A large model application method and system for data retrieval

Abstract

The application provides a large model application method and system for data retrieval, and belongs to the technical field of data management, and specifically comprises the following steps: obtaining constituent data of a matching deviation target, combining matching data of different matching deviation targets and retrieval information items with a retrieval processing number, determining a matching update target of different cache information items according to different update targets and matching conditions of text vectors of the cache information items when it is determined that the update processing of the cache data of the large model does not need to adopt a preset update strategy, obtaining matching data of different cache information items and retrieval information items with a retrieval processing number, and combining a matching update target of different cache information items, a matching update target and a retrieval result matching condition with a retrieval processing number to determine an update processing strategy of different cache information items, thereby improving the timeliness of the update processing of the cache data of the large model.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data management, and particularly relates to a large model application method and system for data retrieval. BACKGROUND

[0002] With the improvement of digital governance capabilities of hospitals, schools and the like, how to realize the access to a large model so as to improve the retrieval capability in the fields of medical information consultation, academic research assistant, customer service intelligent support and the like becomes a technical problem to be solved urgently. In the invention patent application CN202410076898.4 “Unstructured data full-text retrieval method based on a large model”, the meanings of different context features are integrated together, the data retrieval content is more comprehensively converted into a data retrieval vector, different features of multi-modal data are uniformly represented, and matching is more accurate. However, the following technical defects exist.

[0003] In order to improve the query efficiency of a large model, the existing technical solutions often cache commonly used query results. However, with the update of the query target data corresponding to the query results, the timeliness of the query results may be difficult to meet the requirements. Therefore, how to realize the update processing of the cached data according to the update processing result of the query target data to ensure the real-time performance of the update processing of the cached data becomes a technical problem to be solved urgently.

[0004] In view of the above technical problems, specifically, the application provides a large model application method and system for data retrieval. SUMMARY

[0005] To achieve the purpose of the application, the application adopts the following technical solutions:

[0006] Specifically, in a first aspect, the application provides a large model application method for data retrieval, which specifically includes:

[0007] S1, determining an update target based on the update data of the retrieval processing target of a large model, and when the matching deviation of the cached data of the large model meets the requirements in the matching of different update targets in the data retrieval processing process, proceeding to the next step;

[0008] S2, determining a matching deviation target in the update target according to the matching deviation of the text vector of the cached information item in the cached data;

[0009] S3, acquiring the constituent data of the matching deviation target, and combining the matching data of the retrieval information item of different matching deviation targets and retrieval processing times to determine that the update processing of the cached data of the large model does not need to adopt a preset update strategy, and determining the matching update target of different cached information items according to the matching of different update targets and text vectors of cached information items.

[0010] S4 obtains matching data of different cache information items and retrieval information items of retrieval processing times, and determines an update processing strategy of different cache information items in combination with matching update targets of different cache information items and matching conditions of retrieval results of matching update targets and retrieval processing times.

[0011] The present application has the following beneficial effects:

[0012] According to the matching deviation of the update target and the text vector of the cache information item in the cache data, the matching deviation target in the update target is determined, so that the matching condition of the text vector of the cache data and the update target is realized, the determination of the matching deviation target in the update target is realized, and the foundation is laid for further generating a differentiated update processing strategy for the cache data according to the number of matching deviation targets, improving the reliability of the update processing, and ensuring the effectiveness of the cache data.

[0013] According to the matching data of the cache information item and the retrieval information item of the retrieval processing time, the matching update target of different cache information items, and the matching condition of the retrieval result of the matching update target and the retrieval processing time, the update processing strategy of the cache information item is determined, which not only considers the difference in the matching degree of the cache information item and the retrieval demand caused by the difference in the matching condition of the cache information item and the retrieval information item, but also combines the matching update target of the cache information item and the matching condition of the retrieval result, realizes the evaluation of the accuracy deviation of the matching result of the cache information item, and further realizes the determination of the update processing strategy of different cache information items from the perspective of retrieval accuracy and retrieval demand, improves the pertinence and efficiency of the update processing.

