Article cold start recommendation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses an article cold start recommendation method and device, electronic equipment and a storage medium, and relates to the technical field of data processing. The method comprises the steps of determining a to-be-recommended user data set and a to-be-recommended article data set; determining an embedded representation of each to-be-recommended article according to the to-be-recommended article data set; in response to an article recommendation instruction of a target user in the to-be-recommended user data set, determining a target behavior article corresponding to a target behavior of the target user, and generating a target behavior queue; determining the preference degree of the target user for each to-be-recommended article based on the embedded representation of each to-be-recommended article, the target behavior queue and a preset cold start condition; target to-be-recommended articles matched with the target user are determined based on the preference degree, and all the target to-be-recommended articles are recommended and displayed to the target user. According to the scheme provided by the embodiment of the invention, effective recommendation of cold start of the article can be realized, and the exposure degree of the new article is increased.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, electronic device and storage medium for cold start recommendation of items. Background Art

[0002] In today's era of information explosion, recommendation systems help users filter out valuable information and can effectively deal with the problem of information overload. They not only greatly improve users' online experience, but also bring more commercial value to enterprises, while promoting the progress of science and technology and society.

[0003] At present, the recommendation system mainly combines user behavior data to analyze and mine user interest preferences for accurate recommendations for thousands of people. In this process, due to the lack of interactive behavior data of new items, it is difficult for the system to accurately evaluate the user's interest in new items, which affects the exposure of new items and the discovery of high-quality items, resulting in poor recommendation results. At present, for the recommendation of new items (also known as cold start recommendation), methods based on item content similarity, methods based on knowledge graph relevance, or methods based on word2vec or BERT semantic vector similarity are usually used. However, these methods still have problems such as uncontrollable exposure of new items, the need to enhance the semantic relevance of items, and untimely behavioral feedback, which in turn affects the final recommendation effect.

[0004] How to achieve effective recommendations for cold-start items and increase the exposure of new items is a major research issue in the industry. Summary of the invention

[0005] The present invention provides a method, device, electronic device and storage medium for cold start recommendation of items, which can realize effective recommendation of cold start of items and increase the exposure of new items.

[0006] According to one aspect of the present invention, a method for cold start recommendation of items is provided, the method comprising:

[0007] Determine a user data set to be recommended and a data set of items to be recommended; each item to be recommended data includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended for the cold start items is less than a set number threshold;

[0008] Determine an embedding representation of each to-be-recommended item according to the to-be-recommended item dataset;

[0009] In response to an item recommendation instruction of a target user in the to-be-recommended user data set, determining a target behavior item corresponding to a target behavior of the target user, and generating a target behavior queue;

[0010] Determining the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue, and a preset cold start condition;

[0011] Based on the preference, target items to be recommended that match the target user are determined, and each of the target items to be recommended is recommended and displayed to the target user.

[0012] According to another aspect of the present invention, a device for cold-starting an item is provided, the device comprising:

[0013] The first determination module is used to determine a set of user data to be recommended and a set of item data to be recommended; each item data to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended for the cold start items is less than a set number threshold;

[0014] A second determination module, configured to determine an embedding representation of each to-be-recommended item according to the to-be-recommended item dataset;

[0015] A target behavior queue generation module, configured to respond to an item recommendation instruction of a target user in the to-be-recommended user data set, determine a target behavior item corresponding to a target behavior of the target user, and generate a target behavior queue;

[0016] A preference determination module, used to determine the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue and a preset cold start condition;

[0017] The target item to be recommended determining module is used to determine the target item to be recommended that matches the target user based on the preference degree, and recommend and display each of the target items to be recommended to the target user.

[0018] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0019] at least one processor; and

[0020] a memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the item cold start recommendation method described in any embodiment of the present invention.

[0022] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the item cold start recommendation method described in any embodiment of the present invention when executed.

[0023] According to another aspect of the present invention, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method for cold-start recommendation of items according to any embodiment of the present invention is implemented.

