Resource recommendation method and device, electronic equipment and storage medium

Abstract

The invention provides a resource recommendation method, and relates to the technical field of artificial intelligence, in particular to the technical field of resource recommendation and large models. According to the specific implementation scheme, in response to determining that at least one candidate resource description text recommended according to first test flow in a first test stage meets a first preset recommendation condition, at least one to-be-recommended resource is recommended according to at least one second test flow used for at least one candidate recommendation text in a second test stage, the to-be-recommended resources are candidate recommendation resources using candidate resource description texts, and the at least one candidate resource description text is used for describing the candidate recommendation resources; and in response to determining that at least one target candidate resource description text in the at least one candidate resource description text meets a second preset recommendation condition, recommending a candidate recommendation resource using a selected resource description text in the at least one target candidate resource description text based on the target traffic. The invention further provides a resource recommendation device, equipment and a storage medium.

Description

Technical Field

[0001] The present disclosure relates to the fields of artificial intelligence technology, particularly resource recommendation and large-scale model technology, and can be applied to scenarios such as generative search, intelligent document editing, intelligent assistants, virtual assistants, intelligent e-commerce, and live streaming. More specifically, the present disclosure provides a resource recommendation method, apparatus, electronic device, and storage medium. Background Art

[0002] With the development of artificial intelligence technology, the functions and applications of recommendation systems are constantly expanding. Recommendation systems can display resource names or titles to users to encourage them to click and browse resources. Resource names or titles can also be manually generated. Summary of the Invention

[0003] The present disclosure provides a resource recommendation method, apparatus, device, and storage medium.

[0004] According to one aspect of the present disclosure, a resource recommendation method is provided, the method comprising: in response to determining that at least one candidate resource description text recommended according to a first test traffic in a first test phase meets a first preset recommendation condition, in a second test phase, recommending at least one resource to be recommended according to at least one second test traffic for the at least one candidate recommendation text, wherein the resource to be recommended is a candidate recommended resource using the candidate resource description text, and the at least one candidate resource description text is used to describe the candidate recommended resource; in response to determining that at least one target candidate resource description text in the at least one candidate resource description text meets the second preset recommendation condition, recommending a candidate recommended resource using the selected resource description text in the at least one target candidate resource description text based on the target traffic.

[0005] According to another aspect of the present disclosure, a resource recommendation device is provided, which includes: a first recommendation module for, in response to determining that at least one candidate resource description text recommended according to a first test traffic in a first test phase meets a first preset recommendation condition, recommending at least one to-be-recommended resource according to at least one second test traffic for at least one candidate recommendation text in a second test phase, wherein the to-be-recommended resource is a candidate recommended resource using the candidate resource description text, and the at least one candidate resource description text is used to describe the candidate recommended resource; a second recommendation module for, in response to determining that at least one target candidate resource description text in the at least one candidate resource description text meets the second preset recommendation condition, recommending, based on the target traffic, a candidate recommended resource using a selected resource description text in the at least one target candidate resource description text.

[0006] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided according to the present disclosure.

[0007] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided. The computer instructions are used to cause a computer to execute the method provided according to the present disclosure.

[0008] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the method provided according to the present disclosure when executed by a processor.

[0009] 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 disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0011] Figure 1 is a schematic diagram of an exemplary system architecture to which the resource recommendation method and apparatus can be applied according to an embodiment of the present disclosure;

[0012] Figure 2 is a flowchart of a resource recommendation method according to an embodiment of the present disclosure;

[0013] Figure 3 is a schematic diagram of generating resource description text according to an embodiment of the present disclosure;

[0014] Figure 4 is a flowchart of a resource recommendation method according to an embodiment of the present disclosure;

[0015] Figure 5 is a block diagram of a resource recommendation apparatus according to an embodiment of the present disclosure; and

[0016] Figure 6 is a block diagram of an electronic device to which a resource recommendation method can be applied according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0017] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0018] Taking live streaming as an example, in a recommendation system, live streaming titles are usually manually configured by the host or operator. After being reviewed and approved, these titles are distributed to users to attract them to watch the live streaming.

[0019] However, manual titles are not updated in a timely manner, and the live broadcast room may use the same title for a long time. In addition, the quality of manual titles is also unstable, which may have a negative impact on the recommendation effect of the live broadcast.

[0020] Figure 1 This is a schematic diagram of an exemplary system architecture to which the resource recommendation method and apparatus can be applied according to an embodiment of the present disclosure. It should be noted that: Figure 1 The examples shown are merely examples of system architectures to which the embodiments of the present disclosure may be applied, to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure may not be used in other devices, systems, environments or scenarios.

[0021] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0022] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Terminal devices 101, 102, and 103 can be various electronic devices with display screens and support web browsing, including but not limited to smartphones, tablet computers, laptop computers, and desktop computers, etc.

[0023] Server 105 may be a server that provides various services, such as a background management server (for example only) that supports websites browsed by users using terminal devices 101, 102, and 103. The background management server may analyze and process received data such as user requests, and feed back processing results (e.g., web pages, information, or data obtained or generated based on user requests) to the terminal device.

[0024] It should be noted that the resource recommendation method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the resource recommendation device provided in the embodiment of the present disclosure can generally be set in the server 105. The resource recommendation method provided in the embodiment of the present disclosure can also be executed by the terminal devices 101, 102, and 103. Accordingly, the resource recommendation device provided in the embodiment of the present disclosure can also be set in the terminal devices 101, 102, and 103. The resource recommendation method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the resource recommendation device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0025] It can be understood that the above describes the system architecture of the present disclosure, and the following will describe the method of the present disclosure.

[0026] Figure 2 is a flowchart of a resource recommendation method according to an embodiment of the present disclosure.

[0027] like Figure 2 As shown, the method 200 may include operations S210 to S220.

[0028] In operation S210, in response to determining that at least one candidate resource description text recommended based on the first test traffic in the first test phase meets the first preset recommendation condition, at least one resource to be recommended is recommended based on at least one second test traffic for the at least one candidate recommendation text in the second test phase.

[0029] In the disclosed embodiments, the recommended resource is a candidate resource described in a candidate resource description text. The candidate resource can be a variety of resources. For example, the candidate resource can be video, audio, music, graphic information, and other resources. Another example is a live broadcast room. The live broadcast room has one or more hosts and one or more support staff. The hosts can be real people or virtual digital people.

