A material recommendation method, device, electronic device and computer-readable medium

Abstract

Embodiments of the present application provide a method, device, electronic device, and computer-readable medium for material recommendation, including: determining a first proportion and a second proportion of a target material corresponding to a specific tag in a material recommendation process within a historical time period, wherein the first proportion is the ratio of the exposure amount of the target material issued based on a first recall method to the total exposure amount of the target material, and the second proportion is the ratio of the exposure amount of the target material issued based on a second recall method to the total exposure amount of the target material; when the first proportion and the second proportion satisfy a set proportional relationship, determining a weight coefficient of the second recall method according to the first proportion and the second proportion; and performing material recommendation for the target material based on the first recall method, the second recall method, and the weight coefficient.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a material recommendation method, device, electronic device, and computer-readable medium. Background Art

[0002] Material recall aims to filter out content that users may be interested in from millions of materials and recommend the filtered materials to users. Material recall includes model recall and strategy recall. Model recall refers to inputting materials into the recommendation model, and the recommendation model outputs the recommendation level of the material. However, in some specific scenarios, such as sudden hot events, due to the small number of actual exposure samples of hot materials, the recommendation model has not accumulated enough samples, and it is difficult to accurately present candidate materials using only model recall. In this scenario, it is necessary to design a hot strategy related to the hot event, and adopt a combination of model recall and hot strategy recall. Hot strategy recall uses weighted processing to improve the recommendation level of hot materials, so that the exposure of hot materials is increased.

[0003] However, while forcibly using a weighted hotspot strategy to increase the exposure of hotspot materials will also reduce performance indicators. After the popularity of a hot event decreases, manual intervention is still required to exit the relevant hotspot strategy. The manual exit and recall method of the hotspot strategy relies on the experience of technical personnel, has low accuracy and requires a lot of manpower costs. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide a method, device, electronic device and computer-readable medium for material recommendation, which can improve the efficiency index of material recommendation in specific scenarios.

[0005] To solve the above technical problems, the embodiments of the present application are implemented through the following aspects.

[0006] In the first aspect, an embodiment of the present application provides a method for material recommendation, including: determining a first proportion and a second proportion of a target material corresponding to a specific label in a material recommendation process within a historical time period, the first proportion being the ratio of the exposure amount of the target material issued based on a first recall method to the total exposure amount of the target material, and the second proportion being the ratio of the exposure amount of the target material issued based on a second recall method to the total exposure amount of the target material; when the first proportion and the second proportion satisfy a set proportional relationship, determining a weight coefficient of the second recall method according to the first proportion and the second proportion; and making material recommendations for the target material based on the first recall method, the second recall method and the weight coefficient.

[0007] In second aspect, an embodiment of the present application provides a material recommendation device, comprising: a first determination module, for determining a first proportion and a second proportion of a target material corresponding to a specific label in a material recommendation process within a historical time period, the first proportion being the ratio of the exposure amount of the target material issued based on the first recall method to the total exposure amount of the target material, and the second proportion being the ratio of the exposure amount of the target material issued based on the second recall method to the total exposure amount of the target material; a second determination module, for determining a weight coefficient of the second recall method according to the first proportion and the second proportion when the first proportion and the second proportion satisfy a set proportional relationship; a recommendation module, for making material recommendations for the target material based on the first recall method, the second recall method and the weight coefficient.

[0008] In a third aspect, an embodiment of the present application provides an electronic device comprising: a memory, a processor, and computer executable instructions stored on the memory and executable on the processor, wherein the computer executable instructions, when executed by the processor, implement the material recommendation method described in the first aspect above.

[0009] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which is used to store computer-executable instructions. When the computer-executable instructions are executed by a processor, they can implement the material recommendation method described in the first aspect above.