[0024] The technical solution of the embodiment of the present invention is to determine a user data set to be recommended and a data set of items to be recommended; each item data to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended of the cold start items is less than a set number threshold; the embedded representation of each item to be recommended is determined according to the data set of items to be recommended; in response to the item recommendation instruction of the target user in the user data set to be recommended, the target behavior item corresponding to the target behavior of the target user is determined, and a target behavior queue is generated; based on the embedded representation of each item to be recommended, the target behavior queue and the preset cold start conditions are determined to determine the preference of the target user for each item to be recommended; based on the preference, the target item to be recommended that matches the target user is determined, and each target item to be recommended is recommended and displayed to the target user, so as to achieve effective recommendation for cold start of items and increase the exposure of new items.

[0025] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 is a flow chart of a method for cold start recommendation of items provided according to the first embodiment of the present invention;

[0028] Figure 2 is a flow chart of a method for cold start recommendation of items provided according to Embodiment 2 of the present invention;

[0029] Figure 3 is a schematic diagram of the structure of a device for cold-starting an item according to a third embodiment of the present invention;

[0030] Figure 4 It is a schematic diagram of the structure of an electronic device for implementing the item cold start recommendation method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] Embodiment 1

[0034] Figure 1 This is a flowchart of a method for cold-starting an item recommendation according to Embodiment 1 of the present invention. This embodiment is applicable to the case of effectively recommending new items. The method can be executed by a device for cold-starting an item recommendation. The device for cold-starting an item recommendation can be implemented in the form of hardware and / or software. The device for cold-starting an item recommendation can be configured in an electronic device with data processing capabilities such as a computer, a server, or a tablet computer. Specifically, refer to Figure 1 , the method comprising:

[0035] Step 110: Determine a user data set to be recommended and an item data set to be recommended.

[0036] Each piece of data of the item to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended for the cold start items is less than a set number threshold.

[0037] Among them, the users to be recommended can be historical users of the target application, and the target application can be any application, such as a shopping application, a short video application, or an information application, etc.; it can be understood that the items involved in this embodiment can be new products, such as newly listed clothing, mobile phones or home appliances in a shopping application; they can also be newly published articles, short videos or mini programs, etc., which are not limited in this embodiment.

[0038] Optionally, in this embodiment, users who have logged into the target application within a preset time period (for example, the past week, the past month, or the past quarter, etc.) and have performed target behaviors in the target application can be determined as users to be recommended. The target behaviors can be: clicking, reading, purchasing, commenting, or collecting, etc., which are not limited in this embodiment; the user data to be recommended is data that matches the user to be recommended, which can include relevant information of the user to be recommended, such as a user name, user identification code, or user address, etc., which are not limited in this embodiment.

[0039] In this embodiment, the items to be recommended may be products recommended to the user, including new products, such as clothing, mobile phones, home appliances, articles, videos or mini-programs, etc., which are not limited in this embodiment; the data of the items to be recommended is data that matches the items to be recommended, which may include the title of the items to be recommended, the creation time and the number of recommendations, etc., which are not limited in this embodiment.

[0040] It can be understood that in this embodiment, the set of items to be recommended includes cold start items, that is, new items; the number of times the new items have been recommended is less than the set threshold, where the set threshold can be a smaller value such as 1 time, 2 times or 5 times, which is not limited in this embodiment.

[0041] In an optional implementation of the present embodiment, determining the user data set to be recommended and the item data set to be recommended may include: determining reference users who generate target behaviors within a first preset time period, and obtaining basic information of each of the reference users; generating the user data set to be recommended based on the basic information of each of the reference users; determining each item created within a second preset time period, and obtaining the title of each item and the number of times it has been recommended to form data of each item to be recommended; and forming the item data set to be recommended based on the data of each item to be recommended.

[0042] In this embodiment, the first preset time period and the second preset time period may be the same or different, and this is not limited in this embodiment; illustratively, the first preset time period may be the past day, the past week, or the past month, etc.; the second preset time period may be the past day, the past two days, or the past three days, etc.

[0043] The target behavior may include at least one of the following: click, collect, purchase, read, forward and comment; the basic information of the reference user may include the registration information, login information or delivery address of each reference user.

[0044] In an optional implementation of the present embodiment, users who have clicked, collected, purchased, read, forwarded or commented on the target application within a first preset time period can be determined, and these users can be determined as reference users; further, basic information of these reference users can be obtained; in the present embodiment, the basic information of a reference user can be combined into one piece of user data to be recommended; further, the formed user data to be recommended can be added to a preset storage location, such as a target folder or a target database, to obtain the user data set to be recommended involved in the present embodiment.

