combination from the first correlation information, to represent the degree of interest of the target user in the target item in the combination;

[0226] Obtaining a weight of the aggregated items in the combination from the second correlation information, used to represent the degree of interest of the target user in the aggregated items in the combination;

[0227] The weight of the combination is calculated according to the weight of the target item in the combination and the weight of the aggregate items in the combination.

[0228] (5) According to (4), calculating the weight of the combination includes: multiplying the weight of the target item in the combination by the weight of the aggregate item in the combination to determine the weight of the combination.

[0229] (6) According to any one of (1) to (5), further comprising:

[0230] Determine the item feature matrix and the aggregate item feature matrix based on the item and aggregate item click history sequence of each user;

[0231] Determining first correlation information between different items based on the item feature matrix;

[0232] Second correlation information between the items and the polymer items is determined based on the item feature matrix and the polymer item feature matrix.

[0233] (7) According to (6), determining the item feature matrix and the aggregate item feature matrix includes: using the item and aggregate item click history behavior sequence of each user to train the language representation model to obtain the item feature matrix and the aggregate item feature matrix, wherein: the item feature matrix includes the feature vector of each item, and the aggregate item feature matrix includes the feature vector of each aggregate item.

[0234] (8) According to (7), training the language representation model includes:

[0235] Divide each user's item and aggregate item click history into multiple sessions. A session includes a sequence of objects clicked by the same user within a unit time. The objects include items and aggregate items. Each session serves as a training sample.

[0236] The language representation model is trained using the training samples, wherein the training parameters include an item feature matrix and a polymer item feature matrix of an encoding layer of the language representation model;

[0237] Output the item feature matrix and aggregate item feature matrix of the encoding layer of the trained language representation model.

[0238] (9) According to (8), training the language representation model using each training sample includes: inputting any first training sample in the same training batch into the language representation model, and the output layer of the language representation model outputs the object probability distribution of the target position in the first training sample based on the first vector, wherein: the first vector is obtained by splicing the feature vectors of the items or polymer items corresponding to the positive labels in each second training sample, and the second training sample is the other training sample in the same training batch except the first training sample.

[0239] (10) According to any one of (7) to (9), the language representation model includes: a BERT model, an ELMo model, a GTP model, an XLNet model, or a T5 model.

[0240] (11) According to any of (6) to (10):

[0241] The first correlation information between different items includes: a weight of the second item relative to the first item, which is used to represent the degree of interest of the user in the second item when the user has clicked the first item;

[0242] Alternatively, determining the first correlation information between different items includes:

[0243] determining a weight of the second item relative to the first item based on an inner product between a feature vector of the first item and a feature vector of the second item in the item feature matrix;

[0244] or,

[0245] The weight of the second item relative to the first item is determined based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix and the posterior click rate of the second item.

[0246] (12) According to any of (6) to (11):

[0247] The second correlation information between the item and the aggregated item includes: a weight of the first aggregated item relative to the first item, used to represent the degree of interest of the user in the first aggregated item if the user has clicked on the first item;

[0248] Alternatively, determining the second correlation information between the item and the polymer item includes:

[0249] determining a weight of the first polymer item relative to the first item based on an inner product between a feature vector of the first item in the item feature matrix and a feature vector of the first polymer item in the polymer item feature matrix;

[0250] or,

[0251] The weight of the first aggregated item relative to the first item is determined based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix and the posterior click rate of the first aggregated item.

[0252] (13) According to any one of items (1) to (12), it also includes: before recommending to the target user, screening or sorting the combination of the polymer items in the polymer item set and the target items thereof according to one or more of the preset refinement rules or rearrangement rules.

[0253] (14) According to any one of (1) to (13), the aggregated items include, but are not limited to, activities, stores, and aggregated pages.

[0254] (15) An information determination method comprising:

[0255] Determine the item feature matrix and the aggregate item feature matrix based on the item and aggregate item click history sequence of each user;

[0256] Determining first correlation information between different items based on the item feature matrix;

[0257] Second correlation information between the items and the polymer items is determined based on the item feature matrix and the polymer item feature matrix.

[0258] (16) According to (15), determining the item feature matrix and the aggregate item feature matrix includes: using the item and aggregate item click history behavior sequence of each user to train the language representation model to obtain the item feature matrix and the aggregate item feature matrix, wherein: the item feature matrix includes the feature vector of each item, and the aggregate item feature matrix includes the feature vector of each aggregate item.

[0259] (17) According to (16), training the language representation model includes:

[0260] Divide each user's item and aggregate item click history into multiple sessions. A session includes a sequence of objects clicked by the same user within a unit time. The objects include items and aggregate items. Each session serves as a training sample.

