Resource publishing method and device based on LLM model

Abstract

The resource publishing method based on the LLM model is applied to a resource publishing platform, and the resource publishing platform executes a conversion task of converting external resources into standard resources supported by the resource publishing platform based on the accessed LLM model. The method comprises the following steps: acquiring resource information of external resources to be published; generating a cue word corresponding to the conversion task based on the resource information of the external resource, and inputting the cue word into an LLM model, so that the LLM model executes reasoning calculation corresponding to the conversion task based on the cue word, and converts the external resource into the standard resource; and obtaining the standard resource output by the LLM model, and performing resource publishing on the standard resource on the resource publishing platform.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of artificial intelligence technology, and more particularly to a resource publishing method and device based on an LLM model. Background Art

[0002] On some traditional resource publishing platforms, in order to convert external resources into standard resources supported by the resource publishing platform and publish them on the resource publishing platform, a lot of manual operations are usually required.

[0003] For example, the operation personnel responsible for resource publishing are usually required to obtain the resource information of external resources from the providers of external resources offline, and create a data structure of standard resources corresponding to the external resources on the resource publishing platform. Then, the detailed contents of the resource information of the external resources obtained are filled in the created data structure of the standard resources one by one. The whole process is time-consuming and labor-intensive. Summary of the invention

[0004] This specification proposes a resource publishing method based on an LLM model, which is applied to a resource publishing platform. The resource publishing platform performs a conversion task of converting external resources into standard resources supported by the resource publishing platform based on the accessed LLM model; the method includes: Get resource information of external resources to be published; Based on the resource information of the external resource, a prompt word corresponding to the conversion task is generated, and the prompt word is input into the LLM model, so that the LLM model performs an inference calculation corresponding to the conversion task based on the prompt word to convert the external resource into the standard resource; The standard resource output by the LLM model is obtained, and the standard resource is published on the resource publishing platform.

[0005] This specification also proposes a resource publishing device based on the LLM model, which is applied to a resource publishing platform. The resource publishing platform performs a conversion task of converting external resources into standard resources supported by the resource publishing platform based on the accessed LLM model; the method includes: The acquisition module acquires the resource information of the external resources to be published; A conversion module, based on the resource information of the external resource, generates a prompt word corresponding to the conversion task, and inputs the prompt word into the LLM model, so that the LLM model performs an inference calculation corresponding to the conversion task based on the prompt word to convert the external resource into the standard resource; In the above embodiment, the LLM model is introduced into the resource publishing platform to perform the conversion task of converting external resources into standard resources supported by the resource publishing platform. When publishing external resources, prompt words can be generated to guide the LLM model to convert external resources into standard resources. In this way, the external resources can be automatically converted into standard resources for publication without human intervention, thereby reducing labor costs and improving the efficiency of external resource publication. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0007] Figure 1 It is a schematic diagram of the architecture of an Internet service system shown in an embodiment of this specification; Figure 2 It is a flow chart of a resource publishing method based on the LLM model shown in an embodiment of this specification; Figure 3 is a schematic diagram of an event-driven task execution framework shown in an embodiment of this specification; Figure 4 is a flow chart of converting external resources into standard resources shown in an embodiment of this specification; Figure 5 It is an architecture diagram of a task execution framework based on a general LLM model shown in an embodiment of this specification; Figure 6 is a schematic structural diagram of an electronic device shown in an embodiment of this specification; Figure 7 It is a block diagram of a resource publishing device based on the LLM model shown in an embodiment of this specification. DETAILED DESCRIPTION

[0008] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of this specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0009] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0010] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this manual are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0011] This specification aims to propose a technical solution that automatically converts high-value external resources into standard resources and publishes them on the resource publishing platform by introducing the LLM (Large Language Model) model in the resource publishing platform without human intervention.

[0012] Figure 1 FIG. 1 is a schematic diagram of an Internet service system architecture provided by an exemplary embodiment. Figure 1 As shown, the system may include a server 11, a network 12, and several electronic devices, such as a PC (Personal Computer) 13, a mobile phone 14, and the like.

[0013] The server 11 may be a physical server including an independent host, or the server 11 may be a virtual server carried by a host cluster. During operation, the server 11 may run a server-side program of a certain application to implement the relevant functions of the application. For example, when the server 11 runs a program of a certain Internet service, it may be implemented as a corresponding Internet service platform.

[0014] PC 13 and mobile phone 14 are only some types of electronic devices that users can use. In fact, users can obviously also use electronic devices such as the following types: tablet devices, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smart watches, etc.), etc., and one or more embodiments of this specification do not limit this. During operation, the electronic device can run a program on the client side of an application to implement the relevant functions of the application. For example, when the electronic device runs a program for an Internet service, it can be implemented as a client of the Internet service. Among them, the application program of the client of the above-mentioned Internet service can be started and run on the electronic device. The program on the client side can be a native application installed on the electronic device, or the program on the client side can be a small program, a quick application or other similar forms.

[0015] Of course, when using web technologies such as HTML5 or similar, relevant functions can be implemented through pages displayed by browsers. The browser here can be an independent browser application or a browser module embedded in certain applications.

[0016] As for the network 12 for interaction between electronic devices such as PC 13 and mobile phone 14 and server 11, it is possible to select a wired or wireless network to achieve communication based on the communication mode supported by the corresponding electronic device, and this specification does not limit this. For example, PC 13 can support both wired and wireless communication, so it can use a wired or wireless network to achieve communication as needed, while mobile phone 14 usually only supports wireless communication, so it can use a wireless network to achieve communication.

[0017] It should be noted that the above-mentioned Internet services may include any services implemented on the Internet; for example, the above-mentioned Internet services may specifically include resource publishing services (such as travel products) that can be provided to users by a resource publishing platform (such as a product publishing platform), resource trading services (such as travel trading services) and other e-commerce services, or other types of services, etc. This specification does not specifically limit this.

[0018] The technical solution of this specification is described in detail below with reference to the accompanying drawings.

[0019] See also Figure 2 , Figure 2 The flowchart of a resource publishing method based on the LLM model shown in this specification includes the following execution process: Step 202: Obtain resource information of the external resource to be published; The execution subject of the resource publishing method may be a resource publishing platform, wherein in actual application, the resource publishing platform may include a service platform capable of publishing any type of resources.

[0020] In an embodiment shown, the above-mentioned resources may specifically include commodity resources. In this case, the above-mentioned resource publishing platform may specifically include a commodity publishing platform.

[0021] For example, in one example, the above-mentioned commodity may be a travel commodity, and the above-mentioned commodity publishing platform may be a travel transaction service platform that can provide travel transaction services to users. It should be noted that the above-mentioned travel commodity may specifically include any type of travel-related commodity; for example, ticket commodities, air ticket commodities, accommodation commodities, train ticket commodities, etc.

[0022] Of course, in actual applications, the resource publishing platform may be other types of resource publishing platforms other than the product publishing platform, which will not be listed one by one in this specification. For example, the resource publishing platform may also be a content publishing platform for media content such as short videos, news, or other forms.

[0023] In this specification, in order to improve the automation level of the resource publishing platform when publishing external resources, the resource publishing platform can specifically access the LLM model and use the LLM model to perform the conversion task of converting external resources into standard resources supported by the resource publishing platform.

[0024] Among them, the above-mentioned LLM model may specifically include a general LLM model, and may also include an LLM model obtained by fine-tuning and training a data set in a specific application scenario on the basis of the general LLM model, which is not particularly limited in this specification.

[0025] The above-mentioned external resources may specifically include resources provided by external resource providers connected to the above-mentioned resource publishing platform; correspondingly, the above-mentioned standard resources may specifically include resources having a standard information format supported by the resource publishing platform.

[0026] For example, in one example, taking the above resources as commodity resources, the above external resources may specifically be commodities provided by external commodity providers (such as external merchants cooperating with the commodity publishing platform) that are connected to the commodity publishing platform. The above standard resources may specifically be commodities with standard information formats supported by the commodity publishing platform.

[0027] When the resource publishing platform supports publishing external resources provided by an external resource provider, it can connect with the external resource provider to obtain resource information of the external resources to be published from the external resource provider.