[0045] In another optional implementation of the present embodiment, each item created within a second preset time period can be determined, and the title and recommendation count of each item can be obtained. Furthermore, a piece of data on items to be recommended can be formed based on the creation time, title and recommendation count of the item; further, the formed data on each item to be recommended can be added to a preset storage location, such as a target folder or a target database, to obtain the data set of items to be recommended involved in the present embodiment.

[0046] It should be noted that, in this embodiment, the user information is obtained only after the user's authorization, and the method of obtaining is reasonable and legal.

[0047] Step 120: Determine an embedded representation of each to-be-recommended item according to the to-be-recommended item dataset.

[0048] In an optional implementation of the present embodiment, after obtaining the dataset of items to be recommended, the embedded representation of each item to be recommended can be further determined based on the dataset of items to be recommended; in the present embodiment, after obtaining the dataset of items to be recommended, the embedded representation of each item to be recommended can be determined based on the title of each item to be recommended in the dataset of items to be recommended; it can be understood that the embedded representation of each item to be recommended is the vector representation of each item to be recommended.

[0049] Optionally, in this embodiment, the embedded representation of each item to be recommended is determined based on the data set of items to be recommended, including: determining a target dimension of the embedded representation of each item to be recommended; inputting the title of each item to be recommended in the data set of items to be recommended and the target dimension into a pre-trained large language model to obtain an embedded representation of each item to be recommended.

[0050] The target dimension may be any dimension, for example, 300, 768 or 1024, etc., which is not limited in this embodiment. It is understandable that in order to facilitate the subsequent recommendation of the items to be recommended based on the embedded representation of each item to be recommended, the dimensions of the embedded representation of each item to be recommended in this embodiment are the same.

[0051] In an optional implementation of the present embodiment, the target dimension of the embedded representation of each of the items to be recommended may be determined first. Furthermore, the target dimension and the title of the target item to be recommended recorded in the data set of the items to be recommended may be input into a pre-trained large language model, and the embedded representation of the target item to be recommended may be output through the large language model.

[0052] Step 130: In response to the item recommendation instruction of the target user in the to-be-recommended user data set, determine the target behavior item corresponding to the target behavior of the target user, and generate a target behavior queue.

[0053] The target user may be any user in the to-be-recommended user data set, which is not limited in this embodiment.

[0054] In an optional implementation of this embodiment, after receiving an item recommendation instruction for a target user, a target behavior item corresponding to the target behavior of the target user may be further determined, and a target behavior queue may be generated based on relevant information of each target behavior item.

[0055] Optionally, in this embodiment, after receiving an item recommendation instruction for a target user, the items corresponding to the target user's click, collection, purchase, reading, forwarding or comment behavior can be further determined, and these items can be determined as target behavior items; further, the title, creation time or number of recommendations of each target behavior item can be determined, thereby obtaining a target behavior queue.

[0056] Exemplarily, the target behavior queue of the target user can be expressed as:

[0057] b u =[(bc1,bi1,bt1),(bc2,bi2,bt2),...,(bc n ,bi n , bt n )];

[0058] Among them, u is the target user, bc i is the specific behavior of the target user on item i; i Items corresponding to specific behaviors; bt i is the timestamp when a specific behavior occurs, i∈(1,n).

[0059] Step 140: Determine the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue, and a preset cold start condition.

[0060] In an optional implementation of the present embodiment, after obtaining the embedded representation of each item to be recommended in the data set of items to be recommended, the target user's preference for each item to be recommended can be further determined based on the embedded representation of each item to be recommended, the target behavior queue and the preset cold start condition.

[0061] In an optional implementation of this embodiment, for any item p in the to-be-recommended item data set, whether item p is a cold start can be determined based on the number of times it has been recommended and the difference between the creation time and the current time; illustratively, whether item p is a cold start can be expressed by the following formula:

[0062]

[0063] Among them, CUR_TS is the current timestamp, c p is the creation timestamp of item p, DELTA_TS is the timestamp difference threshold of the cold start item, rec_count p is the number of times item p has been recommended, REC_COUNT_THRESHOLD is the threshold of the number of times it has been recommended; it is understandable that is_cold p A value of 0 indicates that item p is not a cold start item, and a value of 1 indicates that it is a cold start item.