[0010] In an embodiment of the present application, by determining the first proportion and second proportion of the target material corresponding to a specific label in the material recommendation process within a historical time period, the first proportion is the ratio of the exposure of the target material based on the first recall method to the total exposure of the target material, and the second proportion is the ratio of the exposure of the target material based on the second recall method to the total exposure of the target material; when the first proportion and the second proportion meet the set proportional relationship, the weight coefficient of the second recall method is determined according to the first proportion and the second proportion; based on the first recall method, the second recall method and the weight coefficient, the target material is recommended, and the recall method adopted can be automatically adjusted according to the historical exposure of the target material corresponding to the specific label under the two recall methods, which not only ensures that the target material has sufficient exposure in the initial stage, but also ensures that the second recall method can be automatically withdrawn in the later stage, thereby improving the efficiency index of material recommendation in specific scenarios, reducing the labor cost of material recommendation, and improving the accuracy of material recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0012] Figure 1 A schematic diagram showing a flow chart of a material recommendation method provided in an embodiment of the present application;

[0013] Figure 2 Another schematic flow chart illustrating a material recommendation method provided in an embodiment of the present application;

[0014] Figure 3 A schematic diagram showing the structure of a material recommendation device provided in an embodiment of the present application is shown;

[0015] Figure 4 A schematic diagram of the hardware structure of an electronic device for executing a material recommendation method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0017] As mentioned previously, when using strategic recall to recommend materials, business strategies are relatively fixed, often tied to company or departmental plans, resulting in long iteration cycles and slow speeds. However, hot materials are characterized by immediacy, widespread public attention, and fluctuating timing and duration. Therefore, these materials are difficult to adapt to existing strategic recall methods.

[0018] On the other hand, hot topics that require high immediacy often significantly increase user visits and engagement. However, in the early stages of a hot topic, due to the small number of real exposure samples, it is difficult to accumulate sufficient samples for the physical examination model. Consequently, the recommendation model does not fully learn the content, resulting in low scores for these content. This makes the content less competitive and lacks penetration during the recall, promotion, and refinement stages of content recommendation. Therefore, in order to present hot topics to users, it is usually necessary to design a hot topic strategy related to the hot topic to assist in recommending these content.

[0019] Hotspot strategies related to hot events (also called weighted strategies) may reduce performance indicators such as user interaction rate, user stay time, user click-through rate, average refresh rate per person, average exposure rate per person, and next-day retention rate while increasing material exposure. In addition, it is difficult to determine the cycle of the hotspot strategy, such as when to start and when to exit the hotspot strategy for hot materials. To determine the cycle of the hotspot strategy, it is necessary to conduct preliminary data analysis, development and launch of the strategy code, ab experiments with small traffic, and manual full-scale launch. It is also necessary to exit the relevant strategy after the popularity of the hot event decreases, and repeat the above-mentioned analysis, experimentation, reduction, and other complex steps before exiting the relevant strategy. If the relevant strategy is not exited at the right time, it will affect the performance indicators and user experience.

[0020] In view of this, the present application proposes a material recommendation method to overcome the above problems.

[0021] Figure 1 A schematic flow chart of a material recommendation method provided in an embodiment of the present application is shown. The method can be executed by an electronic device, such as a server-side device. In other words, the method can be executed by software or hardware installed on the server-side device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0022] This method can be applied to information flow recommendation systems. In such systems, recommending items can include the following steps: an item recall step, used to initially screen items for recommendation; an item rough ranking and re-ranking step, used to further filter out thousands and / or hundreds of highly recommended items from these initial screenings; and a distribution exposure step, used to determine dozens of items from these further screened items for distribution to the client. Optionally, this method can be used in both the item rough ranking and re-ranking steps.

[0023] As shown in the figure, the method may include the following steps.

[0024] Step S110: Determine a first proportion and a second proportion.

[0025] The first proportion and the second proportion are the proportions of the target materials corresponding to the specific tags in the material recommendation process within the historical time period. Specifically, the first proportion is used to indicate the ratio of the exposure of the target material based on the first recall method to the total exposure of the target material within the historical time period, and the second proportion is used to indicate the ratio of the exposure of the target material based on the second recall method to the total exposure of the target material within the historical time period. Optionally, the historical time period is a time period of any length before the current time, such as the past hour. Specific tags used to classify materials are, for example, event tags, topic tags, and character tags. The first recall method and the second recall method are, for example, one of rule recall, collaborative filtering recall, and model recall, and the first recall method is different from the second recall method. The exposure of a material is, for example, the number of materials sent to the client.

[0026] Through this step, the historical recall status of the target material can be determined.

[0027] Step S120: Determine the weight coefficient of the second recall method.

[0028] If the first and second proportions satisfy a predetermined proportional relationship, a weight coefficient for the second recall method is determined based on the first and second proportions. The weight coefficient is used to adjust the proportion of the second recall method. In other words, the weight coefficient is used to adjust the recall method for the target material, for example, reducing the use of the second recall method and increasing the use of the first recall method. The weight coefficient is greater than 0 and less than 1. Optionally, the weight coefficient can be the presence coefficient or the withdrawal coefficient of the second recall method.