[0028] For example, in one example, still taking the above-mentioned resources as commodity resources, the above-mentioned resource publishing platform can specifically be a commodity publishing platform, and the above-mentioned external resource provider can be an external merchant cooperating with the commodity publishing platform. In this case, the external merchant's server can provide an API interface for obtaining external commodities provided by the external merchant. When the resource publishing platform needs to publish the external commodities provided by the external merchant, it can obtain the commodity information of the external commodities provided by the external merchant from the server of the external merchant by calling the API interface.

[0029] It should be noted that the resource information of the external resources to be published obtained by the resource publishing platform from the external resource provider may specifically include any form of information related to resource conversion, which is not particularly limited in this specification.

[0030] For example, in one example, taking the above-mentioned resources as ticket products provided by external product providers, the resource information of the above-mentioned external resources that needs to be obtained may specifically include information such as the name of the ticket product, the name of the attraction, the country to which the attraction belongs, the city to which the attraction belongs, the address of the attraction, the cost of the attraction, etc.

[0031] In actual applications, if the number of external resources that an external resource provider can provide is relatively large, in order to prevent the resource publishing platform that is connected to its system from quickly obtaining all the external resources that the external resource provider can provide in a short period of time, the external resource provider will usually limit the flow of the resource publishing platform that is connected to its system.

[0032] For example, in one example, still taking the above-mentioned resources as commodity resources, the above-mentioned resource publishing platform can be a commodity publishing platform, and the above-mentioned external resource provider can be an external merchant cooperating with the commodity publishing platform. In this case, the external merchant's service end can provide an API interface for obtaining external commodities provided by the external merchant, and perform flow control on the API interface; for example, configure the maximum QPS (Queries PerSecond) for the API. When the resource publishing platform needs to publish the external commodity provided by the external merchant, it can query the commodity information of the external commodity by calling the API interface according to the maximum QPS configured for the API.

[0033] In this case, the resource publishing platform can specifically acquire high-value external resources provided by the external resource provider as much as possible, on the premise that the external resource provider has performed flow limiting processing.

[0034] In one embodiment shown, when the above-mentioned resource publishing platform obtains resource information of external resources to be published from external resource providers, it can specifically pre-calculate the resource value score of each external resource in the external resource set provided by the external resource provider; wherein the resource value score can be specifically used to represent the resource value of the external resource; then, the calculated resource value score can be further mapped into a scheduling priority; wherein the scheduling priority can be specifically positively correlated with the resource value score.

[0035] After the calculated resource value score is mapped into a scheduling priority, the above-mentioned resource publishing platform can use the scheduling priority as a scheduling factor and adopt a periodic resource scheduling method to periodically schedule the external resources in the external resource set to determine the external resources to be released from the external resource set, and then selectively obtain the resource information of these determined external resources to be released from the external resource providers.

[0036] Among them, the scoring strategy adopted by the above-mentioned resource publishing platform when calculating the resource value score of each external resource in the external resource set provided by the external resource provider is no longer specifically limited in this specification. In actual application, the resource value of each external resource in the external resource set can be flexibly evaluated in combination with specific needs.

[0037] In an illustrated implementation manner, the resource publishing platform may evaluate the resource value of the external resources by collecting resource indicators that can reflect the resource value of the external resources.

[0038] In this case, the above-mentioned resource publishing platform can first obtain multiple resource indicators of each external resource in the external resource set; wherein, the multiple resource indicators can specifically be indicators that can represent the resource value of the external resources; then, the multiple resource indicators can be weighted calculated, and the weighted sum obtained by the weighted calculation can be used as the resource value score of each external resource in the external resource set.

[0039] In one embodiment shown, the above-mentioned multiple resource indicators may specifically include a first indicator for representing the traffic value of external resources, and a second indicator for representing the transaction value of external resources; in this case, the above-mentioned resource publishing platform may perform weighted calculation on the first indicator and the second indicator to obtain the resource value score of each external resource in the external resource set.

[0040] The specific representation forms of the first indicator and the second indicator are not particularly limited in this specification. In practical applications, any form of indicator that can be used to represent the flow value of external resources can be used as the first indicator; correspondingly, any form of indicator that can be used to represent the transaction value of external resources can also be used as the second indicator.

[0041] For example, in one example, the traffic value of the external resource can be represented by the ratio of the number of visits to the external resource to the total number of visits to each external resource in the external resource set; in this case, the above-mentioned first indicator can specifically include the ratio of the number of visits to the external resource to the total number of visits to each external resource in the external resource set.

[0042] In another example, the transaction value of the external resource can also be represented by the product of the order quantity of the external resource and the order amount of the external resource (i.e., the order unit price). In this case, the second indicator can specifically include the product of the order quantity of the external resource and the order amount of the external resource.

[0043] It should be noted that, in practical applications, when performing weighted calculation on the above-mentioned first indicator and the above-mentioned second indicator, the weighting coefficients of the first indicator and the second indicator can be flexibly set based on specific needs and will not be specifically limited in this specification.

[0044] In one embodiment shown, the weighted coefficients of the first indicator and the second indicator may specifically be weighted coefficients calculated based on the resource conversion rate of the external resources.

[0045] For instance, in one example, the ratio of the weighted coefficient of the first indicator to the weighted coefficient of the second indicator can specifically be the resource conversion rate of external resources; that is, when setting the weighted coefficients for the first indicator and the second indicator, it can be ensured that the ratio of the first indicator to the second indicator is the resource conversion rate of external resources; for example, assuming that the resource conversion rate of external resources is 5%, then it can be ensured that the ratio of the first indicator to the second indicator is set to 0.05.

[0046] In one embodiment shown, the above-mentioned multiple resource indicators may include a first indicator for indicating the flow value of external resources and a second indicator for indicating the transaction value of external resources, and may also include a third indicator for indicating the change frequency of resource information of external resources. In this case, the above-mentioned resource publishing platform may perform weighted calculation on the first indicator, the second indicator, and the third indicator to obtain the resource value score of each external resource in the external resource set.

[0047] In this way, in the process of calculating the resource value score of external resources, an indicator for representing the frequency of change of the resource information of external resources can be introduced, so that higher resource value scores can be given to external resources whose resource information changes frequently, so that when the above-mentioned resource publishing platform performs resource scheduling on external resources in the external resource set provided by the external resource provider, it can give priority to obtaining the resource information of external resources whose resource information changes frequently, thereby avoiding the problem of order failure due to failure to timely obtain the resource information of external resources whose resource information changes frequently.

[0048] The specific representation form of the third indicator is not particularly limited in this specification. In practical applications, any form of indicator that can be used to represent the frequency of changes in resource information of external resources can be used as the third indicator.

[0049] For example, in one example, the number of times each information field included in the resource information of the external resource changes within a preset time period in the future can be used to represent the change frequency of the resource information of the external resource. In this case, the third indicator can specifically include a weighted sum obtained by weighted calculation of the number of times each information field included in the resource information of the external resource changes within a preset time period in the future.

[0050] Among them, when performing weighted calculation on the number of times each information field contained in the resource information of the external resource changes within a preset time period in the future, the weighted value of the number of times these information fields change within the preset time period in the future can be flexibly set based on specific needs and will not be specifically limited in this specification.

[0051] For example, a weighting coefficient can be set for each field based on its importance; and the importance of each field can be evaluated by the degree of influence of each field on the transaction success rate of external resources. For example, a field with a relatively high degree of influence on the transaction success rate of external resources can be set with a larger weighting coefficient; conversely, a smaller weighting coefficient can be set.

[0052] It should be noted that, in actual applications, the number of times that each information field contained in the resource information of the above-mentioned external resources changes within a preset period of time in the future can be specifically predicted based on the historical number of times that each information field contained in the resource information of the external resources changes in the past historical period.

[0053] For example, in one example, the historical number of times each information field contained in the resource information of an external resource has changed in the past historical period can be collected to build a time series model, and then based on the time series model, the number of times each information field contained in the resource information of the external resource has changed within a preset time period in the future can be predicted.

[0054] When performing weighted calculation on the first indicator, the second indicator and the third indicator, the weighting coefficient of the third indicator can also be flexibly set based on specific needs and will not be specifically limited in this specification.

[0055] In one embodiment shown, the weighted coefficient of the third indicator may be a weighted coefficient calculated based on order information of orders for failed transactions for external resources due to changes in resource information.