[0064] Optionally, in this embodiment, the preset cold start condition is determined by the following formula:

[0065]

[0066] Among them, s is the first item to be recommended, is_cold s =1 means the first recommended item is a cold start; is_cold s =0 means the first recommended item is not a cold start; α, β, γ are constants; CUR_TS is the current timestamp; c s The creation timestamp of the first recommended item; rec_count s The number of times the first item to be recommended has been recommended.

[0067] It can be understood that, in this embodiment, the more times the first to-be-recommended item is recommended, the smaller w(s) is.

[0068] In an optional implementation of the present embodiment, the target user's preference for each item to be recommended can be determined based on the embedded representation of each item to be recommended, the target behavior queue and a preset cold start condition; illustratively, in an example of the present embodiment, the similarity between the first item to be recommended and other items to be recommended can be determined based on the embedded representation of each item to be recommended; further, the weight of each target behavior in the target behavior queue can be determined; further, the target user's preference for each item to be recommended can be determined based on each similarity, the weight of each target behavior and the cold start condition.

[0069] Step 150: Determine target items to be recommended that match the target user based on the preference, and recommend and display each target item to the target user.

[0070] In an optional implementation of the present embodiment, after determining the target user's preference for each item to be recommended, the target items to be recommended that match the target user can be further determined based on each preference, and each target item to be recommended can be recommended and displayed to the target user.

[0071] Optionally, in this embodiment, after determining the target user's preference for each item to be recommended, the preference calculation results can be further sorted in a set order, and the target item to be recommended can be determined based on the sorting results; and the relevant data of the target item to be recommended can be sent to the target page for display.

[0072] The setting order may be from small to large or from large to small, which is not limited in this embodiment.

[0073] In an optional implementation of the present embodiment, after determining the target user's preference for each item to be recommended, the preference calculation results can be further sorted in ascending order, and the result ranked in a set position (for example, the last five or ten, etc., the specific recommendation position can be determined by the number of items actually recommended to the target user) is determined as the target recommended item; further, the relevant data of the target recommended item can be sent to the target page of the target application for display.

[0074] The technical solution of this embodiment is to determine a user data set to be recommended and a data set of items to be recommended; each item data to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended of the cold start items is less than a set number threshold; the embedded representation of each item to be recommended is determined according to the data set of items to be recommended; in response to the item recommendation instruction of the target user in the user data set to be recommended, the target behavior item corresponding to the target behavior of the target user is determined, and a target behavior queue is generated; based on the embedded representation of each item to be recommended, the target behavior queue and the preset cold start condition are used to determine the preference of the target user for each item to be recommended; based on the preference, the target item to be recommended that matches the target user is determined, and each target item to be recommended is recommended and displayed to the target user, so as to achieve effective recommendation for cold start of items and increase the exposure of new items.

[0075] Embodiment 2

[0076] Figure 2 is a flow chart of a method for cold start recommendation of items provided according to the second embodiment of the present invention. This embodiment is a further refinement of the above technical solution. The technical solution in this embodiment can be combined with each optional solution in one or more of the above embodiments. Figure 2 As shown, the method includes:

[0077] Step 210: Determine a user data set to be recommended and an item data set to be recommended.

[0078] Step 220: Determine an embedded representation of each to-be-recommended item according to the to-be-recommended item dataset and the title of each to-be-recommended item.

[0079] Step 230: Determine target-associated items associated with the target to-be-recommended item according to the embedded representation of each to-be-recommended item.

[0080] The target item to be recommended is any item to be recommended in the item to be recommended data set, which is not specifically limited in this embodiment.

[0081] In an optional implementation of the present embodiment, after obtaining the embedded representation of each item to be recommended in the data set of items to be recommended, each target-associated item associated with the target item to be recommended can be further determined based on the embedded representation of each item to be recommended. After obtaining the embedded representation of each item to be recommended in the data set of items to be recommended, the similarity between every two items to be recommended can be further determined based on the embedded representation of each item to be recommended, and each target-associated item associated with the target item to be recommended can be determined based on the similarity calculation result.