[0014] The further technical scheme is that the retrieval processing target is the retrieval target of the retrieval user in the retrieval system.

[0015] The further technical scheme is that the update target is a newly added retrieval processing target in the retrieval processing target.

[0016] The further technical scheme is that the matching condition of the update target in the data retrieval processing process includes the matching number of the update target and the retrieval information item in different retrieval processing times and the retrieval processing time of the retrieval result of the update target.

[0017] The further technical scheme is that the matching deviation condition of the cache data of the large model meets the requirements, specifically including:

[0018] The matching number of the update target and the retrieval information item in different retrieval processing times is determined as the ratio of the matching number to the retrieval information item, and is taken as a retrieval matching coefficient;

[0019] According to the search matching coefficient, determine the information matching search times in the search processing times, take the search processing times of the search result as the update target as the result matching search times, according to the information matching search times of different update targets and the result matching search times, determine the total information matching search times and the total result matching search times;

[0020] With the total information matching search times and the total result matching search times, determine whether the matching deviation situation of the cache data of the large model meets the requirements.

[0021] Further technical solutions are that when any one of the total information matching search times and the total result matching search times is not within the preset search times interval, it is determined that the matching deviation situation of the cache data of the large model does not meet the requirements.

[0022] Further technical solutions are that the information matching search times are search processing times in which the search matching coefficient is greater than the preset search matching times.

[0023] Further technical solutions are that when the matching deviation situation of the cache data of the large model does not meet the requirements, the cache information items in the cache data and the search targets corresponding to the cache information items are both updated.

[0024] Further technical solutions are that the determination method of the update processing strategy of the cache information item is:

[0025] With the matching data of the cache information item and the search information item of the search processing times, determine the matching search times of the cache information item, according to the matching update target search result matching situation in the matching search times, determine the matching search times of the matching update target of the search result, and take it as the matching deviation search times, take the proportion of the matching deviation search times in the matching search times as the deviation search proportion;

[0026] Obtain the number of matching update targets of the cache information item, and determine the target matching deviation coefficient of the cache information item with the ratio of the number of matching update targets to the search target corresponding to the cache information item;

[0027] Determine the search matching deviation coefficient of the cache information item with the mean of the target matching deviation coefficient and the deviation search proportion, and determine the update processing strategy of the cache information item with the search matching deviation coefficient.

[0028] Further technical solutions are that the update processing strategy of the cache information item is determined with the search matching deviation coefficient, specifically including:

[0029] When the retrieval matching deviation coefficient of the cache information item is greater than the preset matching deviation coefficient threshold, it is determined that the retrieval target corresponding to the cache information item needs to be updated;

[0030] When the retrieval matching deviation coefficient of the cache information item is not greater than the preset matching deviation coefficient threshold, it is determined that the retrieval target corresponding to the cache information item does not need to be updated.

[0031] In a second aspect, the present application provides a computer system, comprising a memory and a processor connected in communication, and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to perform the above-mentioned large model application method for data retrieval.

[0032] Other features and advantages will be set forth in the accompanying description, and in part will be apparent from the description and the drawings, or can be learned by practice of the application as claimed in the appended claims.

[0033] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are specifically described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0034] The above-mentioned and other features and advantages of the present application will become more apparent by describing in detail example embodiments thereof with reference to the attached drawings.

[0035] Figure 1 is a flowchart of a large model application method for data retrieval;

[0036] Figure 2 is a flowchart of determining that the matching deviation of the cache data of the large model meets the requirements;

[0037] Figure 3 is a flowchart of a method for determining the matching deviation target in the update target;

[0038] Figure 4 is a flowchart of a method for determining the update processing strategy of the cache information item;

[0039] Figure 5 is a framework diagram of a computer system. DETAILED DESCRIPTION

[0040] In order to enable the person skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in combination with the drawings in the specification. Obviously, the described embodiments are only part of the embodiments of the specification, not all. Based on the embodiments of the specification, all other embodiments obtained by the person skilled in the art without creative labor should be within the protection scope of the specification.