[0029] Through this step, the recall method can be adjusted in a timely manner according to the historical recall of the target material, so as to adopt a recall method that better meets the usage needs.

[0030] Step S130: recommending the target material.

[0031] Based on the first recall method, the second recall method and the weight coefficient, target materials are recommended.

[0032] In an embodiment of the present application, by determining the first proportion and second proportion of the target material corresponding to a specific label in the material recommendation process within a historical time period, the first proportion is the ratio of the exposure of the target material based on the first recall method to the total exposure of the target material, and the second proportion is the ratio of the exposure of the target material based on the second recall method to the total exposure of the target material; when the first proportion and the second proportion meet the set proportional relationship, the weight coefficient of the second recall method is determined according to the first proportion and the second proportion; based on the first recall method, the second recall method and the weight coefficient, the target material is recommended, and the recall method adopted can be automatically adjusted according to the historical exposure of the target material corresponding to the specific label under the two recall methods, which not only ensures that the target material has sufficient exposure in the initial stage, but also ensures that the second recall method can be automatically withdrawn in the later stage, thereby improving the efficiency index of material recommendation in specific scenarios, reducing the labor cost of material recommendation, and improving the accuracy of material recommendation.

[0033] In one possible implementation, determining the first and second proportions of target materials corresponding to specific tags in the material recommendation process within a historical time period includes: according to a set statistical frequency, when each statistical moment arrives, determining the first and second proportions of target materials corresponding to specific tags in the material recommendation process within the historical time period corresponding to the current statistical moment.

[0034] For example, if the historical time period is 1 hour and the statistical frequency is set to every 5 minutes, the first and second percentages of the target material in the past hour will be collected every 5 minutes.

[0035] In a possible implementation, the set proportional relationship includes: the sum of the first proportion and the second proportion is greater than a set proportion threshold, and the ratio between the first proportion and the second proportion is greater than or equal to a set ratio threshold.

[0036] The sum of the first and second proportions is greater than a preset proportion threshold, such as 50%. By setting this proportion threshold, the corresponding adjustment strategy can be triggered only when there is a need to adjust the recall method. Table 1 below shows a possible example of the proportion relationship.

[0037] Table 1

[0038]

[0039] It should be noted that the determined weight coefficient is 0, which indicates that the weight of the second recall method of the material will not be adjusted. In other words, the logic of the second recall method adjustment may only take effect for materials with a weight coefficient not equal to 0.

[0040] Determining the weight coefficient of the second recall method based on the first proportion and the second proportion includes: determining the weight coefficient corresponding to the ratio interval in which the ratio is located based on the ratio between the first proportion and the second proportion, and the correspondence between the set ratio interval and the weight coefficient.

[0041] As shown in Table 1, for example, if the ratio of the first percentage to the second percentage falls within the first interval, this indicates that the number of items exposed through the first recall method is close to or exceeds the number of items exposed through the second recall method. The first interval can be [0.9, +) or any other smaller interval within this range. The interval range can be adjusted based on historical values, experimental results, or user needs. For example, if the first interval in Table 1 is [0.9, 1.2), the weight coefficient is determined to be 0.96; if the second interval is [1.2, 1.6), the weight coefficient is determined to be 0.92.

[0042] When the ratio of the first proportion to the second proportion meets different ratios, the determined weight coefficient is different. Optionally, the correspondence between the ratio of the first proportion to the second proportion and the weight coefficient can be set with reference to the results of a control experiment. When conducting a control experiment, different weight coefficients can be determined for several experimental groups with the same ratio of the first proportion to the second proportion. The optimal solution can be determined based on parameters such as the number of material exposures and material ranking before and after the second recall method is adjusted.

[0043] In one possible implementation, the correspondence includes at least two ratio intervals, and the numerical ranges covered by each ratio interval do not overlap. The smaller the numerical value covered by the ratio interval, the larger the corresponding weight coefficient, and the weight coefficient is used to characterize the exit weight of the second recall method.

[0044] In other words, you can also set a third interval and a fourth interval and determine the corresponding weight coefficients. For example, if the third interval is [1.6, 2.0), the corresponding weight coefficient is 0.88; if the fourth interval is [2.0, +), the corresponding weight coefficient is 0.84.