[0082] Optionally, in this embodiment, the embedded representation of the target item to be recommended can be respectively calculated for similarity with the embedded representations of other items to be recommended in the item dataset to be recommended, so as to determine the similarity between the target item to be recommended and other items to be recommended, obtaining n-1 similarity calculation results, where n is the number of items to be recommended in the item dataset to be recommended; further, the n-1 similarity calculation results can be sorted in descending order, and the top m (where m < n) items to be recommended are determined as the respective target associated items associated with the target item to be recommended.

[0083] The advantage of such a setting is that the items to be recommended in the item dataset to be recommended can be classified in advance, and subsequent recommendations for the target user can be made quickly, improving the execution efficiency of the algorithm.

[0084] Optionally, in this embodiment, for any two items p and q in the item dataset to be recommended, their embedded representations are respectively v p and v q , the semantic similarity sim pq between the two items can be calculated in the following manner, that is:

[0085]

[0086] In this embodiment, the number N of associated recommended items generated is set as a constant. For any item i in the item dataset to be recommended, the semantic similarity between other items to be recommended (except item i) in the item dataset to be recommended and i can be calculated, and they are sorted from high to low according to the similarity, and the top N items are taken as the associated recommendation results of i, denoted as where each item includes the item similar to i and the similarity value, where item1, item2,..., item N are all in the item dataset to be recommended.

[0087] Step 240: In response to the item recommendation instruction of the target user in the user dataset to be recommended, determine the target behavior item corresponding to the target behavior of the target user, and generate a target behavior queue.

[0088] Step 250: Based on the embedded representations of the respective items to be recommended, the target behavior queue, and a preset cold start condition, determine the preference degree of the target user for each of the items to be recommended.

[0089] In an optional implementation of the present embodiment, determining the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue, and the preset cold start condition may include: determining each first associated item to be recommended associated with the first item to be recommended based on the embedded representation of the first item to be recommended and the embedded representations of other items to be recommended in the recommended item data set; determining the behavior weight of each target behavior in the target behavior queue; determining the target user's preference for the first item to be recommended based on each similarity between the first item to be recommended and each of the first associated items to be recommended, each behavior weight, and the preset cold start condition; wherein the first item to be recommended matches the first item to be recommended data; and the preset cold start condition is determined by the following formula:

[0090]

[0091] Among them, s is the first item to be recommended, is_cold s =1 means the first recommended item is a cold start; is_cold s =0 means the first recommended item is not a cold start; α, β, γ are constants; CUR_TS is the current timestamp; c s The creation timestamp of the first recommended item; rec_count s The number of times the first item to be recommended has been recommended.

[0092] The first item to be recommended may be any item to be recommended in the item to be recommended data set, which is not limited in this embodiment.

[0093] In this embodiment, the first associated items to be recommended that are associated with the first item to be recommended may be determined based on step 230, which will not be described in detail herein.

[0094] Optionally, in this embodiment, different weights can be set in advance for different target behaviors. For example, the weight of click behavior is 1, the weight of collection behavior is 2, the weight of purchase behavior is 5, etc. In this embodiment, the behavior weight of the target behavior can be expressed by the following formula: bw=f(bi); wherein bw is the behavior weight of the target behavior, and bi is the target item targeted by the target behavior.

[0095] Furthermore, the target user's preference for the first item to be recommended may be determined based on the similarities between the first item to be recommended and each of the first associated items to be recommended, each of the behavior weights, and the preset cold start condition.

[0096] Optionally, in this embodiment, determining the target user's preference for the first item to be recommended based on the similarities between the first item to be recommended and each of the first associated items to be recommended, each of the behavior weights, and the preset cold start condition may include: multiplying each of the similarities, each of the behavior weights, and the cold start condition based on the following formula, and accumulating the product results to obtain the target user's preference for the first item to be recommended;

[0097]

[0098] Among them, bias up is the target user's preference for the first recommended item, u is the target user, p is the first recommended item, bc is the target behavior, f(bc) is the target behavior weight, bi is the target item for the target behavior, and bt is the timestamp of the target behavior; sim bip is the similarity between the target item and the first item to be recommended; rec_item bi is the associated target item associated with the target item; w(p) is the cold start condition of the first item to be recommended; g(bt) is the time decay function; the time decay function is determined by the following formula:

[0099]

[0100] Among them, g(x) is the time decay function; CUR_TS is the current timestamp; X is the adjustment coefficient, which is a constant; T is a positive integer.