[0056] For instance, in one example, the order information may specifically be the total order amount of orders for external resources whose transactions failed due to changes in resource information, and the weighting coefficient of the third indicator may be the weighting coefficient calculated based on the total order amount of orders for external resources whose transactions failed due to changes in resource information.

[0057] Among them, it should be noted that the specific calculation method of the weighted coefficient of the third indicator based on the total order amount of orders for failed transactions for external resources due to changes in resource information is no longer specifically limited in this specification. In actual applications, the calculation method can be flexibly constructed based on specific needs.

[0058] For example, in one example, the weighted coefficient of the third indicator can be calculated using the following expression: K=

[0059] In the above expression, K represents the weighting coefficient of the above third indicator; N represents the order quantity of the external resource; T represents the total order amount of the orders for the external resource that failed due to the change of resource information; M represents the total order amount of the orders for the external resource that succeeded.

[0060] Among them, the mapping method adopted by the above-mentioned resource publishing platform when further mapping the calculated resource value scores of each external resource into scheduling priorities is no longer specifically limited in this specification. In actual applications, the appropriate mapping strategy can be flexibly selected based on specific needs.

[0061] In one embodiment shown, the resource publishing platform can specifically map the resource value score of the external resource into a scheduling priority by comparing the resource value score of the external resource with the maximum and minimum values ​​of the resource value scores of all external resources in a set of external resources that can be provided by the external resource provider.

[0062] For example, in one example, the resource publishing platform may further map the calculated resource value score of the external resource into a scheduling priority using the following expression: P=

[0063] In this expression, P represents the scheduling priority of the external resource; S represents the resource value score of the external resource; min represents the minimum value of the resource value scores of all external resources in the Shanghai Su external resource set; max represents the maximum value of the resource value scores of all external resources in the above external resource set; g represents the number of priority levels preset for the above scheduling priority (i.e., the number of priority levels). It should be noted that the number of priority levels preset for the above scheduling priority can be flexibly set based on specific needs in actual applications. For example, the number of scheduling priority levels can be divided based on the total number of external resources in the above external resource set.

[0064] Among them, the scheduling method adopted by the above-mentioned resource publishing platform to periodically schedule external resources in the above-mentioned external resource set based on the scheduling priority of each external resource is no longer specifically limited in this specification. In actual application, the appropriate scheduling strategy can be flexibly selected based on specific needs.

[0065] In one embodiment shown, in order to improve the efficiency of resource scheduling, the above-mentioned resource publishing platform may organize the external resources in the above-mentioned external resource set into the form of a binary tree, and complete the resource scheduling of the external resources in the above-mentioned external resource set by traversing the external resources on the binary tree.

[0066] In this case, when the above-mentioned resource publishing platform periodically schedules external resources in the above-mentioned external resource set based on the scheduling priority, it can pre-create a scheduling node corresponding to each external resource in the external resource set; wherein, the information recorded in the scheduling node can specifically include the scheduling priority of the external resource.

[0067] For example, in one example, the information recorded in the scheduling node may include, in addition to the scheduling priority of the external resource, in actual applications, also the resource information of the external resource, the number of times the external resource is scheduled, the timestamp corresponding to the last time it was scheduled, and so on.

[0068] After creating the scheduling nodes corresponding to each external resource in the external resource set, the scheduling priority contained in the information recorded in each scheduling node can be used as a sorting key to construct the scheduling nodes corresponding to all external resources in the external resource set into a binary tree.

[0069] When the binary tree is constructed, the resource publishing platform can periodically traverse the binary tree to search for at least one target scheduling node with the highest scheduling priority recorded in the binary tree; for example, the N scheduling nodes with the highest scheduling priority recorded in the binary tree can be searched; N is a threshold value that can be flexibly set based on specific needs.

[0070] Then, the external resource corresponding to the at least one target scheduling node found can be determined as the external resource to be released, and resource information of the external resource corresponding to the at least one target scheduling node found can be obtained from the external resource provider.

[0071] The specific type of the binary tree is not specifically limited in this specification; for example, in practical applications, a red-black tree may be used. The red-black tree is a self-balancing binary tree. The specific process of constructing the scheduling nodes corresponding to all external resources in the external resource set into the red-black tree is not described in detail in this specification, and those skilled in the art may refer to the records in the relevant technology.

[0072] In an embodiment shown, in order to further improve the efficiency of resource scheduling, the scheduling priority of the external resources in the external resource set may be further mapped into a scheduling weight value, and the scheduling weight value is used as a sorting key to construct a binary tree.

[0073] It should be noted that, since the above-mentioned scheduling priority usually presets several different priority levels, in practical applications, different scheduling weight values ​​can be set for different priority levels. At this time, when the scheduling priority of the external resources in the above-mentioned external resource set is further mapped into a scheduling weight value, the priority level of the scheduling priority of the external resources in the above-mentioned external resource set calculated can be determined first, and then the scheduling weight value set for the priority level can be determined as the scheduling weight value corresponding to the scheduling priority.

[0074] In this case, the information recorded in the above-mentioned scheduling node may also include a scheduling weight value corresponding to the priority level of the scheduling priority of each external resource; wherein the scheduling priority of the external resource may be positively correlated with the scheduling weight value. When constructing a binary tree, the above-mentioned resource publishing platform may specifically use the scheduling weight value obtained by mapping the scheduling priority of the external resource as a sorting key, and construct the scheduling nodes corresponding to all external resources in the above-mentioned external resource set into a binary tree.

[0075] When the binary tree is constructed, the resource publishing platform can periodically traverse the binary tree to search for at least one target scheduling node with the highest scheduling weight value recorded in the binary tree; then, the external resources corresponding to the at least one target scheduling node found can be determined as the external resources to be released, and the resource information of the external resources corresponding to the at least one target scheduling node found can be obtained from the external resource provider.

[0076] By further mapping the scheduling priority of external resources into a scheduling weight value corresponding to the priority level of the scheduling priority, and performing resource scheduling based on the scheduling weight value, resource scheduling is no longer performed in units of external resources, but rather external resources are scheduled in batches in units of priority levels, thereby further improving the efficiency of resource scheduling.

[0077] For example, since the above scheduling weight values ​​usually correspond to a class of external resources with a scheduling priority at a certain priority level, resource scheduling based on the above scheduling weight values ​​is equivalent to batch resource scheduling for a class of external resources with a scheduling priority at that priority level.

[0078] In one embodiment shown, after the resource publishing platform obtains the resource information of the external resources corresponding to the at least one target scheduling node found from the external resource provider, in order to achieve balanced resource scheduling and avoid repeated scheduling of the same external resource, the scheduling priority contained in the information recorded in the at least one target scheduling node can also be updated based on a preset scheduling balancing strategy.

[0079] On the one hand, if the sorting key of the binary tree is the scheduling priority, in this case, after the resource publishing platform obtains the resource information of the external resource corresponding to the at least one target scheduling node found from the external resource provider, it can update the scheduling priority recorded in the at least one target scheduling node based on the preset scheduling balancing strategy. Then, the position of the at least one target scheduling node in the binary tree can be re-determined based on the updated scheduling priority, and the at least one target scheduling node can be reinserted into the binary tree according to the re-determined position, waiting for the execution of the subsequent round of scheduling.

[0080] On the other hand, if the sorting key of the binary tree is the scheduling weight value, in this case, after the resource publishing platform obtains the resource information of the external resource corresponding to the at least one target scheduling node found from the external resource provider, it can update the scheduling priority recorded in the at least one target scheduling node based on the preset scheduling balancing strategy, and synchronously update the scheduling weight value recorded in the at least one target scheduling node based on the updated scheduling priority; for example, in the same manner as above, re-map the updated scheduling priority to the scheduling weight value. Then, the position of the at least one target scheduling node in the binary tree can be re-determined based on the updated scheduling weight value, and the at least one target scheduling node can be reinserted into the binary tree according to the re-determined position, waiting for the execution of subsequent rounds of scheduling.

[0081] The above-mentioned scheduling balance strategy may specifically include any form of update strategy that can achieve balanced scheduling of external resources.

[0082] For example, the scheduling priority of the external resource can be uniformly increased or decreased in an increasing or decreasing direction. For example, in one example, after the resource publishing platform obtains the resource information of the external resource corresponding to the at least one target scheduling node found from the external resource provider, the scheduling priority of the external resource corresponding to the at least one target scheduling node can be uniformly increased by the same value based on the priority value of the original scheduling priority.