[0041] In the prior art, in order to improve the solution processing efficiency of the user's problem, the large model often caches the commonly used user's problem and the matching result to generate cache data, and when encountering similar problems in the cache data, the result can be generated quickly.

[0042] In the present application, in order to solve the update problem of cache data, according to the update situation of the matching search target of the cache information item in the cache data, the matching situation of the cache information item and the search demand of the search user, a differentiated update processing strategy is generated for different cache information items, which ensures the real-time and accuracy of the cache data.

[0043] Embodiment 1 as shown in the present application provides a large model application method for data search, which specifically includes: Figure 1

[0044] S1 determines the update target based on the update data of the search processing target of the large model, and when the matching deviation situation of the cache data of the large model meets the requirements according to the matching situation of different update targets in the data search processing process, the next step is entered;

[0045] It should be noted that when the proportion of the matching result to the number of search processing times of the update target is greater than 0.3, it is determined that the matching deviation situation of the cache data of the large model does not meet the requirements.

[0046] S2 determines the matching deviation target in the update target according to the matching deviation situation of the text vector of the cache information item in the cache data and the update target.

[0047] Specifically, the update target which is different from the text vector of the cache information item in the cache data is taken as the matching deviation target.

[0048] S3 obtains the constituting data of the matching deviation target, and combines the matching data of the search information item of different matching deviation targets and search processing times to determine that the update processing of the cache data of the large model does not need to use the preset update strategy, and determines the matching update target of different cache information items according to the matching situation of different update targets and text vectors of cache information items.

[0049] ​It can be understood that when the proportion of the number of matching deviations in the retrieval processing times is less than 0.05, it is determined that the preset updating strategy is not needed for updating the cache data of the large model, and the update target with a matching number of text vectors of the cache information item greater than 0.7 is taken as the matching update target of the cache information item.

[0050] S4 obtains matching data of different cache information items and retrieval information items of retrieval processing times, and determines an updating processing strategy of different cache information items in combination with matching update targets of different cache information items and matching conditions of the matching update targets and retrieval results of retrieval processing times.

[0051] Further, the retrieval demand is determined as the retrieval processing times of the cache information item, and the retrieval processing times are taken as the matching retrieval processing times. When the proportion of the matching retrieval processing times in the retrieval processing times is greater than 0.03, or the proportion of the retrieval results that are the matching update targets in the matching retrieval processing times is greater than 0.3, it is determined that the cache information item needs to be updated.

[0052] It should be noted that the retrieval processing target is the retrieval target of the retrieval user in the retrieval system.

[0053] Further, the update target is a newly added retrieval processing target in the retrieval processing target.

[0054] Specifically, the matching condition of the update target in the data retrieval processing process includes the number of matches of the update target with retrieval information items in different retrieval processing times and the retrieval processing times with retrieval results being the update target.

[0055] It can be understood that, as shown in Figure 2 The matching deviation condition of the cache data of the large model is determined to meet the requirements, specifically including:

[0056] The matching number of the update target with retrieval information items in different retrieval processing times is determined as the ratio of the matching number to the retrieval information item, and is taken as the retrieval matching coefficient;

[0057] According to the retrieval matching coefficient, the information matching retrieval times in the retrieval processing times are determined, the retrieval processing times with retrieval results being the update target are taken as the result matching retrieval times, and according to the information matching retrieval times and the result matching retrieval times of different update targets, the total number of information matching retrieval times and the total number of result matching retrieval times are determined;

[0058] The matching deviation condition of the cache data of the large model is determined to meet the requirements, specifically including:

[0059] Further, when either the information matching search number or the result matching search number is not within the preset search number interval, it is determined that the matching deviation condition of the cache data of the large model does not meet the requirement.

[0060] It should be noted that the information matching search number is the search processing number in which the search matching coefficient is greater than the preset search matching number.

[0061] It can be understood that when the matching deviation condition of the cache data of the large model does not meet the requirement, the cache information items in the cache data and the search targets corresponding to the cache information items are updated.