[0101] Step 260: Determine target items to be recommended that match the target user based on the preference, and recommend and display each target item to be recommended to the target user.

[0102] The solution of this embodiment can determine each first associated item to be recommended associated with the first item to be recommended based on the embedded representation of the first item to be recommended and the embedded representations of other items to be recommended in the recommended item data set; determine the behavior weight of each target behavior in the target behavior queue; determine the target user's preference for the first item to be recommended based on the similarities between the first item to be recommended and each first associated item to be recommended, each behavior weight and the preset cold start condition. In the process of determining the user's preference for the item, the cold start condition is taken into account, so the cold start of the item can be achieved, the situation of insufficient recommendation of new items is avoided, and the exposure of new items is improved.

[0103] Embodiment 3

[0104] Figure 3 1 is a schematic diagram of the structure of a device for cold-starting recommendation of items provided according to the third embodiment of the present invention. Figure 3 As shown, the device includes: a first determination module 310, a second determination module 320, a target behavior queue generation module 330, a preference determination module 340 and a target to-be-recommended item determination module 350.

[0105] The first determination module 310 is used to determine a set of user data to be recommended and a set of item data to be recommended; each item data to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended for the cold start items is less than a set number threshold;

[0106] A second determination module 320, configured to determine an embedded representation of each to-be-recommended item according to the to-be-recommended item dataset;

[0107] A target behavior queue generation module 330 is used to determine a target behavior item corresponding to the target behavior of the target user in the to-be-recommended user data set in response to the item recommendation instruction of the target user, and generate a target behavior queue;

[0108] A preference determination module 340, configured to determine the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue, and a preset cold start condition;

[0109] The target item to be recommended determining module 350 is used to determine the target items to be recommended that match the target user based on the preference degree, and recommend and display each of the target items to be recommended to the target user.

[0110] The scheme of this embodiment is to determine the user data set to be recommended and the item data set to be recommended by the first determination module; each item data to be recommended includes: title, creation time and number of times recommended; the item set to be recommended includes cold start items, and the number of times recommended of the cold start items is less than the set number threshold; the embedded representation of each item to be recommended is determined by the second determination module according to the item data set to be recommended; the target behavior queue generation module responds to the item recommendation instruction of the target user in the user data set to be recommended, determines the target behavior item corresponding to the target behavior of the target user, and generates a target behavior queue; the preference determination module determines the target user's preference for each item to be recommended based on the embedded representation of each item to be recommended, the target behavior queue and the preset cold start condition; the target item determination module determines the target item to be recommended that matches the target user based on the preference, and recommends and displays each target item to be recommended to the target user, so as to achieve effective recommendation for cold start of items and increase the exposure of new items.

[0111] In an optional implementation of this embodiment, the first determination module 310 is specifically configured to determine reference users who generate target behaviors within a first preset time period, and obtain basic information of each of the reference users;

[0112] Generating the to-be-recommended user data set based on the basic information of each of the reference users;

[0113] Determine each item created within the second preset time period, and obtain the title and the number of times each item has been recommended, so as to form data of each item to be recommended;

[0114] The to-be-recommended item data set is formed based on each of the to-be-recommended item data.

[0115] In an optional implementation of this embodiment, the second determination module 320 is specifically configured to determine a target dimension of the embedded representation of each of the items to be recommended;

[0116] The title of each of the items to be recommended in the item to be recommended data set and the target dimension are input into a pre-trained large language model to obtain an embedded representation of each of the items to be recommended.

[0117] In an optional implementation of this embodiment, the target behavior queue generation module 330 is specifically used to obtain the target behavior generated by the target user within a third preset time period, and determine the items associated with the target behavior as the target behavior items;

[0118] generating the target behavior queue based on the target behavior item data of the target behavior item;

[0119] The target behavior includes at least one of the following: click, collect, and purchase.