[0083] In this way, each time the resource publishing platform obtains the resource information of the external resources corresponding to the at least one target scheduling node found from the external resource provider, the scheduling priority of the external resources corresponding to the at least one target scheduling node will increase or decrease evenly, so that in the next round of periodic scheduling, other external resources with the highest priority or scheduling weight value can be scheduled, so that the same external resource can be repeatedly scheduled, and balanced scheduling of external resources can be achieved.

[0084] Step 104: Generate a prompt word corresponding to the conversion task based on the resource information of the external resource, and input the prompt word into the LLM model, so that the LLM model performs an inference calculation corresponding to the conversion task based on the prompt word to convert the external resource into the standard resource; After the resource publishing platform obtains the resource information of the external resource to be published, it can generate a prompt corresponding to the above conversion task based on the resource information of the external resource, and then input the prompt into the LLM model. After receiving the input prompt, the LLM model can perform the inference calculation corresponding to the conversion task based on the prompt to convert the external resource into a standard resource.

[0085] It should be noted that the content of the prompt words contained in the generated prompt words is not particularly limited in this specification, and in practical applications, it can be flexibly defined based on specific needs.

[0086] In one embodiment shown, in order to cope with the problem that a longer task context may cause the reasoning ability of the LLM model to deteriorate when the LLM model performs reasoning calculations on some complex tasks, on the one hand, the execution link of the conversion task can be split in accordance with the process of manually executing the conversion task, and the conversion task can be split into multiple subtasks. In this case, the conversion task can specifically include multiple subtasks that are executed in sequence; on the other hand, the execution of the multiple subtasks can also be driven by event triggering, and the multiple subtasks can be linked into a complete execution link. In this case, the multiple subtasks contained in the conversion task can specifically be tasks based on event triggering execution.

[0087] In this case, when the LLM model is used to perform the conversion task, a first prompt word corresponding to the first subtask included in the conversion task can be generated based on the resource information of the external resource to be released; wherein the first subtask can specifically refer to the first subtask among the multiple subtasks included in the conversion task. Then, the first prompt word can be input into the LLM model, so that the LLM model performs the reasoning calculation corresponding to the first subtask based on the first prompt word; When the first subtask is completed, the first execution result corresponding to the first subtask output by the LLM model can be obtained, and a first event for triggering the execution of the second subtask included in the conversion task can be generated; wherein the first subtask can specifically refer to the next subtask of the above-mentioned first subtask. Then, it can be determined whether the first event meets the trigger condition; if the first event meets the trigger condition, the first event can be executed at this time, and a second prompt word corresponding to the second subtask can be further generated based on the first execution result, and the second prompt word is input into the LLM model, so that the LLM model can perform the reasoning calculation corresponding to the second subtask based on the second prompt word.

[0088] When the second subtask is completed, the same process can be executed to continue to obtain the second execution result corresponding to the second subtask, and repeat the above process until all subtasks included in the conversion task are completed and the external resource is successfully converted into a standard resource supported by the resource publishing platform, and the above execution process is stopped.

[0089] In the above embodiment, since the execution link of the above-mentioned conversion task is split into multiple sub-tasks, and the multiple sub-tasks are linked into a complete execution link through event triggering, the longer task context of the conversion task can be split into a form that is easier for the LLM model to understand for segmented execution. Therefore, in this way, on the one hand, the problem of the poor reasoning ability of the LLM model for tasks with longer contexts can be overcome, thereby improving the accuracy of the reasoning results of the LLM model; on the other hand, some general LLM models that have not been fine-tuned and do not have professional business knowledge can also be used to execute the above-mentioned conversion tasks, thereby reducing the implementation cost of the above-mentioned conversion tasks and improving the versatility of the overall solution.

[0090] It should be noted that since the first subtask included in the above-mentioned conversion task is the first subtask in the conversion task, the first subtask can usually be manually triggered and executed by the initiator of the conversion task, and the subtasks after the first subtask can be triggered and executed in the aforementioned manner based on the event generated after the previous subtask is executed.

[0091] In actual applications, when the resource publishing platform drives the execution of the multiple subtasks by means of event triggering, it can specifically adopt an event-driven task execution framework to achieve it.

[0092] See also Figure 3 , Figure 3 This is a schematic diagram of an event-driven task execution framework shown in this specification.

[0093] like Figure 3As shown, in the task execution framework, the event producers, event routers, event consumers, and event triggers can be included.

[0094] The event producer is specifically used to generate events for triggering task execution.

[0095] Event routing, which is used to pass events from event producers to appropriate consumers.

[0096] Event consumers are used to execute events that meet the triggering conditions.

[0097] Event triggers are used to detect whether the events generated by the generator meet the trigger conditions.

[0098] For example, Figure 3 As shown in the figure, in actual applications, event triggers can manage events generated by event generators in a publish-subscribe mode, and event consumers can subscribe to events generated by specific event generators through the subscription mechanism. When the event trigger detects that an event generated by an event generator meets the trigger condition, the event that meets the trigger condition can be accurately distributed to the event consumer who subscribed to the event through event routing based on the distribution mechanism, and the event consumer will execute it.

[0099] It should be noted that, in actual applications, the software forms of the above-mentioned event producers, event routers, event consumers and event triggers are not specifically limited in this specification.

[0100] For example, the above-mentioned event trigger can usually be an independent background service program on the service platform, running as an independent daemon process or microservice; the above-mentioned event generator can usually be embedded in the business logic of the application in the form of an integrated component; the above-mentioned event router can usually run in the form of middleware; and the above-mentioned event consumer can usually run in the form of a microservice or an independent thread.

[0101] In this case, when the above-mentioned first subtask is executed, the above-mentioned event producer can obtain the first execution result corresponding to the first subtask output by the LLM model, and generate a first event based on the obtained first execution result to trigger the execution of the second subtask included in the conversion task; wherein, the obtained first execution result can be included in the first event.

[0102] Then, the above-mentioned event trigger can determine whether the first event satisfies the trigger condition; if the first event satisfies the trigger condition, the above-mentioned event trigger can, based on the distribution mechanism, accurately distribute the event that meets the trigger condition to the event consumer that has subscribed to the event through event routing, and the event consumer executes the first event, and further generates a second prompt word corresponding to the second subtask based on the above-mentioned first execution result contained in the first event, and inputs the second prompt word into the LLM model, so that the LLM model can perform the reasoning calculation corresponding to the second subtask based on the second prompt word.

[0103] After the second subtask is executed, the above-mentioned event producer can continue to obtain the second execution result corresponding to the second subtask, and so on, repeating the above process until all subtasks included in the conversion task are executed, and the above execution process is stopped when the external resource is successfully converted into a standard resource supported by the resource publishing platform.

[0104] Among them, in one embodiment shown, since the above-mentioned conversion task contains multiple subtasks, different reasoning capabilities of the LLM model may be applied; therefore, in order to make full use of the differences in reasoning capabilities of different LLM models, each subtask contained in the above-mentioned conversion task can be executed using different LLM models.

[0105] It should be noted that the triggering conditions for the events mentioned above for triggering the execution of the subtasks included in the above conversion tasks are not specifically limited in this specification. In practical applications, the triggering conditions for the events can be flexibly defined based on specific needs.

[0106] In an illustrated embodiment, the triggering condition for triggering the execution of an event of any subtask included in the above conversion task may specifically include that the execution result of the previous subtask of the subtask is confirmed by the user.

[0107] For example, taking the above-mentioned first event as an example, since the first event is generated after the execution of the first subtask is completed, and is an event used to trigger the execution of the next second subtask of the first subtask, therefore, in this case, when determining whether the first event meets the triggering condition, it can be specifically determined whether the user has confirmed the first execution result; if the user has confirmed the first execution result, it can be determined that the first event meets the triggering condition.

[0108] The specific method for the user to confirm the execution result of the subtask is not particularly limited in this specification.

[0109] For example, the execution result output by the LLM model can be output and displayed to the user through a preset display interface; the display interface can specifically include a user option (such as a user button) for confirming the execution result, and the user can manually confirm the execution result output by the LLM model by operating the user option in the display interface.