[0045] In a possible implementation, when the first proportion and the second proportion do not satisfy a set proportional relationship, the current recall mode is maintained and a material recommendation is performed for the target material.

[0046] In one possible implementation, the specific tag includes multiple hot event tags, the first recall method includes model recall, and the second recall method includes hot strategy recall corresponding to each hot event tag; the determination of the first proportion and the second proportion of the target material corresponding to the specific tag in the material recommendation process includes: determining the first proportion and the second proportion corresponding to each hot event tag in the historical time period; the first proportion is the ratio of the exposure of the target material corresponding to the hot event tag based on the model recall to the total exposure of the target material, and the second proportion is the ratio of the exposure of the target material corresponding to the hot event tag based on the corresponding hot strategy recall to the total exposure of the target material. The material recommendation for the target material based on the first recall method, the second recall method and their weight coefficients includes: for any hot event tag, based on the model recall, the hot strategy recall corresponding to the hot event tag and its weight coefficient, the material recommendation for the target material corresponding to the hot event tag is performed.

[0047] Hot events are sudden, and related material samples are relatively rare, making it difficult to provide sufficient hot event material samples for model learning. Therefore, in the early stages of a hot event, the recommendation level of hot event materials determined by model recall is often inaccurate. In this case, model recall is combined with hot event strategy recall, such as weighting hot events, to increase the exposure of materials and accumulate samples for model learning. For example, in the early stages of a hot event, the model determines the recommendation level of a material as 5 points. By combining it with strategy recommendation and weighting it by 200%, the recommendation level of the material reaches 10 points. This allows the material to be ranked higher in subsequent screening stages.

[0048] Multiple hot event tags for the target item can be obtained from the link, such as the top 10 and top 50 search tags. By determining the primary and secondary percentages and weight coefficients corresponding to each tag, when using policy recall for weighting, items with the same tag can be given the same weight. When executing a policy exit, exit can be performed for only a subset of multiple tags, for example, exiting the top 10 tags. This increases the flexibility of policy exit.

[0049] In a possible implementation, the model recall includes at least one of the following: Mid collaborative recall, new target recall, deep recall, factor machine (FM) recall, and user collaborative recall.

[0050] In the embodiment of the present application, in the early stage of a hot event, the hot strategy takes effect normally, and the exposure ratio of the hot strategy recall is high, which helps to accumulate enough samples for model learning; in the middle and late stages of a hot event, when it is calculated that the exposure ratio of the hot event launched by the model recall catches up with or exceeds the hot strategy recall, the hot strategy recall and the related hot strategy tags are automatically exited. This solution not only ensures that the hot materials in the early stage of the hot event have sufficient exposure and provide sufficient learning samples for the model, but also ensures that the strategy recall and strategy tag can be automatically exited in the later stage of the hot spot, relying on the model itself to launch hot materials, so as to no longer have a negative impact on the performance indicators. Using the material recommendation method of the embodiment of the present application, on the one hand, the weight coefficient can be flexibly set, and on the other hand, the strategy exit can be gradually executed according to the decreasing weight coefficient, providing a better user experience.

[0051] Figure 2 Another flow chart illustrating the material recommendation method provided in an embodiment of the present application is shown.

[0052] Referring to step 301, at the beginning of the outbreak of a hot event, 1 / 1000 material sampling is performed based on the thousandth sampling table according to online requests. Sampling covers all stages of material recommendation from coarse sorting, fine sorting, re-ranking to distribution exposure. The objects sampled by the thousandth table include the newly added event identifier Event_id, fine sorting score ranking, final ranking, whether it is a hot field, etc. The sampling results can be used to analyze in real time the proportion of material exposure achieved by model recall and material exposure achieved by strategy recall in the final total distribution exposure. For example, the proportion of model recall exposure and strategy recall exposure of the hot event materials corresponding to each Event_id in the distribution exposure in the past hour is calculated every five minutes. Here, strategy recall mainly refers to the hot strategy for hot events.

[0053] When the model recall exposure ratio and the strategy recall exposure ratio meet a certain ratio, a strategic exit is executed for the item, including tag exit and all strategy exit. If the sum of the model recall exposure ratio and the strategy recall exposure ratio is greater than 50%, and the ratio of the model recall exposure ratio to the strategy recall exposure ratio meets different ratios, the weight coefficient is determined differently. Optionally, ClickHouse performs the above metric calculations. ClickHouse's calculated metrics also include final ranking and refined ranking.