[0030] In the disclosed embodiments, at least one candidate resource description text is used to describe a candidate recommended resource. For example, there may be multiple candidate resource description texts used to describe a candidate recommended resource. Multiple candidate resource description texts may describe the same candidate resource description text. For example, if the candidate recommended resource is a live broadcast room, the multiple candidate resource description texts may be multiple different live broadcast room titles for the live broadcast room.

[0031] In an embodiment of the present disclosure, in the first test phase, candidate recommended resources using the initial resource description text can be recommended based on the first test traffic. From one or more initial resource description texts, one or more initial resource description texts that meet the first preset recommendation condition are determined as one or more candidate resource description texts. In other words, one or more candidate resource description texts have passed the test in the first test phase. Next, a second test phase can be entered after the first test phase. In the second test phase, candidate recommended resources using the candidate resource description text can be recommended based on the second test traffic.

[0032] In the embodiment of the present disclosure, the first preset recommendation condition may be: a first indicator value for the initial resource description text determined in the first testing phase is greater than a first preset threshold.

[0033] In operation S220, in response to determining that at least one target candidate resource description text among the at least one candidate resource description text meets the second preset recommendation condition, a candidate recommended resource of the selected resource description text among the at least one target candidate resource description text is recommended based on target traffic.

[0034] In the embodiment of the present disclosure, one or more candidate resource description texts that meet the second preset recommendation condition are determined from the one or more candidate resource description texts as one or more target resource description texts. The selected resource description text can be determined from the one or more target resource description texts.

[0035] In the embodiment of the present disclosure, the second preset recommendation condition may be: a second indicator value for the candidate resource description text determined in the second testing phase is greater than a second preset threshold.

[0036] In the embodiment of the present disclosure, the target flow rate may be greater than the first test flow rate, or greater than the second test flow rate. The target flow rate may also be referred to as the main flow rate for the candidate recommended resource.

[0037] Through the disclosed embodiments, resource description texts are screened through two testing phases, allowing for the selection of high-quality resource description texts that are more attractive to users. Furthermore, during the testing phase, test traffic different from the main traffic is used to avoid recommending unscreened resource description texts to a large number of users, effectively improving the user experience. The resource description texts selected from one or more target resource description texts that have passed the two-phase test can effectively attract user clicks, improve the user experience, and increase the efficiency of recommending candidate recommended resources.

[0038] It can be understood that the above describes the method of the present disclosure, and the following will describe the initial resource description text of the present disclosure.

[0039] Figure 3The diagram is a schematic diagram of generating resource description text according to an embodiment of the present disclosure.

[0040] In some embodiments, the above method 200 may further include: extracting resource attribute information from candidate recommended resources according to information extraction prompt data using an information extraction macro model.

[0041] like Figure 3 As shown, the information extraction model mlm30 can be a multimodal large model (MLM) capable of processing modal data such as images, audio, and text. For example, the candidate recommendation resource can be a candidate live studio. A studio cover c31 can be obtained for the candidate studio, and one or more live video frames vf32 can be extracted from the candidate studio. The studio cover c31 and live video frames vf32 can be provided to the information extraction model mlm30.

[0042] In some embodiments, the information extraction prompt data includes at least one of a first role prompt text and at least one resource attribute description text. The first role prompt text can instruct the information extraction big model to extract information from the candidate recommended resources as the first role. The type of the first role is consistent with the type of at least one existing role in the candidate recommended resources. For example, the existing role can be an anchor or an auxiliary staff in the live broadcast room. The first role can be an anchor, for example. Taking the candidate live broadcast room as a beauty live broadcast room as an example, the first role prompt text can be "You are a beauty anchor. You can find key information based on the key pictures in the live broadcast and improve the distinguishability of the live broadcast room. Please find the key information." The beauty live broadcast room includes existing beauty anchors. Therefore, the role of the information extraction big model is consistent with the type of the existing role in the live broadcast room, which can improve the accuracy of the resource attribute information extracted by the information extraction big model.

[0043] In some embodiments, the resource attribute description text can instruct the information extraction model to extract attribute information for an existing character from candidate recommended resources. The resource attribute description text can include resource attribute field prompts and multiple preset resource attribute values. For example, using a beauty livestream studio as an example, the multiple resource attribute description texts may include hairstyle attribute description text, style attribute description text, and body shape attribute description text. The hairstyle attribute description text may be "Hairstyle (long straight hair, curly hair, short hair, unknown)." In this hairstyle attribute description text, long straight hair, curly hair, etc. may be preset resource attribute values. The information extraction model can determine the attribute values ​​of the existing character from these multiple preset resource attribute values. The style attribute description text may be "Style (value range: pure, cool, literary, ancient style, sporty, two-dimensional, unknown)." The body shape attribute description text may be "Body shape (value range: slender, athletic, petite, tall, well-proportioned, unknown)." Thus, multiple resource attribute information att30 can be determined from the candidate recommended resources.

[0044] In some embodiments, the information extraction prompt data also includes information format prompt text. The information format prompt text may instruct the information extraction macro model to output resource attribute information according to a preset attribute information format. For example, the information format prompt text may be "The description result is output in JSON format, with key being the resource attribute field and value being the attribute value." The preset attribute information format may be JAON format.

[0045] For another example, the information extraction prompt data including the above-mentioned first role prompt text, resource attribute description text and information format prompt text can be "You are a beauty anchor. You can find key information based on the key pictures in the live broadcast to improve the distinctiveness of your live broadcast room. Please find the key information and output the hairstyle (long straight hair, curly hair, short hair, unknown), style (value range: pure, cool, literary, ancient style, sports, two-dimensional, unknown), figure (value range: slender, athletic, petite, tall, symmetrical, unknown), the description result is output in json format, the key is the resource attribute field, and the value is the attribute value."

[0046] It can be understood that the above describes some methods for obtaining resource attribute information, and the following describes some methods for generating initial resource description text.

[0047] In some embodiments, the above method 200 may further include: generating prompt data and resource attribute information according to the description text using a text generation model to generate at least one initial resource description text. Figure 3 As shown, multiple resource attribute information att30 can be provided to the text generation large model llm30, so that the text generation large model generates prompt data and resource attribute information att30 according to the description text to generate multiple initial resource description texts.