[0110] In one embodiment shown, if the user does not confirm the execution result of a subtask included in the above-mentioned conversion task, it indicates that the user is not satisfied with the output result of the subtask. In this case, a mechanism for iteratively optimizing the prompt words of the subtask can be introduced to guide the LLM model to output an execution result that satisfies the user.

[0111] Among them, the mechanism for iteratively optimizing the prompt words of the subtask specifically refers to iteratively optimizing and adjusting the prompt words generated for the subtask based on a preset optimization and adjustment strategy, until the LLM model outputs a satisfactory execution result based on the optimized and adjusted prompt words for manual confirmation by the user.

[0112] For example, taking the first subtask included in the above-mentioned conversion task as an example, if the user has not confirmed the first execution result of the first subtask, the first prompt word generated is optimized and adjusted based on the preset optimization and adjustment strategy, and the optimized and adjusted first prompt word is re-input into the LLM model, so that the LLM model can perform the reasoning calculation corresponding to the first subtask based on the first prompt word; then, the first execution result corresponding to the first subtask output by the LLM model can be obtained, and it is re-determined whether the user has confirmed the first execution result; if the user still has not confirmed the first execution result, the generated first prompt word can continue to be optimized and adjusted based on the preset optimization and adjustment strategy, and so on, and the above process is repeated until the user confirms the first execution result corresponding to the first subtask output by the LLM model.

[0113] It should be noted that the optimization and adjustment strategy adopted for optimizing and adjusting the prompt words mentioned above is not particularly limited in this specification, and in actual application, it can be flexibly configured based on specific needs.

[0114] In one embodiment shown, in order to avoid the situation where the LLM model still cannot output execution results that satisfy the user for a long time after repeated iterative optimization and adjustment of the prompt words, a mechanism for the user to manually modify the execution results output by the LLM model can also be introduced.

[0115] In this case, the display interface may further include a user option (such as a drop-down list) for modifying the execution result, and the user may manually modify the execution result output by the LLM model by operating the user option in the display interface. The resource publishing platform may obtain the update data corresponding to the execution result output by the LLM model input by the user in response to the user operation on the user option, and then modify and replace the execution result based on the update data.

[0116] It should be noted that the splitting method used when splitting the execution link of the above-mentioned conversion task is not specifically limited in this specification. In actual applications, the above-mentioned conversion task can be flexibly split based on specific needs. Correspondingly, after the conversion task is split, the task objectives of the multiple subtasks contained in the conversion task usually depend on the actual splitting method adopted, which is not specifically limited in this specification.

[0117] The following is an explanation using a specific example of splitting the above conversion task.

[0118] See also Figure 4 , Figure 4 This is a flowchart of converting external resources into standard resources shown in this specification.

[0119] like Figure 4 As shown, in an embodiment shown, after the above conversion task is split, the subtasks included in the above conversion task may specifically include Figure 4 The four subtasks shown are resource allocation subtask (i.e., the first subtask), resource matching subtask (i.e., the second subtask), information extraction subtask (i.e., the third subtask), and detail map production subtask (i.e., the fourth subtask).

[0120] The task objective of the resource classification subtask is specifically used to determine the resource type of the external resource to be released; The task objective of the resource matching subtask is specifically to match the resource information of the external resource with the resource information of the standard resource corresponding to the resource type of the external resource stored in the standard resource database, so as to establish a corresponding relationship between the information items contained in the resource information of the external resource and the standard information items contained in the resource information of the standard resource; The task objective of the above-mentioned information extraction subtask is specifically used for a third subtask of extracting rule information related to the above-mentioned standard resource from the resource information of the external resource; The task objective of the above-mentioned detail map production subtask is specifically used to generate a standard resource detail map corresponding to the external resource based on the resource information of the external resource, the above-mentioned corresponding relationship that has been established, and the above-mentioned extracted rule information.

[0121] Please continue to see Figure 4 ,exist Figure 4 In the process shown, it is still possible to use Figure 3 The event-driven task execution framework shown drives the LLM model to execute the multiple subtasks by event triggering.

[0122] In addition, in the process of executing the multiple subtasks through the event-driven LLM model, whether the user has confirmed the execution result of a certain subtask that has been executed can be used as the trigger condition for the event of driving the next subtask of the subtask. At the same time, a mechanism for iteratively optimizing the prompt words of each subtask can be introduced to guide the LLM model to output an execution result that satisfies the user, and then apply the execution result.

[0123] In this case, when using the above-mentioned LLM model to perform the above-mentioned conversion task, first, based on the resource information of the external resources to be released, a first prompt word corresponding to the above-mentioned resource classification subtask can be generated, and the first prompt word can be input into the LLM model, so that the LLM model can perform the reasoning calculation corresponding to the first subtask based on the first prompt word to determine the resource type of the external resource.

[0124] For example, in actual applications, the resource information of the external resource and the category specification corresponding to the standard resource defined by the platform can be used as the content of the first prompt word to generate the first prompt word. The LLM model can clarify the resource type of the external resource based on matching the resource information of the external resource contained in the prompt word with the category specification.

[0125] When the resource classification subtask is completed, the resource type of the external resource output by the LLM model can be output and displayed to the user through a preset display interface.

[0126] Specifically, the display interface may include a user option (such as a user button) for confirming the execution result, and the user may manually confirm the execution result output by the LLM model by operating the user option in the display interface.

[0127] For example, taking the ticket products provided by the external product provider as the above resources, the product information of the above external products that needs to be obtained may specifically include information such as the name of the ticket product, the name of the attraction, the country to which the attraction belongs, the city to which the attraction belongs, the address of the attraction, the attraction fee, etc. The product type of the external product determined by the LLM model based on this information can be output and displayed in the display interface. In addition, the classification reasons for the LLM model to classify the external product can also be output and displayed in the display interface for user reference. When the user confirms that the product type of the external product is the correct product type supported by the product publishing platform, the above user options can be operated to manually confirm the execution results output by the LLM model.

[0128] Of course, if the product type of the external product output by the LLM model is not the correct product type, the first prompt word can be iteratively optimized based on the above-mentioned mechanism for iterative optimization of the prompt word, until the LLM model outputs a satisfactory execution result based on the optimized and adjusted prompt word for manual confirmation by the user.

[0129] After the resource classification subtask is completed, the resource type of the external resource output by the LLM model can be obtained, and a first event for triggering the resource matching subtask can be generated; then, it can be determined whether the user has confirmed the resource type of the external resource output by the LLM model; if yes, the first event can be executed at this time, and a second prompt word corresponding to the resource matching subtask can be generated based on the resource information of the external resource and the resource type of the external resource; For example, the resource information of the external resource and the resource type of the external resource may be used as the content of the second prompt word to generate the second prompt word.

[0130] Then, the second prompt word can be input into the LLM model, so that the LLM model can perform the inference calculation corresponding to the above-mentioned resource matching subtask based on the second prompt word, and match the resource information of the external resource with the resource information of the standard resource corresponding to the resource type stored in the standard resource database, so as to establish a correspondence between the information items contained in the resource information of the external resource and the standard information items contained in the resource information of the standard resource.

[0131] Of course, if the above correspondence output by the LLM model is not a correct correspondence, the above second prompt word can be iteratively optimized based on the above mechanism for iterative optimization of the prompt word, until the LLM model outputs a satisfactory execution result based on the optimized and adjusted prompt word for manual confirmation by the user.

[0132] In one embodiment shown, the resource information of the external resource may specifically include multiple information items; in this case, multiple matching prompt words corresponding to the multiple information items may be generated based on the resource information of the external resource and the resource type of the external resource.

[0133] Then, the generated multiple matching prompt words can be input into the LLM model respectively, so that the LLM model can perform the reasoning calculation corresponding to the above-mentioned resource matching subtask based on the input matching prompt words, and match the information item corresponding to the matching prompt word contained in the resource information of the external resource with the standard information item contained in the resource information of the standard resource corresponding to the resource type stored in the standard resource database, and establish the correspondence between the information item and the standard information item based on the matching result.