[0062] In another possible embodiment, determining that the matching deviation condition of the cache data of the large model meets the requirement includes:

[0063] Taking the search processing number in which the search result is the update target as the result matching search number;

[0064] Obtaining a proportion of the result matching search number in the search processing number as a result matching proportion;

[0065] Determining, according to the result matching proportion, whether the matching deviation condition of the cache data of the large model meets the requirement.

[0066] Further, when the result matching proportion is greater than a preset result matching proportion threshold, it is determined that the matching deviation condition of the cache data of the large model does not meet the requirement.

[0067] Optionally, determining that the matching deviation condition of the cache data of the large model meets the requirement includes:

[0068] Taking the search processing number in which the search result is the update target as the result matching search number, obtaining a proportion of the result matching search number in the search processing number as a result matching proportion;

[0069] When either the result matching search number or the result matching proportion does not meet the requirement, it is determined that the matching deviation condition of the cache data of the large model does not meet the requirement;

[0070] When both the result matching search number and the result matching proportion meet the requirement:

[0071] Determine a ratio of the number of matches to the search information item based on the number of matches of the search target in different search processing times, and take the ratio as a search match coefficient, determine the information match search times in the search processing times based on the search match coefficient, and when either the sum of the information match search times of different update targets or the proportion of the sum of the information match search times in the search processing times does not meet the requirements, it is determined that the match deviation of the cache data of the large model does not meet the requirements;

[0072] When the sum of the information match search times of different update targets and the proportion of the sum of the information match search times in the search processing times both meet the requirements:

[0073] Determine the target search match coefficient of different update targets based on the information match search times and the result match search times of different update targets, and when the target search match coefficient of different update targets is less than the preset search match coefficient threshold, it is determined that the match deviation of the cache data of the large model meets the requirements;

[0074] When there is an update target whose target search match coefficient is not less than the preset search match coefficient threshold:

[0075] Take the update target whose target search match coefficient is not less than the preset search match coefficient threshold as the search match update target, and when the number of search match update targets does not meet the requirements, it is determined that the match deviation of the cache data of the large model does not meet the requirements;

[0076] When the number of search match update targets meets the requirements:

[0077] Determine the update target match coefficient based on the target search match coefficient of different update targets, and use the update target match coefficient to determine whether the match deviation of the cache data of the large model meets the requirements.

[0078] Specifically, when the update target match coefficient is greater than the preset target match coefficient setting value, it is determined that the match deviation of the cache data of the large model does not meet the requirements.

[0079] It should be noted that, as Figure 3 The method for determining the match deviation target in the update target is:

[0080] Determine the number of matches with the text vector in different cache information items based on the match deviation of the text vector of the update target and the text vector of the cache information item in the cache data;

[0081] determine a matching coefficient of the update target and the different cache information items according to the number of matches of the text vectors and the ratio of the text vectors in the different cache information items;

[0082] determine whether the update target is a matching deviation target according to the matching coefficients of the different cache information items.

[0083] Further, when the maximum value of the matching coefficients of the update target and the different cache information items is less than a preset cache information matching coefficient, it is determined that the update target is a matching deviation target.

[0084] It can be understood that the determination that the preset update strategy is not required for the update processing of the cache data of the large model specifically includes:

[0085] determine the matching data of the different matching deviation targets and the retrieval information items of the retrieval processing times, determine the matching number of the matching deviation targets and the retrieval information items of the retrieval processing times, determine the ratio of the matching number to the retrieval information items as a retrieval matching coefficient;

[0086] determine the information matching retrieval times in the retrieval processing times according to the retrieval matching coefficient, and determine the matching deviation targets whose information matching retrieval times are greater than a preset matching retrieval time as retrieval matching deviation targets based on the information matching retrieval times of the different matching deviation targets;

[0087] determine whether the preset update strategy is required for the update processing of the cache data of the large model based on the number of the retrieval matching deviation targets.

