Information retrieval method and device, equipment and storage medium

Abstract

The invention discloses an information retrieval method and device, equipment and a storage medium, and the method comprises the steps: obtaining initial recall information returned by a preset knowledge base, carrying out the abstract processing of the initial recall information through employing a preset large language model, and obtaining abstract information, the input information of the preset large language model comprises a preset structured input sequence, the preset structured input sequence is used for guiding the preset large language model to generate an output result expected by a user, performing secondary information retrieval based on the summary information to obtain secondary recall information returned by the preset knowledge base, and generating a target information retrieval result based on the primary recall information and the secondary recall information. According to the scheme, secondary information retrieval is performed based on the summary information, so that the accuracy of information retrieval is improved; besides, the first recall information is subjected to abstract processing by adopting the preset large language model, so that the obtained abstract information is more accurate, and the accuracy of information retrieval is further improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to information retrieval methods, devices, equipment, and storage media. Background Art

[0002] Current information recall technology is mainly applied to scenarios that require retrieving relevant information from a large number of documents. This technology matches the user's question with the content in the document library to provide the most relevant answers or information. For example, in a knowledge base question-and-answer system, a large language model (LLM) can assist the retrieval system by generating queries related to the user's question, improving the accuracy and efficiency of the retrieval. Currently, there are many scenarios where LLM is used to assist in information recall. By using LLM, the dependence on traditional keyword matching can be reduced, and instead, more complex semantic understanding and generation capabilities can be utilized to provide a higher-quality user interaction experience.

[0003] Specifically, in a knowledge base question-and-answer system, first, vector matching is performed in the local knowledge base according to the "retrieval target", and then the local knowledge base returns the results with a higher matching degree, which are used as the "recalled information". These information will directly be used as the background information for the LLM to generate the answer content when interacting with the LLM. Finally, the "retrieval target" and the "recalled information" are simultaneously passed to the LLM, and the LLM summarizes and creates the question-and-answer result for return.

[0004] However, the above information retrieval method has the problem of low accuracy. Summary of the Invention

[0005] This application aims to at least solve the technical problems existing in the prior art. To this end, a first aspect of this application proposes an information retrieval method, which includes:

[0006] Obtain the initially recalled information returned by the preset knowledge base;

[0007] Use the preset large language model to perform summary processing on the initially recalled information to obtain summary information; wherein, the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate the output result expected by the user;

[0008] Perform secondary information retrieval based on the summary information to obtain the secondarily recalled information returned by the preset knowledge base;

[0009] Generate a target information retrieval result based on the initially recalled information and the secondarily recalled information.

[0010] In a possible implementation manner, generating a target information retrieval result based on the initially recalled information and the secondarily recalled information includes:

[0011] Obtain the first recall similarity corresponding to the initial recall information and the second recall similarity corresponding to the secondary recall information respectively;

[0012] Calculate the initial recall weight based on the first recall similarity and a preset similarity threshold, and calculate the secondary recall weight based on the second recall similarity and the preset similarity threshold;

[0013] Calculate the target contribution corresponding to the secondary recall information based on the initial recall weight, the secondary recall weight, and a preset recall result contribution coefficient;

[0014] Obtain the target information retrieval result based on the target contribution.

[0015] In a possible implementation manner, obtaining the target information retrieval result based on the target contribution includes:

[0016] Obtain at least one optimized contribution; wherein, the optimized contribution is obtained by optimizing at least one initial parameter in the information retrieval process according to at least one preset correction strategy in a preset correction strategy set;

[0017] Obtain the target information retrieval result based on the optimized contribution and the target contribution.

[0018] In a possible implementation manner, obtaining the target information retrieval result based on the optimized contribution and the target contribution includes:

[0019] Obtain the difference between the maximum value and the minimum value of the optimized contribution and the target contribution;

[0020] Determine the target information retrieval result based on the difference and a preset difference threshold.

[0021] In a possible implementation manner, determining the target information retrieval result based on the difference and the preset difference threshold includes:

[0022] If the difference is less than the preset difference threshold, use the secondary recall information corresponding to the target contribution as the target information retrieval result;

[0023] If the difference is greater than or equal to the preset difference threshold, optimize the corresponding parameters in the information retrieval based on the preset correction strategy corresponding to the optimized contribution to obtain new secondary recall information;

[0024] Use the new secondary recall information as the target information retrieval result.