[0134] For example, still taking the ticket products provided by the above-mentioned resources as an external product provider as an example, the product information of the above-mentioned external products may specifically include information items such as attractions, charging item names, ticket types, sessions, and regions. However, since these information items included in the product information of the external products usually do not conform to the information specifications of the above-mentioned product publishing platform for standard products, it is necessary to match the information items such as attractions, charging item names, ticket types, sessions, and regions contained in the product information of the external products with the standard information items of ticket products stored in the standard product database, so as to establish a correspondence between these non-standard information items contained in the product information of the external products and the standard information items of ticket products stored in the standard product database.

[0135] In one embodiment shown, before generating the second prompt word corresponding to the above-mentioned resource matching subtask, it is also possible to pre-calculate the similarity between the information items contained in the resource information of the external resource and the standard information items contained in the resource information of the standard resource corresponding to the resource type of the external resource stored in the standard resource database, and obtain at least one standard information item stored in the standard resource database that has the highest similarity with the information items contained in the resource information of the external resource.

[0136] For example, in one example, the above similarity can be specifically represented by vector distance. In this case, when calculating the similarity between the information items contained in the resource information of the external resource and the standard information items contained in the resource information of the standard resource corresponding to the resource type of the external resource stored in the standard resource database, the information items contained in the resource information of the external resource can be specifically converted into an information vector, and the standard information items contained in the resource information of the standard resource corresponding to the resource type of the external resource stored in the standard resource database can also be converted into a standard information vector, and then the vector distance between the information vector and the standard information vector can be calculated, and at least one standard information vector with the largest vector distance between the information vectors corresponding to the information items contained in the resource information of the external resource stored in the standard resource database can be determined; finally, at least one standard information item corresponding to at least one standard information vector stored in the standard resource database with the largest vector distance can be obtained as the at least one standard information item with the highest similarity.

[0137] In this case, when generating the second prompt word corresponding to the resource matching subtask, the second prompt word corresponding to the resource matching subtask may be generated based on the resource information of the external resource and the pre-screened at least one standard information item with the highest similarity. For example, the resource information of the external resource and the at least one standard information item with the highest screened similarity may be used as the content of the second prompt word to generate the second prompt word.

[0138] Then, the second prompt word can be input into the LLM model, so that the LLM model can perform the inference calculation corresponding to the above-mentioned resource matching subtask based on the second prompt word, match the information items contained in the resource information of the external resource with the above-mentioned at least one standard information item contained in the prompt word respectively, and establish the correspondence between the information items contained in the resource information of the external resource and the above-mentioned at least one standard information item based on the matching result.

[0139] In this way, at least one pre-screened standard information item with the highest similarity can be directly added to the prompt word, and the matching scope of the LLM model can be pre-defined, which can help improve the accuracy of the matching results output by the LLM model.

[0140] After the resource matching subtask is completed, the corresponding relationship output by the LLM model can be further obtained, and a second event for triggering the execution of the information extraction subtask can be generated; then, it can be determined whether the user has confirmed the corresponding relationship; if yes, the second event can be executed at this time, and a third prompt word corresponding to the information extraction subtask can be further generated based on the resource information of the external resource; For example, the resource information of the external resource may be used as the content of the third prompt word to generate the third prompt word.

[0141] Then, the third prompt word can be input into the LLM model, so that the LLM model can perform the inference calculation corresponding to the above-mentioned information extraction subtask based on the third prompt word, and extract the rule information related to the standard resource from the resource information of the external resource.

[0142] Of course, if the rule information output by the LLM model is not the correct rule information, the third prompt word can be iteratively optimized based on the above-mentioned mechanism for iterative optimization of the prompt word, until the LLM model outputs a satisfactory execution result based on the optimized and adjusted prompt word for manual confirmation by the user.

[0143] The rule information may specifically include any form of rules related to standard resources.

[0144] For example, still taking the above-mentioned resources as goods provided by external product providers as an example, the above-mentioned rule information may include product rules; for example, taking the above-mentioned goods as tickets as an example, the above-mentioned product rules may specifically include any one or more combinations of the following: reservation rules, cancellation rules, refund rules, purchase limit rules, and usage rules.

[0145] After the above-mentioned information extraction subtask is executed, the above-mentioned rule information output by the LLM model can continue to be obtained, and a third event for triggering the execution of the above-mentioned detail map production subtask can be generated; then, it can be determined whether the user has confirmed the above-mentioned extracted rule information; if so, the third event can be executed at this time, and further based on the resource information of the external resource, the above-mentioned corresponding relationship established (that is, the execution result of the above-mentioned resource matching subtask), and the above-mentioned rule information extracted (that is, the execution result of the above-mentioned extraction subtask), a fourth prompt word corresponding to the above-mentioned detail map production subtask is generated.

[0146] For example, the resource information of the external resource, the established corresponding relationship, and the extracted rule information may be used as the content of the fourth prompt word to generate the fourth prompt word.

[0147] Then, the fourth prompt word can be input into the LLM model, so that the LLM model can perform the reasoning calculation corresponding to the fourth subtask based on the fourth prompt word, and utilize the image generation capability of the LLM model to generate a standard resource details map corresponding to the above external resource based on the resource information of the external resource, the above-mentioned corresponding relationship established, and the above-mentioned rule information extracted.

[0148] Among them, the LLM model uses its own image generation capability to generate a standard resource detail map corresponding to the above external resource based on the resource information of the external resource, the above correspondence established and the above rule information extracted, which will not be described in detail in this manual.

[0149] For example, in one example, still taking the ticket products provided by the above-mentioned resources as an external product provider as an example, when generating a standard product details map based on the product information of the ticket products, the above-mentioned corresponding relationship established, and the above-mentioned rule information extracted, in one case, the standard information item contained in the above-mentioned corresponding relationship can be used as the key, and the non-standard information item corresponding to the standard information item contained in the product information of the ticket products can be used as the value corresponding to the key, and the association display is performed in the generated standard resource details map; or, in another case, the non-standard information item contained in the product information of the ticket products can be directly replaced with the standard information item corresponding to the non-standard information item, so as to convert the product information contained in the ticket products into a standard product that conforms to the information format supported by the above-mentioned product publishing platform. As for the above-mentioned rule information, since it is itself a product rule related to the standard product, these product rules can be directly displayed in the standard resource details map.

[0150] It should be noted that the specific layout of the various information contained in the above-mentioned standard resource details map finally generated on the details map will no longer be specifically limited in this specification. In practical applications, it can be flexibly designed according to specific needs.

[0151] When using the LLM model's own image generation capabilities to generate a standard resource details map corresponding to the above external resource based on the resource information of the external resource, the above-mentioned correspondence established, and the above-mentioned rule information extracted, since the richness of resource information of different external resources may be significantly different, if a unified standard resource details map is generated according to a unified standard resource details map structure, it may be impossible to generate a standard resource details map that can reasonably display the resource information of the external resource.

[0152] For example, if the structure of the standard resource details map is too complex, a lot of information on the details map may be missing; and if the structure of the standard resource details map is too simple, it will be difficult to reflect the real information of the external resources.

[0153] Based on this, in one embodiment shown, when using the LLM model to utilize its own image generation capabilities, based on the resource information of the external resource, the established corresponding relationship, and the extracted rule information, to generate a standard resource details map corresponding to the external resource, a unified standard resource details map structure may no longer be used, but based on the amount of information actually contained in the resource information of the external resource, a standard resource details map with a higher degree of freedom may be adaptively generated for the external resource. That is, for different external resources, a standard resource details map with a completely consistent structure may be generated, but a standard resource details map that can cover the resource information of the external resource may be generated based on the amount of information actually contained in different external resources.

[0154] In this way, a standard resource detail map that better meets user needs can be generated for users.

[0155] For example, taking the above-mentioned resources as commodities, a unified structure is no longer used. Instead, a product detail picture with a higher degree of freedom is adaptively generated for the external commodity based on the actual amount of information contained in the external commodity. This can obviously stimulate the user's desire to buy and reduce pre-sales and after-sales communication costs.

[0156] Step 106: Acquire the standard resource output by the LLM model, and publish the standard resource on the resource publishing platform.

[0157] After the external resources are converted into standard resources based on the LLM model, the resource publishing platform can obtain the converted standard resources output by the LLM model, and then publish the standard resources on the resource publishing platform.

[0158] For example, the standard resource detail diagram corresponding to the external resources output by the LLM model can be published on the resource publishing platform.