[0054] As shown in step 302, the recommendation engine reads the AB strategy switch and data from Redis. If the AB switch is on and the event_id exit weight is non-zero, the strategy exit logic takes effect. The corresponding event is recommended by the model, with strategy support. For items not recalled using the hotspot strategy, exit is not executed.

[0055] Assume that in the initial stages of a hot event, the weight corresponding to the model recall method is 1, and the weight corresponding to the hot strategy recall method is 2. When the strategy exit logic takes effect, the weight of the hot strategy method is adjusted by introducing a weight coefficient (such as the exit weight). If the final weight of the item after the strategy exit is executed (that is, the sum of the weight corresponding to the model recall method and the weight corresponding to the hot strategy recall method) is less than 1, the final weight can be adjusted to 1.

[0056] In one possible implementation, the method further includes storing the results such as exposure, proportion, weight coefficient in a non-relational database, such as a redis database, in a key-value pair format. When storing, overall data and grouped data can be stored. Among them, the key value of the overall data is event_id, and the value value includes information such as the model exposure proportion, the strategy exposure proportion, the total exposure amount issued, the model exposure amount issued, and the strategy exposure amount issued. The ratio of the model exposure proportion and the strategy exposure proportion is used to design the exit weight, and the other value values are used for real-time monitoring. The key value of the grouped data is experimental group + event_id, and the value value includes information such as the exposure proportion of this event_id in this experimental group, the total exposure amount issued by this experimental group, the number of exposures of this event_id in this experimental group, and the exit weight of this event_id, which is used for grouped real-time monitoring. The experimental group is the grouping during the ab control experiment.

[0057] For example, you can monitor changes in the model exposure share and strategy exposure share of a certain Event_id to trigger the strategy exit mechanism; you can monitor changes in the exposure share of a certain Event_id in each experimental group in the past hour to determine whether the exposure of hot materials has dropped sharply after the hot strategy exit mechanism takes effect, to prevent the exit strategy from being too strong and affecting the exposure share of hot materials; you can monitor changes in the exit weight of a certain Event_id in each experimental group to monitor whether the hot exit strategy is effective and changes in the exit weight value of each experimental group; you can also monitor the exposure share ranking and model score ranking of a certain Event_id.

[0058] Through real-time monitoring of indicators, when abnormal indicator value changes occur after the execution of strategy exit, timely adjustments can be made to ensure the stable operation of the strategy exit mechanism.

[0059] In an embodiment of the present application, by real-time monitoring of the exposure ratios of the model recall and the strategy recall for materials with the same hot event label over a period of time, it is determined whether the model recall has been fully learned and can push out the relevant materials in the final released materials. Since the strategy recall method is relatively "rigid", if the number of such hot materials recalled by the model can exceed the hot strategy recall and the re-arrangement business strategy, the hot strategy recall and the re-arrangement business strategy will be weakened, allowing the model to take the lead in launching such hot materials. On the contrary, material recommendations are made in the form of a strategy-assisted model. Therefore, the complementarity of the model recall and the strategy recall method can be achieved, solving the problems of the current hot material business strategy, such as low performance indicators, insufficient timeliness, frequent manual modification and offline of business strategies, and the need for additional offline analysis.

[0060] The material recommendation method provided by the embodiment of the present application has a higher degree of automation, and does not require frequent changes to online traffic and policy codes. It integrates offline analysis of business policy experiments into automated online analysis of hot spot data for a certain period of time, and returns the exit weight of "whether the hot spot policy should be exited" for use by the recommendation system. It not only ensures that hot spots can be launched by strategies in the early stage of occurrence due to insufficient model training, but also ensures that after the model has fully learned the samples, the model itself will launch hot spot materials, thereby reducing the impact on performance indicators such as user stay time, user click-through rate, and average exposure rate. In addition, the solution has strong scalability and can be modified accordingly according to different business development stages.

[0061] Figure 3 A schematic structural diagram of a material recommendation device provided in an embodiment of the present application is shown. The device 400 includes: a first determination module 410, a second determination module 420 and a recommendation module 430.