[0025] In a possible implementation manner, obtaining at least one optimized contribution includes:

[0026] Obtain at least one preset correction strategy in the preset correction strategy set;

[0027] Optimize at least one initial parameter corresponding to the information retrieval process by using at least one preset correction strategy to obtain optimized parameters; wherein, the initial parameters include a preset structured input sequence, a preset recall result contribution coefficient, and a preset maximum number of secondary recall loops.

[0028] Perform secondary information retrieval based on the optimized parameters and the summary information to obtain the secondary recall information returned by the preset knowledge base, and calculate the optimized contribution degree corresponding to the secondary recall information.

[0029] In a possible implementation manner, the method further includes:

[0030] If the second recall similarity is greater than the preset similarity threshold, repeat the step of performing secondary information retrieval based on the summary information according to the preset maximum number of secondary recall loops to obtain the secondary recall information returned by the preset knowledge base, and generate the target information retrieval result based on the primary recall information and the secondary recall information.

[0031] If the second recall similarity is less than or equal to the preset similarity threshold, continue to perform the step of generating the target information retrieval result based on the primary recall information and the secondary recall information.

[0032] A second aspect of this application proposes an information retrieval device, which includes:

[0033] An acquisition module, configured to acquire the primary recall information returned by the preset knowledge base;

[0034] A processing module, configured to perform summary processing on the primary recall information by using a preset large language model to obtain summary information; wherein, the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate the output result expected by the user.

[0035] A retrieval module, configured to perform secondary information retrieval based on the summary information to obtain the secondary recall information returned by the preset knowledge base;

[0036] A generation module, configured to generate a target information retrieval result based on the primary recall information and the secondary recall information.

[0037] In a possible implementation manner, the above generation module is specifically configured to:

[0038] Respectively obtain the first recall similarity corresponding to the primary recall information and the second recall similarity corresponding to the secondary recall information;

[0039] Calculate the primary recall weight based on the first recall similarity and the preset similarity threshold, and calculate the secondary recall weight based on the second recall similarity and the preset similarity threshold;

[0040] Calculate the target contribution corresponding to the secondary recall information based on the primary recall weight, secondary recall weight, and preset recall result contribution coefficient;

[0041] Obtain the target information retrieval result based on the target contribution.

[0042] In one possible implementation, the above-mentioned generation module is further configured to:

[0043] Obtain at least one optimized contribution; wherein, the optimized contribution is obtained by optimizing at least one initial parameter in the information retrieval process according to at least one preset correction strategy in the preset correction strategy set;

[0044] Obtain the target information retrieval result based on the optimized contribution and the target contribution.

[0045] In one possible implementation, the above-mentioned generation module is further configured to:

[0046] Obtain the difference between the maximum value and the minimum value in the optimized contribution and the target contribution;

[0047] Determine the target information retrieval result based on the difference and the preset difference threshold.

[0048] In one possible implementation, the above-mentioned generation module is further configured to:

[0049] If the difference is less than the preset difference threshold, use the secondary recall information corresponding to the target contribution as the target information retrieval result;

[0050] If the difference is greater than or equal to the preset difference threshold, optimize the corresponding parameters in the information retrieval based on the preset correction strategy corresponding to the optimized contribution to obtain new secondary recall information;

[0051] Use the new secondary recall information as the target information retrieval result.

[0052] In one possible implementation, the above-mentioned generation module is further configured to:

[0053] Obtain at least one preset correction strategy in the preset correction strategy set;

[0054] Optimize at least one initial parameter in the information retrieval process using at least one preset correction strategy to obtain optimized parameters; wherein, the initial parameters include a preset structured input sequence, a preset recall result contribution coefficient, and a preset maximum number of secondary recall loops;

[0055] Perform secondary information retrieval based on the optimized parameters and the abstract information to obtain the secondary recall information returned by the preset knowledge base, and calculate the optimized contribution corresponding to the secondary recall information.

[0056] In a possible implementation manner, the above information retrieval device is further configured to:

[0057] If the second recall similarity is greater than the preset similarity threshold, then based on the preset maximum number of secondary recall loops, repeatedly execute the step of performing secondary information retrieval based on the summary information to obtain the secondary recall information returned by the preset knowledge base, and generate the target information retrieval result based on the primary recall information and the secondary recall information;

[0058] If the second recall similarity is less than or equal to the preset similarity threshold, then continue to execute the step of generating the target information retrieval result based on the primary recall information and the secondary recall information.