[0088] Further, when the number of the retrieval matching deviation targets is greater than a preset deviation target number threshold, it is determined that the preset update strategy is required for the update processing of the cache data of the large model.

[0089] It should be noted that when the preset update strategy is required for the update processing of the cache data of the large model, the cache information items in the cache data and the retrieval targets corresponding to the cache information items are all updated.

[0090] In another possible embodiment, the determination that the preset update strategy is not required for the update processing of the cache data of the large model specifically includes:

[0091] determine the matching data of the different matching deviation targets and the retrieval information items of the retrieval processing times, determine the matching number of the matching deviation targets and the retrieval information items of the retrieval processing times, determine the ratio of the matching number to the retrieval information items as a retrieval matching coefficient;

[0092] According to different matching deviation targets, a search matching coefficient is determined based on the number of search processing times, and information matching search times in the number of search processing times are determined;

[0093] Based on the information matching search times, it is determined whether the preset update strategy needs to be used for the update processing of the cached data of the large model.

[0094] Further, when the information matching search times are greater than a preset matching search times threshold, it is determined that the preset update strategy needs to be used for the update processing of the cached data of the large model.

[0095] Optionally, it is determined that the preset update strategy does not need to be used for the update processing of the cached data of the large model, specifically including:

[0096] The number of matching deviation targets is obtained, and when the number of matching deviation targets is greater than a preset deviation target number, it is determined that the preset update strategy needs to be used for the update processing of the cached data of the large model.

[0097] When the number of matching deviation targets is not greater than the preset deviation target number:

[0098] When the number of matching deviation targets is within a preset matching deviation target number interval, it is determined that the preset update strategy does not need to be used for the update processing of the cached data of the large model.

[0099] When the number of matching deviation targets is not within the preset matching deviation target number interval:

[0100] The matching data of different matching deviation targets and search information items of search processing times is used to determine the matching number of the matching deviation targets and the search information items of search processing times, to determine the ratio of the matching number to the search information items as a search matching coefficient.

[0101] According to the search matching coefficient, information matching search times in the number of search processing times are determined, and when the information matching search times of the matching deviation targets do not meet the requirements, it is determined that the preset update strategy needs to be used for the update processing of the cached data of the large model.

[0102] When the information matching search times of the matching deviation targets meet the requirements:

[0103] Based on the information matching search times of different matching deviation targets, the matching deviation targets with information matching search times greater than a preset matching search times are determined as search matching deviation targets, and when the number of search matching deviation targets does not meet the requirements, it is determined that the preset update strategy needs to be used for the update processing of the cached data of the large model.

[0104] When the number of the search matching deviation targets meets the requirement:

[0105] Based on the information matching search times of different matching deviation targets and the search matching coefficients of different information matching search times, the update demand coefficient of the matching deviation target is determined, and whether the preset update strategy needs to be used for the update processing of the cache data of the large model is determined by using the update demand coefficient.

[0106] Further, when the update demand coefficient is greater than the preset update demand coefficient threshold, it is determined that the preset update strategy needs to be used for the update processing of the cache data of the large model.

[0107] Specifically, the matching update target of the cache information item is an update target with a cache information matching coefficient of the cache information item greater than an information matching coefficient preset value.

[0108] It should be noted that the search result matching condition of the matching update target and the search processing times includes the search processing times of the matching update target.

[0109] It can be understood that, as shown in the method for determining the update processing strategy of the cache information item: Figure 4

[0110] The matching search times of the cache information item are determined based on the matching data of the cache information item and the search information item of the search processing times, the matching search times of the matching update target with the search result matching condition are determined according to the matching search times, and the matching search times of the matching update target with the search result matching condition are taken as the matching deviation search times, and the proportion of the matching deviation search times in the matching search times is taken as the deviation search proportion;

[0111] The number of the matching update targets of the cache information item is obtained, and the target matching deviation coefficient of the cache information item is determined based on the ratio of the number of the matching update targets to the search target corresponding to the cache information item;

[0112] The search matching deviation coefficient of the cache information item is determined by using the mean value of the target matching deviation coefficient and the deviation search proportion, and the update processing strategy of the cache information item is determined by using the search matching deviation coefficient.