[0120] In an optional implementation of this embodiment, the preference determination module 340 is specifically configured to determine first associated items to be recommended that are associated with the first item to be recommended based on the embedded representation of the first item to be recommended and the embedded representations of other items to be recommended in the recommended item dataset;

[0121] Determining a behavior weight of each target behavior in the target behavior queue;

[0122] Determining the target user's preference for the first item to be recommended based on the similarities between the first item to be recommended and each of the first associated items to be recommended, each of the behavior weights, and the preset cold start condition; wherein the first item to be recommended matches the first item to be recommended data;

[0123] The preset cold start condition is determined by the following formula:

[0124]

[0125] Among them, s is the first item to be recommended, is_cold s =1 means the first recommended item is a cold start; is_cold s =0 means the first item to be recommended is not a cold start; α, β, γ are constants; CUR_TS is the current timestamp; cs is the creation timestamp of the first item to be recommended; rec_count s The number of times the first item to be recommended has been recommended.

[0126] In an optional implementation of this embodiment, the preference determination module 340 is further specifically configured to multiply each of the similarities, each of the behavior weights, and the cold start condition based on the following formula, and accumulate the product results to obtain the target user's preference for the first item to be recommended;

[0127]

[0128] Among them, bias up is the target user's preference for the first recommended item, u is the target user, p is the first recommended item, bc is the target behavior, f(bc) is the target behavior weight, bi is the target item for the target behavior, and bt is the timestamp of the target behavior; sim bip is the similarity between the target item and the first item to be recommended; rec_item bi is the associated target item associated with the target item; w(p) is the cold start condition of the first item to be recommended; g(bt) is the time decay function; the time decay function is determined by the following formula:

[0129]

[0130] Among them, g(x) is the time decay function; CUR_TS is the current timestamp; X is the adjustment coefficient, which is a constant; T is a positive integer.

[0131] In an optional implementation of this embodiment, the target item to be recommended determining module 350 is specifically configured to sort the preference calculation results according to a set order, and determine the target item to be recommended based on the sorting results;

[0132] The relevant data of the target item to be recommended is sent to the target page for display.

[0133] The item cold start recommendation device provided in the embodiment of the present invention can execute the item cold start recommendation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0134] In the technical solutions of the embodiments of the present invention, the collection, storage, use, processing, transmission, provision and disclosure of user personal information (such as registration information, identity information, etc.) are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0135] Embodiment 4

[0136] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0137] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0138] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0139] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as an item cold-start recommendation method, which includes: determining a user data set to be recommended, and a data set of items to be recommended; each item data to be recommended includes: a title, a creation time, and a number of times it has been recommended; the set of items to be recommended includes cold-start items, and the number of times the cold-start items have been recommended is less than a set threshold; determining an embedded representation of each item to be recommended based on the data set of items to be recommended; in response to an item recommendation instruction of a target user in the user data set to be recommended, determining a target behavior item corresponding to a target behavior of the target user, and generating a target behavior queue; determining a preference of the target user for each item to be recommended based on the embedded representation of each item to be recommended, the target behavior queue, and a preset cold-start condition; determining a target item to be recommended that matches the target user based on the preference, and recommending and displaying each target item to be recommended to the target user.

[0140] In some embodiments, the item cold start recommendation method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the item cold start recommendation method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the item cold start recommendation method in any other appropriate manner (e.g., by means of firmware).

[0141] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0142] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0143] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0144] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0145] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0146] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0147] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0148] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for cold start recommendation of items, characterized in that: include: Determine the user data set to be recommended and the item data set to be recommended; Each item to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended for the cold start items is less than the set number threshold; Determine an embedding representation of each to-be-recommended item according to the to-be-recommended item dataset; In response to an item recommendation instruction of a target user in the to-be-recommended user data set, determining a target behavior item corresponding to a target behavior of the target user, and generating a target behavior queue; Determining the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue, and a preset cold start condition; Based on the preference, target items to be recommended that match the target user are determined, and each of the target items to be recommended is recommended and displayed to the target user.

2. The method for cold start recommendation of items according to claim 1, characterized in that: The step of determining the user dataset to be recommended and the item dataset to be recommended includes: Determine reference users who have generated target behavior within a first preset time period, and obtain basic information of each of the reference users; Generating the to-be-recommended user data set based on the basic information of each of the reference users; Determine each item created within the second preset time period, and obtain the title and the number of times each item has been recommended, so as to form data of each item to be recommended; The to-be-recommended item data set is formed based on each of the to-be-recommended item data.