[0059] A third aspect of the present application provides an electronic device, which includes a processor and a memory. At least one instruction, at least one program, a code set, or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the information retrieval method as described in the first aspect.

[0060] A fourth aspect of the present application provides a computer-readable storage medium, in which at least one instruction, at least one program, a code set, or an instruction set is stored, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the information retrieval method as described in the first aspect.

[0061] The embodiments of the present application have the following beneficial effects:

[0062] The information retrieval method provided by the embodiments of the present application includes: obtaining the primary recall information returned by the preset knowledge base, performing summary processing on the primary recall information by using a preset large language model to obtain summary information, where the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate the output result expected by the user, performing secondary information retrieval based on the summary information to obtain the secondary recall information returned by the preset knowledge base, and generating the target information retrieval result based on the primary recall information and the secondary recall information. This solution improves the accuracy of information retrieval by performing secondary information retrieval based on the summary information; in addition, by using the preset large language model to perform summary processing on the primary recall information, the obtained summary information is more accurate, thereby further improving the accuracy of information retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a block diagram of a computer device provided by an embodiment of the present application;

[0064] Figure 2 It is a step flowchart of an information retrieval method provided by an embodiment of the present application;

[0065] Figure 3 A flowchart of steps for generating a target information retrieval result provided by an embodiment of the present application;

[0066] Figure 4 A flowchart of steps for obtaining a target information retrieval result provided by an embodiment of the present application;

[0067] Figure 5 A flowchart of steps for obtaining an optimization contribution degree provided by an embodiment of the present application;

[0068] Figure 6 A flowchart of steps for generating a target information retrieval result provided by an embodiment of the present application;

[0069] Figure 7 A flowchart of steps for determining a target information retrieval result provided by an embodiment of the present application;

[0070] Figure 8 A structural block diagram of an information retrieval device provided by an embodiment of the present application. Detailed implementation manners

[0071] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without making creative efforts belong to the scope of protection of the present application.

[0072] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present disclosure, unless otherwise stated, the meaning of "a plurality" is two or more. Additionally, the use of "based on" or "according to" is meant to be open and inclusive, because a process, step, calculation, or other action "based on" or "according to" one or more of the stated conditions or values may in practice be based on additional conditions or values beyond those stated.

[0073] The information retrieval method provided by the present application can be applied to a computer device (electronic device). The computer device can be a server or a terminal. Among them, the server can be a single server or a server cluster composed of multiple servers. The embodiments of the present application do not make specific limitations in this regard. The terminal can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices.

[0074] Taking the computer device as a server as an example, Figure 1 A block diagram of a server is shown, as Figure 1 shown, the server may include a processor and a memory connected by a system bus. Among them, the processor of the server is used to provide computing and control capabilities. The memory of the server includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the computer program is executed by the processor, it implements an information retrieval method.

[0075] Those skilled in the art can understand that Figure 1 the structure shown in is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the server to which the solution of the present application is applied. Optionally, the server may include more or fewer components than those shown in the figure, or combine certain components, or have a different component layout.

[0076] It should be noted that the execution subject of the embodiments of the present application may be a computer device or an information retrieval device. In the following method embodiments, the computer device is used as the execution subject for description.

[0077] Figure 2 is a step flowchart of an information retrieval method provided by an embodiment of the present application. As Figure 2 shown, the method includes the following steps:

[0078] Step 202, obtain the initial recall information returned by the preset knowledge base.

[0079] Among them, in the process of information retrieval, the process of obtaining the results directly retrieved from the preset knowledge base is called initial recall. Optionally, after obtaining the initial recall information, it can be first determined whether the initial recall information meets the requirements of the preset similarity threshold. If it meets the requirements of the preset similarity threshold, the initial recall information is the final retrieval result; otherwise, retrieval enhancement will be performed. Here, the initial recall information can be denoted as s1_result, and the preset similarity threshold can be denoted as c_limit.

[0080] The preset similarity threshold can be set accordingly for different business scenarios. Optionally, it can be given by a business expert for the preset similarity threshold in a specific scenario. At this time, the set preset similarity threshold is the initialization threshold.

[0081] The initial threshold is often given by experts under ideal circumstances. In actual recall, the initial threshold can be optimized according to factors such as the scenario, hardware and software environment, etc. Exemplarily, if m questions to be retrieved are selected and retrieved one by one using the system, and the expert gives the expected retrieval results, the semantic similarity between each actual retrieval result and the expected retrieval result given by the expert is calculated, and finally the similarity value is obtained. The average value of the similarity values corresponding to these m questions is used as the optimized preset similarity threshold.