[0159] In one embodiment shown, the LLM model accessed by the resource publishing platform may specifically be a general LLM model; the resource publishing platform may specifically be equipped with a task execution framework based on the general LLM model.

[0160] Specifically, the task execution framework can adopt a modular hierarchical structure. Through clear hierarchical division and functional modularization, the task execution framework can be highly efficient, scalable and flexible. In the task execution framework adopting a modular hierarchical structure, the multiple subtasks included in the above conversion task can usually be functional modules defined by the user in the task execution framework.

[0161] See also Figure 5 , Figure 5This is an architecture diagram of a task execution framework based on a general LLM model shown in this specification; like Figure 5 As shown, the task execution framework can specifically adopt a modular hierarchical structure, which can specifically include an interface layer, a data processing layer, a task processing layer, a functional module layer and a bottom support layer. In addition, the task execution framework can also provide an independent tool module for users.

[0162] The interface layer is used to provide an API interface related to task execution.

[0163] like Figure 5 As shown, the interfaces provided by the interface layer may specifically include APIs such as an API for viewing a task list, an API for viewing task details, an API for triggering task execution, an API for manually modifying the interface output by the LLM model, an API for viewing the task execution results output by the LLM model, etc. In the process of executing the above-mentioned conversion tasks based on the task execution framework, the user can implement the service functions related to task execution by calling these API interfaces provided by the interface layer.

[0164] For example, Figure 5 As shown, when the user is not satisfied with the execution result output by the LLM model, the user can call the above-mentioned API for manually modifying the interface output by the LLM model by operating on the display interface for displaying the execution result, and then manually modify the execution result output by the LLM model in the display interface.

[0165] The above data processing layer is specifically used to provide data processing capabilities related to task execution.

[0166] like Figure 5 As shown, the data processing layer can specifically provide the ability to pre-process data related to task execution, the ability to convert data related to task execution, the ability to store data related to task execution, and so on.

[0167] For example, as mentioned in the previous embodiment, before generating the second prompt word corresponding to the resource matching subtask, the similarity between the information items contained in the resource information of the external resource and the standard information items contained in the resource information of the standard resource corresponding to the resource type of the external resource stored in the standard resource database can be pre-calculated, and the processing operation of obtaining at least one standard information item stored in the standard resource database with the highest similarity to the information items contained in the resource information of the external resource is to utilize the data preprocessing capability provided by the data processing layer. As mentioned in the previous embodiment, the processing operation of converting the information items contained in the resource information of the external resource into information vectors is to utilize the data conversion capability provided by the data processing layer.

[0168] The above-mentioned task processing layer can be specifically used to provide task execution functions related to task execution.

[0169] like Figure 5 As shown, the task execution functions related to task execution provided by the above-mentioned task processing layer may specifically include a task scheduling function for performing task scheduling, a task execution engine for performing tasks, and a task status management function for performing task status management, etc.

[0170] The above-mentioned functional module layer is the core layer of the above-mentioned task execution framework, and is used to provide functional modules corresponding to the multiple subtasks included in the above-mentioned conversion task.

[0171] In actual applications, after splitting the execution link of the conversion task into multiple subtasks by imitating the process of manually executing the conversion task, the functional modules corresponding to the subtasks included in the above-mentioned conversion task can be defined on the functional module based on the expansion capability of the functional module layer. Through the functional module layer, the above-mentioned task execution framework can support different types of functional modules and quickly access new functional modules by utilizing the expansion capability of the functional module layer.

[0172] When the execution link of the conversion task is split into multiple subtasks, and the execution results of the multiple subtasks do not meet the user's expectations, the user can also optimize and adjust the execution link of the conversion task, re-split the conversion task, and then define functional modules corresponding to the re-split subtasks on the functional module. And so on, the above optimization process can be iterated until the execution results of the multiple split subtasks meet the user's expectations.

[0173] like Figure 5As shown, taking the splitting of the above-mentioned conversion task into the above-mentioned resource allocation subtask, resource matching subtask, information extraction subtask and detail map production subtask as an example, at the functional module layer, corresponding functional modules can be defined for these four types of subtasks respectively.

[0174] The above underlying support layer is used to provide some underlying support related to task execution.

[0175] like Figure 5 As shown, the underlying support provided by the above underlying support layer may specifically include SQL, configuration management, vector database, etc.

[0176] The above tool module can be specifically used to provide some daily tools related to the task execution framework.

[0177] like Figure 5 As shown, the daily tools provided by the above tool module may specifically include log management tools, performance monitoring tools, cost assessment tools, result sampling tools, cache tools, prompt word maintenance tools, etc.

[0178] Based on the above task execution framework, a complete supply chain can be provided to convert external resources provided by external resource providers into standard resources and publish them on the resource publishing platform. When the resource publishing platform needs to publish external resources provided by external resource providers, no human intervention is required in the entire process. It only needs to remotely pull the resource information of the external resource from the external resource provider through API calls, and then it can automatically convert the external resource into a standard resource and publish it on the resource publishing platform, thereby significantly reducing the labor cost when publishing external resources and improving the efficiency of external resource publishing.

[0179] For example, still taking the above-mentioned resources as ticket products provided by external product providers, and the above-mentioned resource publishing platform as a travel transaction service platform as an example, based on the above-mentioned task execution framework, a complete product supply chain can be provided for converting ticket products provided by external product providers into standard ticket products in an information format supported by the travel transaction service platform, and publishing them on the resource publishing platform. When the travel transaction service platform needs to publish ticket products provided by external product providers, there is no need for manual participation in the entire process. It only needs to remotely pull the product information of the ticket products (such as the name of the ticket product, the name of the attraction, the country to which the attraction belongs, the city to which the attraction belongs, the address of the attraction, the attraction fee, etc.) from the external product provider through an API call, and then the ticket products can be automatically converted into standard ticket products and sold on the travel transaction service platform.

[0180] In the above technical scheme, the LLM model is introduced into the resource publishing platform to perform the conversion task of converting external resources into standard resources supported by the resource publishing platform. When publishing external resources, prompt words can be generated to guide the LLM model to convert external resources into standard resources. In this way, external resources can be automatically converted into standard resources for publishing without human intervention, thereby reducing labor costs and improving the efficiency of external resource publishing.

[0181] Corresponding to the embodiments of the aforementioned method, this specification also provides embodiments of an apparatus, an electronic device, and a storage medium.

[0182] Figure 6 is a schematic structural diagram of an electronic device provided by an exemplary embodiment. Figure 6 At the hardware level, the device includes a processor 602, an internal bus 604, a network interface 606, a memory 608, and a non-volatile memory 610, and may also include other required hardware. One or more embodiments of this specification may be implemented based on software, such as the processor 602 reading the corresponding computer program from the non-volatile memory 610 into the memory 608 and then running it. Of course, in addition to the software implementation, one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0183] like Figure 7 As shown, Figure 7 is a block diagram of a resource publishing device based on an LLM model according to an exemplary embodiment of the present specification, and the device can be applied to Figure 6 In the electronic device shown in the figure, the technical solution of this specification is implemented. Wherein, the resource publishing platform performs a conversion task of converting external resources into standard resources supported by the resource publishing platform based on the accessed LLM model; the device 1200 includes: Acquisition module 701, acquiring resource information of external resources to be published; A conversion module 702 generates a prompt word corresponding to the conversion task based on the resource information of the external resource, and inputs the prompt word into the LLM model, so that the LLM model performs an inference calculation corresponding to the conversion task based on the prompt word to convert the external resource into the standard resource; The publishing module 703 obtains the standard resource output by the LLM model, and publishes the standard resource on the resource publishing platform.

[0184] Accordingly, the present specification also provides an electronic device, which includes a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to implement the steps in all the method flows described above.

[0185] Accordingly, the present specification also provides a computer-readable storage medium on which executable computer program instructions are stored; wherein, when the instructions are executed by a processor, the steps in all the method flows described above are implemented.

[0186] Accordingly, the present specification also provides a computer program product having executable computer program instructions stored thereon; wherein, when the computer program instructions are executed by a processor, the steps in all the method flows described above are implemented.

[0187] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a server system. Of course, it is not excluded that with the development of computer technology in the future, the computer that implements the functions of the above embodiments may be, for example, a personal computer, a laptop computer, a vehicle-mounted human-computer interaction device, a cellular phone, a camera phone, a smart phone, 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.