[0062] The first determination module 410 is used to determine the first proportion and the second proportion of the target material corresponding to the specific label in the material recommendation process within a historical time period, wherein the first proportion is the ratio of the exposure of the target material based on the first recall method to the total exposure of the target material, and the second proportion is the ratio of the exposure of the target material based on the second recall method to the total exposure of the target material; the second determination module 420 is used to determine the weight coefficient of the second recall method according to the first proportion and the second proportion when the first proportion and the second proportion meet the set proportional relationship; the recommendation module 430 is used to make material recommendations for the target material based on the first recall method, the second recall method and the weight coefficient.

[0063] In one possible implementation, the first determination module 410 is specifically configured to determine, according to a set statistical frequency, at each statistical moment, a first proportion and a second proportion of the target material corresponding to the specific tag in the material recommendation process within a historical time period corresponding to the current statistical moment.

[0064] In one possible implementation, the set proportional relationship includes: the sum of the first proportion and the second proportion is greater than a set proportion threshold, and the ratio between the first proportion and the second proportion is greater than or equal to the set ratio threshold, and the second determination module 420 is specifically used to determine the weight coefficient corresponding to the ratio interval in which the ratio is located based on the ratio between the first proportion and the second proportion, and the correspondence between the set ratio interval and the weight coefficient.

[0065] In one possible implementation, the correspondence includes at least two ratio intervals, and the numerical ranges covered by each ratio interval do not overlap. The smaller the numerical value covered by the ratio interval, the larger the corresponding weight coefficient, and the weight coefficient is used to characterize the exit weight of the second recall method.

[0066] In a possible implementation, the device is configured to maintain the current recall mode and make a material recommendation for the target material when the first proportion and the second proportion do not satisfy a set proportional relationship.

[0067] In one possible implementation, the specific tag includes multiple hot event tags, the first recall method includes model recall, and the second recall method includes hot strategy recall corresponding to each hot event tag. The first determination module 410 is specifically used to determine the first proportion and the second proportion corresponding to each hot event tag in the historical time period; the first proportion is the ratio of the exposure of the target material corresponding to the hot event tag based on the model recall to the total exposure of the target material, and the second proportion is the ratio of the exposure of the target material corresponding to the hot event tag based on the corresponding hot strategy recall to the total exposure of the target material. In a possible implementation, the second determination module 420 is specifically used to make material recommendations for the target material corresponding to any hot event tag based on the model recall, the hot strategy recall corresponding to the hot event tag and its weight coefficient.

[0068] In a possible implementation, the model recall includes at least one of the following: Mid collaborative recall, new target recall, deep recall, FM recall, and user collaborative recall.

[0069] The device 400 provided in the embodiment of the present application can execute the various methods described in the above method embodiments and realize the functions and beneficial effects of the various methods described in the above method embodiments, which will not be repeated here.

[0070] Figure 4 A schematic diagram of the hardware structure of an electronic device that executes a material recommendation method provided in an embodiment of the present application is shown. Referring to the figure, at the hardware level, the electronic device includes a processor 510, and optionally, an internal bus 520, a network interface 530, and a memory. Among them, the memory may include a memory 540, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory) 550, such as at least one disk storage, etc. Of course, the electronic device may also include hardware required for other services.

[0071] The processor 510, the network interface 530, and the memory can be interconnected via an internal bus 520. This internal bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Such buses can be classified as address buses, data buses, control buses, and the like. For ease of illustration, only one bidirectional arrow is used in this figure, but this does not imply that there is only one bus or only one type of bus.

[0072] The memory is used to store programs. Specifically, the program may include program code, which includes computer operating instructions. The memory may include internal memory 540 and non-volatile memory 550, and provides instructions and data to the processor 510.

[0073] The processor 510 reads the corresponding computer program from the non-volatile memory 550 into the memory 540 and then runs it, forming a device for locating the target user at the logical level. The processor 510 executes the program stored in the memory and is specifically used to execute Figures 1 to 2 The method described in the embodiment can achieve the same or corresponding technical effects.

[0074] The above application Figures 1 to 2The methods disclosed in the illustrated embodiments can be applied to a processor or implemented by processor 510. Processor 510 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 510 or software instructions. The above-mentioned processor 510 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in conjunction with the embodiments of this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor 510 reads the information in the memory and performs the steps of the above method in conjunction with its hardware.

[0075] The electronic device can also execute the methods described in the above method embodiments and realize the functions and beneficial effects of the methods described in the above method embodiments, which will not be described in detail here.