[0048] The description text generation prompt data includes at least one of text style generation prompt text, text content generation prompt text, and text size prompt text.

[0049] The text style generation prompt instructs the large text generation model to generate initial resource description text according to the target text style. Multiple target text styles can be selected, including colloquial and suspenseful styles. For example, if the resource description text is a live broadcast title, the text style generation prompt might be "The new title should be colloquial and highly engaging, sparking users' strong curiosity and desire to watch. This can be achieved through the use of interrogative sentences, suspense, or engaging vocabulary."

[0050] The text content generation prompt text is used to instruct the text generation model to generate initial resource description text according to at least one of the resource category information and resource attribute information for the candidate recommended resource. The resource category information can be a first-level classification for the candidate recommended resource facility. Multiple first-level classifications can include entertainment, games, and finance, etc. Taking the first-level classification of finance as an example, multiple second-level classifications under the first-level classification can include bonds, stocks, etc. For example, the text content generation prompt text can be "The new title content should be closer to the given first-level classification and the host description, and more creative and detailed."

[0051] The text size description text instructs the text generation model to generate initial resource description text based on a target character count. The target character count is greater than or equal to a first preset character count threshold and less than or equal to a second preset character count threshold. The first preset character count threshold can be 15, and the second preset character count threshold can be 30. The text size description text can be, "The title must be at least 15 characters long and no more than 30 characters long, but may be adjusted as needed to ensure the completeness and appeal of the title."

[0052] In addition, the description text generation prompt data may also include a second role prompt text. The second role prompt text is used to instruct the text generation macro model to generate initial resource description text for the second role. For example, the second role prompt text may be "You are an expert in title creation, skilled in applying psychological principles and in-depth research on live broadcast user behavior. Now, please create a new live broadcast title based on the following information."

[0053] For another example, the description text generation prompt data including the second role prompt text, the description text generation prompt data including the text style generation prompt text, the text content generation prompt text, and the text size prompt text may be: "You are an expert in title creation, skilled in applying psychological principles and in-depth research on live broadcast user behavior. Now, please create a new live broadcast title based on the following information. You need to ensure that the new title:

[0054] - The new title should be colloquial and highly engaging, sparking the user's curiosity and desire to watch. This can be achieved by using questions, creating suspense, or using intriguing vocabulary.

[0055] - The new title content should be closer to the given first-level category and anchor description, and more creative and detailed.

[0056] - The title should be at least 15 words and no more than 30 words, but can be adjusted as needed to ensure the completeness and appeal of the title.

[0057] Original title:

[0058] ###

[0059] {{title}}

[0060] ###

[0061] First-level classification:

[0062] ###

[0063] {{cate}}

[0064] ###

[0065] Anchor information:

[0066] ###

[0067] {{star_info}}

[0068] ###

[0069] Please follow the above requirements, carefully understand the meaning of the original title, combine the first-level classification and the host description to create a new title, and output it in the following format:

[0070] "Title 1":

[0071] "Title 2":

[0072] "Title 3":"

[0073] The description text generation prompt text also includes multiple input information slots. The multiple input information slots are the original title {{title}} slot, the first-level category {{cate}} slot, and the anchor information {{star_info}} slot. The description text generation prompt text also includes multiple output information slots. The multiple output information slots include "Title 1", "Title 2", and "Title 3".

[0074] Therefore, if Figure 3As shown, the text generation model can be used to generate initial resource description text t31, initial resource description text t32, and initial resource description text t33. In addition, artificial text t34 can also be used as the initial resource description text. These multiple initial resource description texts describe the anchor of the beauty live broadcast room from different dimensions. Using the large model to generate different resource description texts based on text style prompt text and text content prompt text, etc., it can be efficiently tested to screen out resource description texts that are more attractive to users. It also helps to screen out description texts that are more compatible with the resources, which can further improve the user experience and improve the efficiency of recommendation.

[0075] Next, operation S301 can be performed on the initial resource description text t31, initial resource description text t32, and initial resource description text t33 to screen the multiple initial resource description texts based on preset rules. The preset rules can be laws or regulations. Operation S301 can be performed manually or by an audit model. The audit model can be a different model from the text generation model.

[0076] If the initial resource description text t31, the initial resource description text t32, and the initial resource description text t33 all meet the preset rules, operations S311 and S312 can be performed in sequence to perform the test of the first test phase and the test of the second test phase. Next, for at least one target resource description text that passes the second test phase, operation S321 is performed to determine the selected resource description text in at least one target candidate resource description text. Based on the selected resource description text tc30, operation S322 can be performed to recommend the candidate recommended resources using the selected resource description text based on the target traffic. In addition, based on the selected resource description text tc30, the text generation model can also be fine-tuned.

[0077] It is understood that the above description of the generation method of the initial description text of the present disclosure is described. In order to screen the resource description text, the initial resource description text can be evaluated, which will be described below.

[0078] In some embodiments, a first test traffic volume for the candidate resource description text in the first test phase is less than or equal to a second test traffic volume for the candidate resource description text in the second test phase. The first test traffic volume is the portion of the recommendation system traffic allocated to the initial resource description text in the first test phase. The second test traffic volume is the portion of the recommendation system traffic volume allocated to the candidate resource description text in the second test phase.

[0079] For example, a recommendation system can be used to execute the above method 200. The total traffic of the recommendation system can be the total traffic of multiple candidate recommendation resources. In one example, the recommendation system can recommend multiple resources to a first user within a preset time period. The first user can click on and browse the resource content of K resources among the multiple resources. In addition, the recommendation system can also recommend multiple resources to a second user. The second user can click on and browse the resource content of M resources among the multiple resources. If the users of the recommendation system are the first user and the second user, the total traffic of the recommendation system within the preset time period can be K+M. K is an integer greater than or equal to 1. M can be an integer greater than or equal to 1.

[0080] In the first test phase, the total traffic for candidate recommended resources can be divided into target traffic and a first available traffic. The first available traffic can be Q% of the total traffic for candidate recommended resources. The target traffic can be, for example, 1-Q% of the total traffic for candidate recommended resources. Q can be, for example, 20, and the target traffic can be 80%. Next, at least one first test traffic for at least one initial resource description text can be determined. Taking N initial resource description texts as an example, the first test traffic can be, for example, Q% / N. It can be understood that the first test traffic for different initial resource description texts can be the same or different. But the sum of multiple first test traffic can be the first available traffic. N can be an integer greater than or equal to 1. Q can be a non-negative value.