[0113] Further, the update processing strategy of the cache information item is determined by using the search matching deviation coefficient, specifically including:

[0114] When the search matching deviation coefficient of the cache information item is greater than a preset matching deviation coefficient threshold, it is determined that the search target corresponding to the cache information item needs to be updated; ​

[0115] When the retrieval matching deviation coefficient of the cache information item is not greater than a preset matching deviation coefficient threshold, it is determined that the retrieval target corresponding to the cache information item does not need to be updated.

[0116] Optionally, the further technical scheme is characterized in that the method for determining the update processing strategy of the cache information item comprises:

[0117] S41 acquiring the number of matching update targets of the cache information item, and determining a target matching deviation coefficient of the cache information item according to the ratio of the number of matching update targets to the retrieval target corresponding to the cache information item;

[0118] S42 determining the matching retrieval times of the cache information item according to the matching data of the cache information item and the retrieval information item, determining the matching retrieval times of the retrieval result matching update target according to the matching update target retrieval result matching condition in the matching retrieval times, taking the matching retrieval times of the retrieval result matching update target as the matching deviation retrieval times, and determining the retrieval update demand coefficient of the cache information item according to the matching retrieval times and the matching deviation retrieval times;

[0119] S43 determining the retrieval matching deviation coefficient of the cache information item according to the target matching deviation coefficient and the retrieval update demand coefficient, and determining the update processing strategy of the cache information item according to the retrieval matching deviation coefficient.

[0120] Optionally, the retrieval update demand coefficient of the cache information item is determined according to the matching retrieval times and the matching deviation retrieval times, and specifically comprises:

[0121] determining a basic update demand coefficient according to a preset update demand coefficient corresponding to the matching retrieval times;

[0122] determining the retrieval update demand coefficient of the cache information item according to the product of the basic update demand coefficient and the proportion of the matching deviation retrieval times in the matching retrieval times.

[0123] Optionally, the step S41 comprises the following content:

[0124] S411 acquiring the number of matching update targets of the cache information item, and determining that the retrieval target corresponding to the cache information item needs to be updated when the number of matching update targets is greater than a preset target number threshold, and turning to step S412 when the number of matching update targets is not greater than the preset target number threshold;

[0125] S412 determines the target matching deviation coefficient of the cached information item by the ratio of the number of matching update targets to the retrieval target corresponding to the cached information item. When the target matching deviation coefficient of the cached information item does not meet the requirements, it is determined that the retrieval target corresponding to the cached information item needs to be updated. When the target matching deviation coefficient of the cached information item meets the requirements, the process proceeds to step S42.

[0126] Optionally, step S42 above includes the following:

[0127] S421 uses the matching data of the cached information item and the retrieval information item of the retrieval processing number to determine the matching retrieval number of the cached information item. When the matching retrieval number of the cached information item is greater than the preset matching retrieval number threshold, proceed to step S422. When the matching retrieval number of the cached information item is not greater than the preset matching retrieval number threshold, proceed to step S423.

[0128] S422 When the target matching deviation coefficient of the cached information item is within the preset target matching deviation coefficient range, it is determined that the retrieval target corresponding to the cached information item needs to be updated. When the target matching deviation coefficient of the cached information item is not within the preset target matching deviation coefficient range, proceed to step S423.

[0129] S423 determines the number of times the search result matches the target update and uses it as the number of times the search result matches the target update. If the number of times the search result matches the target update, it is determined that the target update process needs to be performed on the cached information item. If the number of times the search result matches the target update, the process proceeds to step S424.

[0130] S424 determines the retrieval update requirement coefficient of the cached information item based on the number of matching retrievals and the number of matching deviation retrievals. When the retrieval update requirement coefficient is greater than the preset retrieval update requirement coefficient threshold, it is determined that the retrieval target corresponding to the cached information item needs to be updated. When the retrieval update requirement coefficient is not greater than the preset retrieval update requirement coefficient threshold, the process proceeds to step S43.