3. The method for cold start recommendation of items according to claim 1, characterized in that: The step of determining the embedding representation of each item to be recommended according to the item data set to be recommended includes: Determining a target dimension of the embedding representation of each of the items to be recommended; The title of each of the items to be recommended in the item to be recommended data set and the target dimension are input into a pre-trained large language model to obtain an embedded representation of each of the items to be recommended.

4. The method for cold start recommendation of items according to claim 1, characterized in that: The determining of the target behavior items corresponding to the target behavior of the target user and generating a target behavior queue includes: Acquire a target behavior generated by the target user within a third preset time period, and determine an item associated with the target behavior as a target behavior item; generating the target behavior queue based on the target behavior item data of the target behavior item; The target behavior includes at least one of the following: click, collect, and purchase.

5. The method for cold start recommendation of items according to claim 1, characterized in that: The determining the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue, and a preset cold start condition includes: Determining first associated items to be recommended that are associated with the first item to be recommended based on the embedded representation of the first item to be recommended and the embedded representations of other items to be recommended in the recommended item dataset; Determining a behavior weight of each target behavior in the target behavior queue; Determining the target user's preference for the first item to be recommended based on the similarities between the first item to be recommended and each of the first associated items to be recommended, each of the behavior weights, and the preset cold start condition; wherein the first item to be recommended matches the first item to be recommended data; The preset cold start condition is determined by the following formula: Among them, s is the first item to be recommended, is_cold s =1 means the first recommended item is a cold start; is_cold s =0 means the first recommended item is not a cold start; α, β, γ are constants; CUR_TS is the current timestamp; c s The creation timestamp of the first recommended item; rec_count s The number of times the first item to be recommended has been recommended.

6. The method for cold start recommendation of items according to claim 5, characterized in that: The determining the target user's preference for the first item to be recommended based on the similarities between the first item to be recommended and each of the first associated items to be recommended, each of the behavior weights, and the preset cold start condition includes: Based on the following formula, each of the similarities, each of the behavior weights, and the cold start condition are multiplied respectively, and the product results are accumulated to obtain the preference of the target user for the first item to be recommended; Among them, bias up is the target user's preference for the first recommended item, u is the target user, p is the first recommended item, bc is the target behavior, f(bc) is the target behavior weight, bi is the target item for the target behavior, and bt is the timestamp of the target behavior; sim bip is the similarity between the target item and the first item to be recommended; rec_item bi is the associated target item associated with the target item; w(p) is the cold start condition of the first item to be recommended; g(bt) is the time decay function; the time decay function is determined by the following formula: Among them, g(x) is the time decay function; CUR_TS is the current timestamp; X is the adjustment coefficient, which is a constant; T is a positive integer.

7. The method for cold start recommendation of items according to claim 5, characterized in that: The determining the target items to be recommended that match the target user based on the preference, and recommending and displaying each of the target items to be recommended to the target user, includes: Sort the calculation results of each preference according to the set order, and determine the target item to be recommended based on the sorting results; The relevant data of the target item to be recommended is sent to the target page for display.

8. An item cold start recommendation device, characterized in that: include: A first determination module is used to determine a user data set to be recommended and a data set of items to be recommended; Each item to be recommended includes: title, creation time and number of times recommended; the set of items to be recommended includes cold start items, and the number of times recommended for the cold start items is less than the set number threshold; A second determination module, configured to determine an embedding representation of each to-be-recommended item according to the to-be-recommended item dataset; A target behavior queue generation module, configured to respond to an item recommendation instruction of a target user in the to-be-recommended user data set, determine a target behavior item corresponding to a target behavior of the target user, and generate a target behavior queue; A preference determination module, used to determine the target user's preference for each of the items to be recommended based on the embedded representation of each of the items to be recommended, the target behavior queue and a preset cold start condition; The target item to be recommended determining module is used to determine the target item to be recommended that matches the target user based on the preference degree, and recommend and display each of the target items to be recommended to the target user.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the item cold start recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the item cold start recommendation method according to any one of claims 1 to 7 when executed.

Images