[0081] In the second test phase, the total traffic for the candidate recommended resources can be divided into target traffic and a second available traffic. The second available traffic can also be Q% of the total traffic for the candidate recommended resources. The target traffic can also be 1-Q% of the total traffic for the candidate recommended resources. Next, at least one second test traffic for at least one candidate resource description text can be determined. Taking J candidate resource description texts as an example, the second test traffic can be Q% / J, for example. It can be understood that the second test traffic for different candidate resource description texts can be the same or different. The sum of multiple second test traffic can be the second available traffic. J can be an integer greater than or equal to 1 and less than or equal to N.

[0082] In some embodiments, the above method 200 may further include: determining whether to display the candidate recommended resources using the initial resource description text to the object based on the first test traffic for the initial resource description text. For example, a first recommendation probability may be determined based on the first test traffic. Determine whether to display the candidate recommended resources using the initial resource description text to the object based on the first recommendation probability. At the start of the first test phase, the first recommendation probability may be the same as the value of the first test traffic. Next, the first recommendation probability may change dynamically so that at the end of the first test phase, the traffic difference between the actual traffic of the candidate recommended resources using the initial resource description text and the first test traffic is less than or equal to a preset traffic difference threshold.

[0083] In some embodiments, the above method 200 may further include: determining whether to display the candidate recommended resource using the candidate resource description text to the object based on a second test traffic for the candidate resource description text. For example, a second recommendation probability may be determined based on the second test traffic. Determine whether to display the candidate recommended resource using the candidate resource description text to the object based on the second recommendation probability. At the start moment of the second test phase, the second recommendation probability may be the same as the numerical value of the second test traffic. Next, the second recommendation probability may change dynamically so that at the end of the second test phase, the traffic difference between the actual traffic of the candidate recommended resource using the candidate resource description text and the second test traffic is less than or equal to a preset traffic difference threshold.

[0084] In some embodiments, the method 200 may further include at least one of the following operations: evaluating candidate recommended resources using the initial resource description text in the first test phase to obtain a first evaluation result; evaluating candidate recommended resources using the candidate resource description text in the second test phase to obtain a second evaluation result; and evaluating candidate recommended resources using the reference resource description text within a reference period to obtain a reference evaluation result. The reference period is the same as the duration of the second test phase.

[0085] The evaluation process can determine at least one indicator value of the candidate recommended resource within the period to be evaluated. The period to be evaluated can be the first test period of the above-mentioned first test phase, the second test period of the above-mentioned second test phase, or the reference period. The duration of the reference period is consistent with the duration of the second test period. By evaluating the candidate recommended resources using different resource description texts, it is possible to more accurately determine the situations in which the same candidate recommended resource using different resource description texts is clicked or browsed by users, which helps to more efficiently filter out high-quality resource description texts from different resource description texts.

[0086] At least one metric value may include impressions. The impressions can be the number of times a candidate recommended resource was displayed during the evaluation period. As described above, for candidate recommended resources using different resource descriptions, whether to display them to a target can be determined based on different traffic levels. The target can be a user. For example, if a candidate recommended resource was displayed to a target 10 times during the evaluation period, the impressions can be 10.

[0087] At least one indicator value may also include a click-through rate (CTR). The CTR may be a ratio of clicks to impressions. The CTR is the number of times a candidate recommendation resource is clicked during the evaluation period. For example, after a candidate recommendation resource is presented to a user, the user may determine whether to click on the candidate recommendation resource based on the resource recommendation text used by the candidate recommendation resource. If, during the evaluation period, the candidate recommendation resource is presented to the user 10 times and clicked 5 times, the CTR may be 50%.

[0088] At least one indicator value may also include a short-time display indicator value. The short-time display indicator value is the ratio between the number of short-time displays and the number of clicks. The number of short-time displays is: the number of times that the candidate recommended resource displays the resource content after being clicked is less than or equal to the preset time threshold. For example, the preset time threshold may be 3 seconds. The short-time display indicator value may also be referred to as the 3-second fast sliding rate. If, within the period to be evaluated, the candidate recommended resource is displayed to the object 10 times and is clicked 5 times. After the candidate recommended resource is clicked, the resource content of the candidate recommended resource may be displayed on the visual interface. If the user exits the viewing after only watching for 2 seconds, this display may be regarded as a short-time display. If 1 short-time display occurs, the short-time display indicator value may be determined to be 20%.

[0089] At least one index value may also include a weighted fusion index value. The weighted fusion index value is obtained by weighting the fusion index value using the evaluation weight. The fusion index value is obtained based on the processed click-through rate and the processed subtraction result. The processed click-through rate is the exponential operation result with the click-through rate as the base and the first parameter as the exponent. The processed subtraction result is the exponential operation result with the subtraction result as the base and the second parameter as the exponent. The subtraction result is obtained by subtracting the short-time display index value from the first preset value. The evaluation weight is obtained by dividing the addition result by the multiplication result. The addition result is obtained by adding the minimum value between the display difference value and the product result to the second preset value. The display difference value is the difference between the display amount and the display threshold. The product result is obtained by multiplying the adjustable parameter by the display threshold. For example, the fusion index value is obtained by multiplying the processed click-through rate and the processed subtraction result. The weighted fusion index value can be determined by the following formula :

[0090]

[0091] It can be the processed click rate. It can be click-through rate. Can be the first parameter. It can be used to display indicator values ​​for a short time. The first preset value may be 1. Can be the second parameter. It can be a fusion index value.

[0092] Can be used for evaluation weights. The second preset value may be 1. It can be used to display the minimum value between the difference value and the product result. Can be used to display difference values. It can be the display volume. Can be a display threshold. It can be the product result. This parameter can be an adjustable parameter. Through the disclosed embodiments, the degree of weighting can be adjusted using the adjustable parameter n. Resource description texts that display more content may have a higher click-through rate, while short-term display index values ​​may also be lower. By adjusting the fusion index value using evaluation weights based on the adjustable parameter, resource description texts with high display volumes can have higher weighted fusion index values, thereby selecting more accurate and attractive resource description texts.

[0093] It can be understood that the above describes the evaluation method of the present disclosure, and the following will further describe the method of the present disclosure.