[0131] Example 2, Second aspect, such as Figure 5 As shown, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-described method for large model application for data retrieval when running the computer program.

[0132] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0133] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0134] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A large model application method for data retrieval, characterized in that, Specifically comprising: Determine the update target based on the updated data of the retrieval processing target of the large model, and determine the matching deviation of the cached data of the large model when the matching deviation meets the requirements, and enter the next step; Determine the matching deviation target in the update target according to the matching deviation of the text vector of the cached information item in the cached data; When it is determined that the update processing of the cached data of the large model does not need to use the preset update strategy, the matching update target of different cached information items is determined according to the matching situation of different update targets and text vectors of cached information items; Obtain the matching data of different cached information items and retrieval information items of retrieval processing times, and determine the update processing strategy of different cached information items in combination with the matching update target of different cached information items, the matching update target and the retrieval result matching situation of the retrieval processing times; The method for determining the update processing strategy of the cached information item is: Determine the matching retrieval times of the cached information item based on the matching data of the cached information item and the retrieval information item of the retrieval processing times, and determine the matching retrieval times of the retrieval result as the matching update target according to the retrieval result matching situation of the matching update target in the matching retrieval times, and take it as the matching deviation retrieval times, and take the proportion of the matching deviation retrieval times in the matching retrieval times as the deviation retrieval proportion; Obtain the number of matching update targets of the cached information item, and determine the target matching deviation coefficient of the cached information item based on the ratio of the number of matching update targets to the retrieval target corresponding to the cached information item; Determine the retrieval matching deviation coefficient of the cached information item by using the mean value of the target matching deviation coefficient and the deviation retrieval proportion, and determine the update processing strategy of the cached information item by using the retrieval matching deviation coefficient.

2. The large model application method for data retrieval of claim 1, wherein, The retrieval processing target is the retrieval target of the retrieval user in the retrieval system.

3. The large model application method for data retrieval of claim 1, wherein, The update target is a newly added retrieval processing target in the retrieval processing target.

4. The large model application method for data retrieval of claim 1, wherein, Determine the matching deviation of the cached data of the large model to meet the requirements, specifically comprising: Determine the ratio of the matching number to the retrieval information item based on the matching number of the update target in different retrieval processing times, and take it as the retrieval matching coefficient; Determine the information matching retrieval times in the retrieval processing times according to the retrieval matching coefficient, take the retrieval processing times of the retrieval result as the result matching retrieval times, and determine the information matching retrieval total times and the result matching retrieval total times according to the information matching retrieval times of different update targets and the result matching retrieval times; Determine whether the matching deviation of the cached data of the large model meets the requirements based on the information matching retrieval total times and the result matching retrieval total times.

5. The large model application method for data retrieval of claim 4, wherein, When either of the information matching search total number and the result matching search total number is not within the preset search number interval, it is determined that the matching deviation condition of the cache data of the large model does not meet the requirement.

6. The method of claim 1, wherein, When the matching deviation condition of the cache data of the large model does not meet the requirement, the cache information item in the cache data and the search target corresponding to the cache information item are updated.

7. The large model application method for data retrieval of claim 1, wherein, The matching update target and the search result matching condition of the search processing number include the search processing number of the matching update target.

8. The method of claim 1, wherein, The update processing strategy of the cache information item is determined by using the search matching deviation coefficient, and specifically includes: When the search matching deviation coefficient of the cache information item is greater than a preset matching deviation coefficient threshold, it is determined that the search target corresponding to the cache information item needs to be updated. When the search matching deviation coefficient of the cache information item is not greater than the preset matching deviation coefficient threshold, it is determined that the search target corresponding to the cache information item does not need to be updated.

9. A computer system comprising: The memory and processor connected in communication, and the computer program stored on the memory and capable of running on the processor, characterized in that the processor executes the computer program to perform the large model application method for data search according to any one of claims 1-8.

Images