[0261] The language representation model is trained using the training samples, wherein the training parameters include an item feature matrix and a polymer item feature matrix of an encoding layer of the language representation model;

[0262] Output the item feature matrix and aggregate item feature matrix of the encoding layer of the trained language representation model.

[0263] (18) According to (17), using each training sample to train the language representation model includes:

[0264] Input any first training sample in the same training batch into the language representation model, and the output layer of the language representation model outputs the object probability distribution of the target position in the first training sample based on the first vector.

[0265] The first vector is obtained by concatenating the feature vectors of the items or polymer items corresponding to the positive labels in each second training sample, and the second training sample is other training samples in the same training batch except the first training sample.

[0266] (19) According to any one of (16) to (18), the language representation model includes: a BERT model, an ELMo model, a GTP model, an XLNet model, or a T5 model.

[0267] (20) According to any of (15) to (19):

[0268] The first correlation information between different items includes: a weight of the second item relative to the first item, which is used to represent the degree of interest of the user in the second item when the user has clicked the first item;

[0269] Alternatively, determining the first correlation information between different items includes:

[0270] determining a weight of the second item relative to the first item based on an inner product between a feature vector of the first item and a feature vector of the second item in the item feature matrix;

[0271] or,

[0272] The weight of the second item relative to the first item is determined based on the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix and the posterior click rate of the second item.

[0273] (21) According to any of (15) to (20):

[0274] The second correlation information between the item and the aggregated item includes: a weight of the first aggregated item relative to the first item, used to represent the degree of interest of the user in the first aggregated item if the user has clicked on the first item;

[0275] Alternatively, determining the second correlation information between the item and the polymer item includes:

[0276] determining a weight of the first polymer item relative to the first item based on an inner product between a feature vector of the first item in the item feature matrix and a feature vector of the first polymer item in the polymer item feature matrix;

[0277] or,

[0278] The weight of the first aggregated item relative to the first item is determined based on the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix and the posterior click rate of the first aggregated item.

[0279] (22) According to any one of (15)-(21): Aggregate items include, but are not limited to, events, stores, and aggregate pages.

[0280] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, and storage of user personal information involved in the technical solutions disclosed herein all comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken with respect to user personal information to prevent unauthorized access to user personal information data and to safeguard the security of user personal information, network security, and national security.

[0281] Those skilled in the art will appreciate that embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more non-transitory computer-readable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer program code.

[0282] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0283] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0284] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0285] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. An information recommendation method, comprising: Using the target user's item click history behavior sequence, query the first correlation information between different items to obtain an item set related to the target user's historically clicked items; Using the target user's item click history behavior sequence, query the second correlation information between the item and the aggregated item, and obtain the aggregated item set related to the target user's historically clicked items; Determining target items of the polymerized items in the polymerized items set for the target user by matching the polymerized items in the polymerized items set with the items in the item set; The combination of the polymer items in the polymer item set and the target items is recommended to the target user.

2. According to the method of claim 1, determining the target items of the aggregated items in the aggregated items set for the target user comprises: Matching the polymer items in the polymer item set with items in the item set according to the attribute information satisfied by the target item of the polymer items set in the matching rule; The items in the item set that match the polymer item in the polymer item set are determined as target items of the polymer item for the target user.

3. According to the method of claim 1, recommending to the target user a combination of the aggregated items in the aggregated item set and the target items thereof comprises: Calculating the weight of the combination of the aggregated items in the aggregated items set and the target items thereof, so as to represent the degree of interest of the target user in the combination; Filter combinations based on the weights of each combination; Recommend the selected combinations to target users.

4. According to the method of claim 3, calculating the weight of the combination of the polymer item in the polymer item set and the target item thereof comprises: For any combination of the polymer product in the polymer product set and its target product: Acquire a weight of the target item in the combination from the first correlation information, which is used to represent the degree of interest of the target user in the target item in the combination; Obtaining the weight of the aggregated items in the combination from the second correlation information, which is used to represent the degree of interest of the target user in the aggregated items in the combination; The weight of the combination is calculated according to the weight of the target item in the combination and the weight of the aggregated items in the combination.

5. The method according to any one of claims 1 to 4, further comprising: According to the item and aggregate item click history behavior sequence of each user, determine the item feature matrix and the aggregate item feature matrix; Determining first correlation information between different items according to the item feature matrix; According to the item feature matrix and the polymer item feature matrix, second correlation information between the items and the polymer items is determined.