[0094] Figure 4 is a schematic flowchart of a resource recommendation method according to an embodiment of the present disclosure.

[0095] like Figure 4 As shown, method 400 may include operations S401 to S404.

[0096] In operation S401, a plurality of initial resource description texts are generated.

[0097] For example, the above-mentioned information extraction model can be used to extract the resource attribute information of the candidate recommended resources. Next, the above-mentioned text generation model can be used to generate multiple initial resource description texts based on the resource attribute information.

[0098] In operation S402 , traffic is allocated.

[0099] For example, within the total traffic used for candidate recommended resources, available traffic can be allocated to multiple initial resource description texts, and target traffic can also be allocated to reference resource description texts. It is understood that in this embodiment, the available traffic can be the first available traffic or the second available traffic, i.e., the first available traffic and the second available traffic are equal. Furthermore, multiple initial resource description texts and reference resource description texts are all used to describe the same candidate recommended resource. In one example, multiple initial resource description texts and reference resource description texts can be different live room titles for the same live room.

[0100] In operation S403 , a reference resource description text is recommended based on the target traffic.

[0101] For example, candidate recommended resources using reference resource description text can be recommended based on target traffic. Target traffic can also be called primary traffic. Reference resource description text can be the primary recommendation text.

[0102] In operation S404 , a plurality of initial resource description texts are recommended based on available traffic.

[0103] For example, the available traffic can be redistributed to multiple initial resource description texts. Each initial resource description text is allocated a first test traffic. Candidate recommended resources using the initial resource description text can be recommended based on the first test traffic. The available traffic can be, for example, Q% of the total traffic used for the candidate recommended resources, where Q is a non-negative value.

[0104] Next, the above operation S210 may be performed, which will be further described below.

[0105] like Figure 4 As shown, in the first test stage stage 41, the traffic allocated to the initial resource description text t41, the initial resource description text t42, and the initial resource description text t43 is X% of the total traffic for the candidate recommended resources. It can be understood that, taking the three initial resource description texts as an example, X=Q / 3. X is a non-negative value.

[0106] The above-mentioned evaluation process can be performed on the candidate recommended resources using the initial resource description text t41 to obtain a first evaluation result for the initial resource description text t41, and the first evaluation result can include at least one first index value. As mentioned above, the evaluation process can determine at least one index value of the content of the candidate recommended resource in the time period to be evaluated. For the candidate recommended resources using the initial resource description text t41, the evaluation process can determine at least one first index value of the candidate recommended resource using the initial resource description text t41 within the first test time period. The at least one first index value includes at least one of the first display volume, the first click-through rate, the first short-time display index value, and the first weighted fusion index value. It can be understood that the above description of the display volume, click-through rate, short-time display index value, and weighted fusion index value also applies to the first display volume, the first click-through rate, the first short-time display index value, and the first weighted fusion index value, and the present disclosure will not repeat them here.

[0107] In addition, the above evaluation process can also be performed on the candidate recommended resources using the initial resource description text t42 and the candidate recommended resources using the initial resource description text t43, respectively, to obtain a first evaluation result for the initial resource description text t42 and a first evaluation result for the initial resource description text t43. It is understood that the description of the first evaluation result for the initial resource description text t42 and the first evaluation result for the initial resource description text t43 is the same or similar to the description of the first evaluation result for the initial resource description text t41, and is not further elaborated herein.

[0108] Next, operation S411 may be performed to determine whether the initial resource description text satisfies a first preset recommendation condition. The first preset recommendation condition may include at least one of the following: a first impression volume greater than or equal to a first preset impression threshold; a first click-through rate greater than or equal to a first preset click-through rate threshold; a first short-term impression index value less than or equal to a first preset short-term impression duration threshold; or a first weighted fusion index value greater than or equal to a first preset weighted fusion threshold.

[0109] For example, based on the first evaluation result for the initial resource description text t41, the first evaluation result for the initial resource description text t42, and the first evaluation result for the initial resource description text t43, if it is determined that the initial resource description text t41 meets any sub-condition of the first preset recommendation condition, the initial resource description text t41 can be used as the candidate resource description text t41'. The candidate resource description text t41' can enter the second test stage stage42. In the second test stage, based on the second test traffic of the candidate resource description text t41', the candidate recommended resource using the candidate resource description text t41' can be recommended. The second test traffic can be Y% of the total traffic for the candidate recommended resource, where Y is a non-negative value. Y can be 2X.

[0110] The above-mentioned evaluation process can be performed on the candidate recommended resources using the candidate resource description text t41' to obtain a second evaluation result for the candidate resource description text t41', and the second evaluation result can include at least one second index value. As mentioned above, the evaluation process can determine at least one index value of the content of the candidate recommended resource in the time period to be evaluated. For the candidate recommended resources using the candidate resource description text t41', the evaluation process can determine at least one second index value of the candidate recommended resource using the candidate resource description text t41' within the second test time period. The at least one second index value includes at least one of a second display volume, a second click-through rate, a second short-time display index value, and a second weighted fusion index value. It can be understood that the above description of the display volume, click-through rate, short-time display index value, and weighted fusion index value also applies to the second display volume, the second click-through rate, the second short-time display index value, and the second weighted fusion index value, and the present disclosure will not repeat them here.

[0111] In addition, the above-mentioned evaluation process can be performed on the candidate recommended resources using the reference resource description text to obtain a reference evaluation result for the reference resource description text, and the reference evaluation result can include at least one second index value. As mentioned above, the evaluation process can determine at least one index value of the content of the candidate recommended resource in the time period to be evaluated. For the candidate recommended resources using the reference resource description text, the evaluation process can determine at least one reference index value of the candidate recommended resource using the reference resource description text within the reference time period. The at least one reference index value includes at least one of a reference display volume, a reference click-through rate, a reference short-time display index value, and a reference weighted fusion index value. It can be understood that the above description of the display volume, click-through rate, short-time display index value, and weighted fusion index value also applies to the reference display volume, reference click-through rate, reference short-time display index value, and reference weighted fusion index value, and the present disclosure will not repeat them here.

[0112] Next, the above-mentioned operation S220 may be performed, which will be described below in conjunction with operations S421 to S423 .