[0188] Although one or more embodiments of the present specification provide method operation steps as described in the embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way of executing the order of many steps, and does not represent the only execution order. When the device or terminal product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment). The terms "include", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, product or device including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such a process, method, product or device. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or device including the elements. For example, if the words first, second, etc. are used to represent the name, they do not represent any specific order.

[0189] For the convenience of description, the above devices are described in various modules according to their functions. Of course, when implementing one or more of the present specification, the functions of each module can be implemented in the same or more software and / or hardware, or the module implementing the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0190] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0191] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0192] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0193] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0194] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0195] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. 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 disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage, graphene 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 temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0196] It should be understood by those skilled in the art that one or more embodiments of the present specification may be provided as a method, system or computer program product. Therefore, one or more embodiments of the present specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, one or more embodiments of the present specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0197] One or more embodiments of this specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of this specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0198] Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. In the description of this specification, the description of the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of this specification. In this specification, the schematic representation of the above terms does not necessarily target the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples without contradiction.

[0199] The above description is only an example of one or more embodiments of this specification and is not intended to limit one or more embodiments of this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this specification shall be included in the scope of the claims.

Claims

1. A resource publishing method based on an LLM model, applied to a resource publishing platform, wherein the resource publishing platform performs a conversion task of converting external resources into standard resources supported by the resource publishing platform based on the accessed LLM model; the method comprises: Get resource information of external resources to be published; Based on the resource information of the external resource, a prompt word corresponding to the conversion task is generated, and the prompt word is input into the LLM model, so that the LLM model performs an inference calculation corresponding to the conversion task based on the prompt word to convert the external resource into the standard resource; The standard resource output by the LLM model is obtained, and the standard resource is published on the resource publishing platform.

2. The method according to claim 1, obtaining resource information of the external resource to be published, comprising: Calculate the resource value score of each external resource in the external resource set provided by the external resource provider, and further map the calculated resource value score into a scheduling priority; wherein the resource value score represents the resource value of the external resource; and the scheduling priority is positively correlated with the resource value score; Periodically performing resource scheduling on the external resources in the external resource set based on the scheduling priority, so as to determine the external resources to be released from the external resource set; The determined resource information of the external resource to be published is obtained from the external resource provider.

3. The method according to claim 2, wherein calculating the resource value score of each external resource in the external resource set provided by the external resource provider comprises: Acquire multiple resource indicators of each external resource in the external resource set provided by the external resource provider; wherein the multiple resource indicators are indicators used to represent the resource value of the external resource; the multiple resource indicators include a first indicator used to represent the flow value of the external resource, and a second indicator used to represent the transaction value of the external resource; The first indicator and the second indicator are weightedly calculated to obtain a resource value score of each external resource in the external resource set.

4. The method according to claim 3, wherein the plurality of resource indicators further comprises a third indicator for indicating a frequency of change of resource information of the external resource; The first indicator and the second indicator are weightedly calculated to obtain a resource value score of each external resource in the external resource set, including: The first indicator, the second indicator and the third indicator are weightedly calculated to obtain a resource value score of each external resource in the external resource set.

5. The method according to claim 4, wherein the first indicator comprises the ratio of the number of visits to the external resource to the total number of visits to each external resource in the external resource set; the second indicator comprises the product of the number of orders for the external resource and the order amount of the external resource; the third indicator comprises a weighted sum obtained by weighted calculation of the number of times each information field contained in the resource information of the external resource changes within a preset time period in the future; in, The weighting coefficients of the first indicator and the second indicator are weighting coefficients calculated based on the resource conversion rate of the external resources; the weighting coefficient of the third indicator is a weighting coefficient calculated based on the order information of orders for failed transactions for the external resources due to changes in resource information.

6. The method of claim 2, further mapping the calculated resource value score into a scheduling priority, comprising: The calculated resource value score is further mapped to a scheduling priority based on the following expression: P= In this expression, P represents the scheduling priority of the external resource; S represents the resource value score of the external resource; min represents the minimum value of the resource value scores of all the external resources in the external resource set; max represents the maximum value of the resource value scores of all the external resources in the external resource set; g represents the number of priority levels preset for the scheduling priority.

7. The method according to claim 6, periodically performing resource scheduling on the external resources in the external resource set based on the scheduling priority to determine the external resources to be released from the external resource set, comprising: Creating a scheduling node corresponding to an external resource in the external resource set; wherein the information recorded in the scheduling node includes the scheduling priority of the external resource; Using the scheduling priority as a sorting key, constructing a binary tree of scheduling nodes corresponding to the external resources in the external resource set, and periodically traversing the binary tree to search for at least one target scheduling node with the highest scheduling priority; The external resource corresponding to the at least one found target scheduling node is determined as the external resource to be released.

8. The method according to claim 7, wherein the information recorded in the scheduling node further comprises a scheduling weight value corresponding to the priority level of the scheduling priority of the external resource; wherein, The scheduling priority of the external resource is positively correlated with the scheduling weight value; Using the scheduling priority as a sorting key, constructing the scheduling nodes corresponding to the external resources in the external resource set into a binary tree, and periodically traversing the binary tree to find at least one target scheduling node with the highest scheduling priority, including: Using the scheduling weight value as a sorting key, the scheduling nodes corresponding to the external resources in the external resource set are constructed into a binary tree, and the binary tree is periodically traversed to find at least one target scheduling node with the highest scheduling weight value.

9. The method of claim 1, wherein the conversion task comprises a plurality of subtasks executed based on event triggering; Based on the resource information of the external resource, a prompt word corresponding to the conversion task is generated, and the prompt word is input into the LLM model, so that the LLM model performs reasoning calculation based on the prompt word to convert the external resource into the standard resource, including: Based on the resource information of the external resource, generate a first prompt word corresponding to a first subtask included in the conversion task, and input the first prompt word into the LLM model so that the LLM model performs an inference calculation corresponding to the first subtask based on the first prompt word; Obtaining a first execution result corresponding to the first subtask output by the LLM model, and generating a first event for triggering the execution of a second subtask included in the conversion task; determining whether the first event satisfies a trigger condition; if the first event satisfies the trigger condition, executing the first event, further generating a second prompt word corresponding to the second subtask based on the first execution result, and inputting the second prompt word into the LLM model, so that the LLM model performs an inference calculation corresponding to the second subtask based on the second prompt word; Continue to obtain the second execution result corresponding to the second subtask, and repeat the above process until all subtasks included in the conversion task are executed, and the external resource is converted into the standard resource.

10. The method according to claim 9, wherein the triggering condition for triggering the event of executing any subtask included in the conversion task comprises that the execution result of the previous subtask of the subtask is confirmed by the user; Determining whether an event for triggering the execution of any subtask included in the conversion task satisfies a triggering condition includes: determining whether the user has confirmed the execution result of the previous subtask of any subtask included in the conversion task; If the user confirms the execution result of the previous subtask, it is determined that the event meets the trigger condition.

11. The method of claim 10, further comprising: If the user has not confirmed the execution result of the previous subtask, the generated prompt word corresponding to the previous subtask is optimized and adjusted based on a preset optimization and adjustment strategy, and the optimized and adjusted prompt word is re-input into the LLM model, so that the LLM model performs the reasoning calculation corresponding to the previous task based on the prompt word; and, the execution result corresponding to the previous subtask output by the LLM model is obtained, and it is determined whether the user has confirmed the execution result; the above process is repeated until the user confirms the execution result.

12. The method of claim 10, wherein the plurality of subtasks comprises: A first subtask for determining a resource type of the external resource; A second subtask for matching the resource information of the external resource with the resource information of the standard resource corresponding to the resource type stored in the standard resource database to establish a correspondence between the information items included in the resource information of the external resource and the standard information items included in the resource information of the standard resource; A third subtask for extracting rule information related to the standard resource from the resource information of the external resource; A fourth subtask is used to generate a standard resource detail map corresponding to the external resource based on the resource information of the external resource, the established corresponding relationship, and the extracted rule information.

13. According to the method as claimed in claim 12, the resource publishing platform includes a product publishing platform; the external resources include products provided by external product providers; the standard resources include products in an information format supported by the product publishing platform; and the rule information includes product rules.

14. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 13 are implemented.

15. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

Images