6. The method of claim 5, wherein determining the item feature matrix and the aggregate item feature matrix comprises: The language representation model is trained using the historical click behavior sequences of each user's items and aggregated items to obtain an item feature matrix and an aggregated item feature matrix, where the item feature matrix includes the feature vector of each item, and the aggregated item feature matrix includes the feature vector of each aggregated item.

7. According to the method of claim 6, training the language representation model comprises: Divide the item and aggregate item click history behavior sequence of each user into multiple sessions, where the same session includes the object sequence clicked by the same user within a unit time, where the objects include items and aggregate items, and each session is used as a training sample; The language representation model is trained using each training sample, wherein the training parameters include an item feature matrix and a polymer item feature matrix of an encoding layer of the language representation model; Output the item feature matrix and aggregate item feature matrix of the encoding layer of the trained language representation model.

8. According to the method of claim 7, using each training sample to train the language representation model comprises: Input any first training sample in the same training batch into the language representation model, and the output layer of the language representation model outputs the object probability distribution of the target position in the first training sample based on the first vector, Wherein: the first vector is obtained by concatenating the items or aggregate items corresponding to the positive labels in each second training sample The second training sample is other training samples in the same training batch except the first training sample.

9. The method according to any one of claims 6 to 8, wherein the language representation model comprises: BERT model, ELMo model, GTP model, XLNet model, or T5 model.

10. The method according to claim 5, wherein the first correlation information between different items comprises: The weight of the second item relative to the first item, which is used to represent the user's interest in the second item if the user has clicked on the first item; Alternatively, determining the first correlation information between different items includes: determining a weight of the second item relative to the first item according to an inner product between a feature vector of the first item and a feature vector of the second item in the item feature matrix; or, The weight of the second item relative to the first item is determined according to the inner product between the feature vector of the first item and the feature vector of the second item in the item feature matrix and the posterior click rate of the second item.

11. The method according to claim 5, wherein the second correlation information between the item and the polymer item comprises: The weight of the first aggregated item relative to the first item, used to represent the user's interest in the first aggregated item if the user has clicked on the first item; Alternatively, determining the second correlation information between the item and the polymer item includes: determining a weight of the first polymerized item relative to the first item based on an inner product between a feature vector of the first item in the item feature matrix and a feature vector of the first polymerized item in the polymerized item feature matrix; or, The weight of the first aggregated item relative to the first item is determined according to the inner product between the feature vector of the first item in the item feature matrix and the feature vector of the first aggregated item in the aggregated item feature matrix and the posterior click rate of the first aggregated item.

12. The method according to any one of claims 1 to 4, further comprising: Before recommending to the target user, the combination of the polymer items in the polymer item set and the target items is screened or sorted according to one or more of the preset refined sorting rules or rearrangement rules.

13. A method for determining information, comprising: According to the item and aggregate item click history behavior sequence of each user, determine the item feature matrix and the aggregate item feature matrix; Determining first correlation information between different items according to the item feature matrix; According to the item feature matrix and the polymer item feature matrix, second correlation information between the items and the polymer items is determined.

14. An information recommendation device, comprising: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the information recommendation method according to any one of claims 1 to 12 based on instructions stored in the memory.

15. An information determination device, comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the information determination method of claim 13 based on instructions stored in the memory.

16. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

17. An information recommendation device, comprising: An item set acquisition unit is configured to query first correlation information between different items by using the item click history behavior sequence of the target user, and obtain an item set related to the items clicked by the target user in history; The aggregated item set acquisition unit is configured to query the second correlation information between the item and the aggregated item by using the item click history behavior sequence of the target user, and obtain the aggregated item set related to the item clicked by the target user in history; a target item determination unit configured to determine a target item for a target user from the aggregated items in the aggregated item set by matching the aggregated items in the aggregated item set with the items in the item set; The information recommendation unit is configured to recommend polymer products in the polymer product set and their The target item combination.

18. An information determination device, comprising: A feature determination unit is configured to determine an item feature matrix and an aggregate item feature matrix according to the item and aggregate item click history behavior sequence of each user; a first correlation information determining unit, configured to determine first correlation information between different items according to the item feature matrix; The second correlation information determination unit is configured to determine second correlation information between the items and the polymer items based on the item feature matrix and the polymer item feature matrix.

19. A computer program comprising: Instructions, when executed by a processor, cause the processor to perform the method according to any one of claims 1-13.

20. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

Citation Information

Patent Citations

  • Article recommendation method and device based on user click, equipment and storage medium

    CN112148980A

  • Commodity recommendation method and device and computer equipment

    CN112837106A

  • Commodity recommendation method and device, equipment and storage medium

    CN114581190A

  • Recommendation method and device and storage medium

    CN114764471A

  • Personalized news recommendation method and system based on user global interest migration perception

    CN115481325A