[0113] In operation S421, it is determined whether the candidate resource description text meets the second preset recommendation condition. The second preset recommendation condition includes at least one of the following: the second display volume is greater than or equal to the second preset display threshold; the second click-through rate is greater than or equal to the reference click-through rate; the second short-term display index value is less than or equal to the reference short-term display duration threshold; the second weighted fusion index value is greater than or equal to the reference weighted fusion index value. The second preset recommendation condition involves the reference index value and the second index value, which can directly compare the candidate resource description text with the reference resource description text, and can efficiently determine whether to replace the reference resource description text with the candidate resource description text, which can effectively reduce labor costs and improve the efficiency of resource description text updates.

[0114] For example, based on the second evaluation result for candidate resource description text t41' and the reference evaluation result for the reference resource description text, if it is determined that candidate resource description text t41' meets any sub-condition of the second preset recommendation condition, candidate resource description text t41' may be selected as the resource description text. In other words, compared to the reference resource description text, candidate resource description text t41' is a more preferred resource description text.

[0115] In operation S422, the reference resource description text is replaced with the selected resource description text.

[0116] In operation S423 , candidate recommended resources using the selected resource description text are recommended based on the target traffic.

[0117] It is understood that the above description of the present disclosure is based on an example in which there is one candidate resource description text that meets the second preset recommendation condition. However, the present disclosure is not limited thereto, and the following description will be based on an example in which there is a candidate resource recommendation text that does not meet the second recommendation condition.

[0118] If a candidate resource recommendation text exists that does not meet the second recommendation condition, the candidate resource recommendation text can be used as the unselected resource description text tuc40. Furthermore, the initial resource description text t42 and the initial resource description text t43 described above, which do not meet the first preset recommendation condition, can also be used as the unselected resource description text tuc40. The following will be explained in conjunction with these three unselected resource description texts. The three unselected resource description texts are the first unselected resource description text, the second unselected resource description text, and the third unselected resource description text.

[0119] like Figure 4 As shown, method 400 may further include operation S431, determining whether the missing resource description text meets the first text elimination condition. The first text elimination condition may include: the click rate is less than a preset click rate threshold; the short-term display index value is greater than a preset short-term display duration threshold; the weighted fusion index value is less than a preset weighted fusion threshold

[0120] In response to determining that the missed resource description text satisfies the first text elimination condition, operation S433 may be executed to delete the resource description text. If the click-through rate of the first missed resource description text is less than a preset click-through rate threshold, the short-term display index value is greater than a preset short-term display duration threshold, or the weighted fusion index value is less than a preset weighted fusion threshold, operation S433 may be executed to delete the first missed resource description text.

[0121] In response to determining that the missed resource description text does not meet the first text elimination condition, operation S432 may be performed to determine whether the missed resource description text meets the second text elimination condition. For example, if the second missed resource description text and the third missed resource description text do not meet the first text elimination condition, operation S432 may be performed. The second text elimination condition may be: the click-through rate is the minimum click-through rate; the short-term display index value is the maximum short-term display index value; and the weighted fusion index value is the minimum weighted fusion index value. If, between the second missed resource description text and the third missed resource description text, the second missed resource description text has the lowest click-through rate, the highest short-term display index value, and the lowest weighted fusion index value, the second missed resource description text may be deleted. The third missed resource description text may be used as the initial resource description text and returned to the first test phase of the subsequent test cycle. In addition, the replaced reference resource description text may also be used as the initial resource description text and returned to the first test phase or the second test phase of the subsequent test cycle. If the reference resource description text and the missed resource description text do not meet the elimination condition, they may be returned to the test phase for retesting. When recommending candidate resources using the selected resource description based on target traffic, even if the reference resource description has been replaced, the evaluation results of the selected resource description based on the target traffic may differ from the evaluation results of the selected resource description based on the test traffic, due to the difference between the test traffic and the target traffic. Therefore, the reference resource description and the missed resource description are returned to the testing phase for retesting, which can quickly provide better alternative descriptions for the candidate resources, further improving recommendation efficiency.

[0122] Next, after deleting one or more missing resource description texts, the process may return to operation S401 to generate multiple initial resource description texts for subsequent test cycles. It is understood that each test cycle may include a first test phase and a second test phase. The length of a test cycle may be greater than 24 hours. It is understood that the first test phase and the second test phase may be different stages of the race.

[0123] It can be understood that the above describes the second preset recommendation condition of the present disclosure. In other embodiments, the second preset recommendation condition may also include at least one of the following: the second click-through rate is greater than or equal to the second preset click-through rate threshold; the second short-time display index value is less than or equal to the second preset short-time display duration threshold; the second weighted fusion index value is greater than or equal to the second preset weighted fusion threshold.

[0124] It can be understood that the above describes the method of the present disclosure, and the following will describe the device of the present disclosure.

[0125] Figure 5is a block diagram of a resource recommendation apparatus according to an embodiment of the present disclosure.

[0126] like Figure 5 As shown, the apparatus 500 may include a first recommendation module 510 and a second recommendation module 520 .

[0127] The first recommendation module 510 is configured to, in response to determining that at least one candidate resource description text recommended based on the first test traffic flow during the first test phase satisfies a first preset recommendation condition, recommend at least one resource to be recommended based on at least one second test traffic flow for the at least one candidate recommendation text during the second test phase. The resource to be recommended is a candidate resource for recommendation using the candidate resource description text, and the at least one candidate resource description text is used to describe the candidate resource for recommendation.

[0128] The second recommendation module 520 is configured to recommend, based on target traffic, a candidate recommended resource of the selected resource description text in the at least one target candidate resource description text in response to determining that the at least one target candidate resource description text in the at least one candidate resource description text meets a second preset recommendation condition.

[0129] In some embodiments, the candidate resource description text is one of at least one initial resource description text recommended in the first test phase, and the first test traffic for the candidate resource description text in the first test phase is less than or equal to the second test traffic for the candidate resource description text in the second test phase. The first test traffic is the portion of the recommendation system traffic allocated to the initial resource description text in the first test phase. The second test traffic is the portion of the recommendation system traffic allocated to the candidate resource description text in the second test phase.

[0130] In some embodiments, the candidate resource description text is one of at least one initial resource description text recommended in the first testing phase, and the target traffic is a portion of the recommendation system traffic allocated to the reference resource description text. The apparatus 500 further includes at least one of the following modules: a first determination module configured to determine, based on a first test traffic for the initial resource description text, whether to display the candidate recommended resource using the initial resource description text to the subject. A second determination module configured to determine, based on a second test traffic for the candidate resource description text, whether to display the candidate recommended resource using the candidate resource description text to the subject.