[0076] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0077] The embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, which, when executed by an electronic device including multiple application programs, enables the electronic device to execute Figures 1 to 2 The method described in the embodiment can achieve the same or corresponding technical effects.

[0078] The computer-readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0079] Furthermore, the embodiment of the present application also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, when the program instructions are executed by a computer, Figures 1 to 2 The method described in the embodiment can achieve the same or corresponding technical effects.

[0080] In short, the above description is only a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0081] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0082] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0083] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0084] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

Claims

1. A material recommendation method, characterized in that: include: Determine a first proportion and a second proportion of a target material corresponding to a specific tag in a material recommendation process within a historical time period, where the first proportion is a ratio of exposures of the target material issued based on a first recall method to the total exposures of the target material, and the second proportion is a ratio of exposures of the target material issued based on a second recall method to the total exposures of the target material; When the first proportion and the second proportion satisfy a set proportional relationship, determining a weight coefficient of the second recall method according to the first proportion and the second proportion; Based on the first recall method, the second recall method and the weight coefficient, a material recommendation is performed for the target material.

2. The method according to claim 1, wherein Determine the first and second proportions of target materials corresponding to specific tags in the material recommendation process within a historical period, including: According to the set statistical frequency, when each statistical moment arrives, the first and second proportions of the target material corresponding to the specific tag in the material recommendation process within the historical time period corresponding to the current statistical moment are determined.

3. The method according to claim 1, wherein The set proportional relationship includes: the sum of the first proportion and the second proportion is greater than a set proportion threshold, and the ratio between the first proportion and the second proportion is greater than or equal to a set ratio threshold; The determining of the weight coefficient of the second recall method according to the first proportion and the second proportion includes: According to the ratio between the first proportion and the second proportion, and the correspondence between the set ratio interval and the weight coefficient, the weight coefficient corresponding to the ratio interval in which the ratio is located is determined.

4. The method according to claim 3, wherein: The corresponding relationship includes at least two ratio intervals, and the numerical ranges covered by each ratio interval do not overlap. The smaller the numerical value covered by the ratio interval, the larger the corresponding weight coefficient. The weight coefficient is used to characterize the exit weight of the second recall method.

5. The method according to claim 1, wherein Also includes: When the first proportion and the second proportion do not satisfy the set proportional relationship, the current recall mode is maintained and material recommendation is performed for the target material.

6. The method according to claim 1, wherein The specific tag includes a plurality of hot event tags, the first recall method includes model recall, and the second recall method includes hot strategy recall corresponding to each hot event tag; Determining the first proportion and the second proportion of the target material corresponding to the specific tag in the material recommendation process includes: Determine the first proportion and second proportion corresponding to each hot event label within the historical time period; the first proportion is the ratio of the exposure of the target material corresponding to the hot event label based on the model recall to the total exposure of the target material, and the second proportion is the ratio of the exposure of the target material corresponding to the hot event label based on the corresponding hot strategy recall to the total exposure of the target material.

7. The method according to claim 6, wherein: The recommending the target material based on the first recall method, the second recall method and the weight coefficient thereof includes: For any hot event tag, based on the model recall, the hot strategy recall corresponding to the hot event tag and its weight coefficient, material recommendations are made for the target material corresponding to the hot event tag.

8. The method according to claim 6, wherein: The model recall includes at least one of the following: Mid collaborative recall, new target recall, deep recall, FM recall and user collaborative recall.

9. A material recommendation device, characterized in that: include: A first determination module is configured to determine a first proportion and a second proportion of a target material corresponding to a specific tag in a material recommendation process within a historical time period, wherein the first proportion is a ratio of exposures of the target material issued based on the first recall method to the total exposures of the target material, and the second proportion is a ratio of exposures of the target material issued based on the second recall method to the total exposures of the target material; a second determining module, configured to determine a weight coefficient of the second recall method according to the first proportion and the second proportion when the first proportion and the second proportion satisfy a set proportional relationship; A recommendation module is used to recommend the target material based on the first recall method, the second recall method and the weight coefficient.

10. An electronic device comprising: processor; as well as A memory arranged to store computer-executable instructions, wherein when the instructions are executed, the processor performs the material recommendation method according to any one of claims 1 to 8.

11. A computer-readable medium storing one or more programs, which, when executed by an electronic device comprising a plurality of application programs, enable the electronic device to execute the material recommendation method according to any one of claims 1 to 8.

Images