[0131] In some embodiments, the apparatus 500 further includes at least one of the following modules: a first evaluation processing module configured to evaluate candidate recommended resources using the initial resource description text in the first test phase to obtain a first evaluation result; a second evaluation processing module configured to evaluate candidate recommended resources using the candidate resource description text in the second test phase to obtain a second evaluation result; and a third evaluation processing module configured to evaluate candidate recommended resources using the reference resource description text within a reference period to obtain a reference evaluation result, wherein the reference period is the same as the second test phase.

[0132] In some embodiments, the evaluation process is used to determine at least one indicator value of the candidate recommended resource within the time period to be evaluated, and the time period to be evaluated is the first test time period of the first test phase, the second test time period of the second test phase, or the reference time period. The at least one indicator value includes at least one of the display volume, click-through rate, and short-time display indicator value. The display volume is the number of times the candidate recommended resource is displayed within the time period to be evaluated. The click-through rate is the ratio between the number of clicks and the display volume, and the number of clicks is the number of times the candidate recommended resource is clicked within the time period to be evaluated. The short-time display indicator value is the ratio between the number of short-time displays and the number of clicks, and the number of short-time displays is: the number of times the candidate recommended resource displays the resource content after being clicked is less than or equal to the preset time threshold.

[0133] In some embodiments, at least one index value also includes a weighted fusion index value. The weighted fusion index value is obtained by weighting the fusion index value using the evaluation weight. The fusion index value is obtained based on the processed click-through rate and the processed subtraction result, the processed click-through rate is the exponential operation result with the click-through rate as the base and the first parameter as the exponent, the processed subtraction result is the exponential operation result with the subtraction result as the base and the second parameter as the exponent, and the subtraction result is obtained by subtracting the short-time display index value from the first preset value. The evaluation weight is obtained by dividing the addition result by the multiplication result, the addition result is obtained by adding the minimum value between the display difference value and the product result to the second preset value, the display difference value is the difference between the display amount and the display threshold, and the product result is obtained by multiplying the adjustable parameter by the display threshold.

[0134] In some embodiments, the first evaluation result includes a first impression volume, a first click-through rate, a first short-term impression index value, and a first weighted fusion index value, and the first preset recommendation condition includes at least one of the following: the first impression volume is greater than or equal to a first preset impression threshold; the first click-through rate is greater than or equal to a first preset click-through rate threshold; the first short-term impression index value is less than or equal to a first preset short-term impression duration threshold; and the first weighted fusion index value is greater than or equal to the first preset weighted fusion threshold.

[0135] In some embodiments, the second evaluation result includes a second amount of impressions, a second click-through rate, a second short-term display index value, and a second weighted fusion index value, and the reference evaluation result includes a reference amount of impressions, a reference click-through rate, a reference short-term display index value, and a reference weighted fusion index value. The second preset recommendation condition includes at least one of the following: the second amount of impressions is greater than or equal to a second preset impression threshold; the second click-through rate is greater than or equal to the reference click-through rate; the second short-term display index value is less than or equal to the reference short-term display duration threshold; and the second weighted fusion index value is greater than or equal to the reference weighted fusion index value.

[0136] In some embodiments, the second recommendation module 520 includes: a replacement submodule for replacing the reference resource description text with the selected resource description text; and a recommendation submodule for recommending candidate recommended resources using the selected resource description text based on target traffic.

[0137] In some embodiments, the apparatus 500 further includes an extraction module configured to extract resource attribute information from the candidate recommended resources using an information extraction macromodel based on information extraction prompt data. A generation module configured to generate prompt data and resource attribute information based on the description text using a text generation macromodel to generate at least one initial resource description text.

[0138] In some embodiments, the information prompt data includes at least one of a first role prompt text and at least one resource attribute description text. The first role prompt text is used to instruct the information extraction macromodel to extract information from the candidate recommended resource as a first role, where the type of the first role is consistent with the type of at least one existing role in the candidate recommended resource. The resource attribute description text is used to instruct the information extraction macromodel to extract attribute information of the existing role from the candidate recommended resource.

[0139] In some embodiments, the description text generation prompt data includes at least one of a text style generation prompt text, a text content generation prompt text, and a text size prompt text. The text style generation prompt text is used to instruct the text generation model to generate an initial resource description text according to a target text style. The text content generation prompt text is used to instruct the text generation model to generate an initial resource description text according to at least one of resource category information and resource attribute information for a candidate recommended resource. The text size description text is used to instruct the text generation model to generate an initial resource description text based on a target number of characters, where the target number of characters is greater than or equal to a first preset character number threshold and less than or equal to a second preset character number threshold.

[0140] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0141] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0142] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure 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 assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0143] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. Computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to bus 604.

[0144] Various components in device 600 are connected to I / O interface 605, including an input unit 606, such as a keyboard, mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a magnetic disk, optical disk, etc.; and a communication unit 609, such as a network card, modem, wireless communication transceiver, etc. The communication unit 609 allows device 600 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0145] The computing unit 601 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the resource recommendation method. For example, in some embodiments, the resource recommendation method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the resource recommendation method described above can be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to execute the resource recommendation method in any other appropriate manner (eg, by means of firmware).

[0146] Various embodiments of the systems and techniques described 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 parts (ASSPs), system on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes 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.

[0147] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0148] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or apparatus. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory (EPROM) or flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0149] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) display or a liquid crystal display (LCD)) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0150] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0151] Computer systems may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The client and server relationship arises through computer programs running on the respective computers and having a client-server relationship to each other.

[0152] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0153] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A resource recommendation method, comprising: In response to determining that at least one candidate resource description text recommended according to the first test traffic in the first test phase meets the first preset recommendation condition, at least one resource to be recommended is recommended according to at least one second test traffic for at least one candidate recommendation text in the second test phase, wherein the resource to be recommended is a candidate recommendation resource using the candidate resource description text, and the at least one candidate resource description text is used to describe the candidate recommendation resource; In response to determining that at least one target candidate resource description text in at least one of the candidate resource description texts meets a second preset recommendation condition, based on target traffic, a candidate recommended resource of the selected resource description text in at least one of the target candidate resource description texts is recommended for use.

2. The method according to claim 1, wherein The candidate resource description text is one of the at least one initial resource description text recommended in the first test phase, a first test flow rate for the candidate resource description text in the first test phase is less than or equal to a second test flow rate for the candidate resource description text in the second test phase, The first test traffic is a portion of the traffic of the recommendation system allocated to the initial resource description text in the first test phase; The second test traffic is a portion of the traffic of the recommendation system allocated to the candidate resource description text in the second test phase.

3. The method according to claim 1, wherein The candidate resource description text is one of the at least one initial resource description text recommended in the first testing phase, and the target traffic is a portion of the traffic of the recommendation system allocated to the reference resource description text; It also includes at least one of the following operations: determining, based on the first test traffic for the initial resource description text, whether to display the candidate recommended resource using the initial resource description text to an object; Determine whether to display the candidate recommended resource using the candidate resource description text to the object based on the second test traffic for the candidate resource description text.

4. The method according to claim 3, wherein: It also includes at least one of the following operations: Evaluating the candidate recommended resources using the initial resource description text in the first testing phase to obtain a first evaluation result; evaluating the candidate recommended resources using the candidate resource description text in the second testing phase to obtain a second evaluation result; The evaluation process is performed on candidate recommended resources that use the reference resource description text within a reference period to obtain a reference evaluation result, wherein the duration of the reference period is consistent with the duration of the second test phase.

5. The method according to claim 4, wherein The evaluation process is used to determine at least one indicator value of the candidate recommended resource within a to-be-evaluated period, where the to-be-evaluated period is the first test period of the first test phase, the second test period of the second test phase, or the reference period. At least one of the indicator values ​​includes at least one of the display volume, click-through rate, and short-time display indicator value. The display quantity is the number of times the candidate recommendation resource is displayed during the evaluation period. The click rate is the ratio between the number of clicks and the number of impressions, and the number of clicks is the number of times the candidate recommendation resource is clicked during the evaluation period. The short-time display index value is the ratio between the number of short-time displays and the number of clicks, and the number of short-time displays is: the number of times that the candidate recommended resource displays resource content after being clicked is less than or equal to a preset duration threshold.

6. The method according to claim 5, wherein: At least one of the index values ​​further includes a weighted fusion index value, The weighted fusion index value is obtained by weighting the fusion index value using the evaluation weight. The fusion index value is obtained based on the processed click-through rate and a processed subtraction operation result, wherein the processed click-through rate is an exponential operation result with the click-through rate as the base and the first parameter as the exponent, and the processed subtraction operation result is an exponential operation result with the subtraction operation result as the base and the second parameter as the exponent, and the subtraction operation result is obtained by subtracting the short-term display index value from the first preset value; The evaluation weight is obtained by dividing the addition result by the multiplication result, the addition result is obtained by adding the minimum value between the display difference value and the multiplication result to the second preset value, the display difference value is the difference between the display amount and the display threshold, and the multiplication result is obtained by multiplying the adjustable parameter by the display threshold.

7. The method according to claim 6, wherein: The first evaluation result includes a first display volume, a first click-through rate, a first short-term display index value, and a first weighted fusion index value. The first preset recommendation condition includes at least one of the following: The first display amount is greater than or equal to a first preset display threshold; The first click rate is greater than or equal to a first preset click rate threshold; The first short-term display indicator value is less than or equal to a first preset short-term display duration threshold; The first weighted fusion index value is greater than or equal to a first preset weighted fusion threshold.

8. The method according to claim 6, wherein: The second evaluation result includes a second display volume, a second click-through rate, a second short-term display index value, and a second weighted fusion index value; the reference evaluation result includes a reference display volume, a reference click-through rate, a reference short-term display index value, and a reference weighted fusion index value. The second preset recommendation condition includes at least one of the following: The second display amount is greater than or equal to a second preset display threshold; The second click-through rate is greater than or equal to the reference click-through rate; The second short-term display indicator value is less than or equal to the reference short-term display duration threshold; The second weighted fusion index value is greater than or equal to the reference weighted fusion index value.

9. The method according to claim 1, wherein: The recommending, based on the target traffic, to use at least one candidate recommended resource having the selected resource description text in the target candidate resource description text comprises: Replacing the reference resource description text with the selected resource description text; Recommend candidate recommended resources using the selected resource description text based on the target traffic.

10. The method according to claim 2, further comprising: Extracting resource attribute information from the candidate recommended resources based on information extraction prompt data using an information extraction model; The text generation model is used to generate prompt data and the resource attribute information according to the description text, thereby generating at least one initial resource description text.

11. The method according to claim 10, wherein: The information prompt data includes at least one of a first role prompt text and at least one resource attribute description text, The first role prompt text is used to instruct the information extraction model to extract information from the candidate recommendation resource as the first role, and the type of the first role is consistent with the type of at least one existing role in the candidate recommendation resource. The resource attribute description text is used to instruct the information extraction model to extract the attribute information of the existing role from the candidate recommendation resource.

12. The method according to claim 10, wherein: The description text generation prompt data includes at least one of text style generation prompt text, text content generation prompt text and text size prompt text, The text style generation prompt text is used to instruct the text generation model to generate the initial resource description text according to the target text style; The text content generation prompt text is used to instruct the text generation model to generate the initial resource description text according to at least one of the resource category information and the resource attribute information for the candidate recommended resource. The text scale description text is used to instruct the text generation model to generate the initial resource description text based on a target number of characters, where the target number of characters is greater than or equal to a first preset character number threshold and less than or equal to a second preset character number threshold.

13. A resource recommendation device, comprising: A first recommendation module is configured to, in response to determining that at least one candidate resource description text recommended according to the first test traffic in the first test phase meets the first preset recommendation condition, recommend at least one to-be-recommended resource according to at least one second test traffic for at least one of the candidate recommendation texts in the second test phase, wherein the to-be-recommended resource is a candidate recommended resource using the candidate resource description text, and the at least one candidate resource description text is used to describe the candidate recommended resource; The second recommendation module is used to recommend the candidate recommended resource of the selected resource description text in at least one of the candidate resource description texts based on the target traffic in response to determining that at least one target candidate resource description text in at least one of the candidate resource description texts meets the second preset recommendation condition.

14. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12.

15. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 12.

16. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images