[0082] Optionally, the first recall similarity of the initial recall information can be denoted as current_rate1, and its similarity will be calculated automatically in the recall mechanism. Then, the first recall similarity current_rate1 is compared with the above-mentioned preset similarity threshold c_limit. If current_rate1 is greater than the preset similarity threshold c_limit, it is directly returned and this recall ends. If current_rate1 is less than or equal to the preset similarity threshold c_limit, retrieval enhancement is performed.

[0083] It should be noted that the initial recall information can include multiple items, that is, the initial recall information can be a list, and the initial recall information in the list can be sorted from high to low according to the first recall similarity. When making a comparison, the first recall similarity corresponding to the first initial recall information can be directly used for comparison with the preset similarity threshold.

[0084] Step 204: Use a preset large language model to perform abstract processing on the initial recall information to obtain abstract information.

[0085] Among them, when performing retrieval enhancement, the preset large language model can be used to perform abstract processing on the initial recall information to obtain abstract information. The input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate the output result expected by the user.

[0086] Optionally, for different application tasks or business scenarios, a preset structured input sequence, that is, a Prompt, can be constructed. Its function is to tell the preset large language model the precautions, key points for abstracting, data formats, taboos, etc. during abstract processing, so as to guide the preset large language model to generate the output result expected by the user. The specific implementation method can refer to the existing technology and will not be elaborated here.

[0087] Step 206: Perform secondary information retrieval based on the abstract information to obtain the secondary recall information returned by the preset knowledge base.

[0088] Among them, the content of the abstract information is usually relatively small. After obtaining the abstract information, secondary information retrieval can be performed based on the abstract information, so as to obtain the secondary recall information returned by the preset knowledge base. The secondary recall information can be denoted as s2_result. The process of secondary information retrieval is the same as that of the primary information retrieval.

[0089] Step 208: Generate a target information retrieval result based on the primary recall information and the secondary recall information.

[0090] Among them, in some optional embodiments, after obtaining the primary recall information and the secondary recall information, the primary recall information and the secondary recall information can be directly combined to generate a target information retrieval result.

[0091] In other optional embodiments, as Figure 3 shown, Figure 3 is a step flowchart for generating a target information retrieval result provided by an embodiment of the present application, including:

[0092] Step 302: Obtain the first recall similarity corresponding to the primary recall information and the second recall similarity corresponding to the secondary recall information respectively.

[0093] Step 304: Calculate the primary recall weight based on the first recall similarity and the preset similarity threshold, and calculate the secondary recall weight based on the second recall similarity and the preset similarity threshold.

[0094] Step 306: Calculate the target contribution corresponding to the secondary recall information based on the primary recall weight, the secondary recall weight, and the preset recall result contribution coefficient.

[0095] Step 308: Obtain the target information retrieval result based on the target contribution.

[0096] Among them, the first recall similarity corresponding to the primary recall information is current_rate1, and the second recall similarity corresponding to the secondary recall information can be denoted as current_rate2, and their similarities will be calculated automatically in the recall mechanism.

[0097] After obtaining the second recall similarity corresponding to the secondary recall information, the effectiveness of the secondary recall information can be judged first based on the second recall similarity. The effectiveness discrimination is to ensure that the information does not deviate from the original problem and intention after the secondary retrieval. Optionally, the second recall similarity can be compared with the preset similarity threshold c_limit. If the second recall similarity is greater than the preset similarity threshold, the secondary recall information is effective. If the second recall similarity is less than or equal to the preset similarity threshold, the secondary recall information is invalid.

[0098] In some optional embodiments, if the second recall similarity is greater than the preset similarity threshold, then based on the preset maximum number of secondary recall loops, the step of repeatedly performing secondary information retrieval based on the summary information is executed to obtain the secondary recall information returned by the preset knowledge base, and based on the primary recall information and the secondary recall information, the step of generating the target information retrieval result is performed. If the second recall similarity is less than or equal to the preset similarity threshold, then the step of generating the target information retrieval result based on the primary recall information and the secondary recall information is continued.

[0099] Next, the primary recall weight can be calculated based on the first recall similarity and the preset similarity threshold, and the secondary recall weight can be calculated based on the second recall similarity and the preset similarity threshold. Among them, the primary recall weight can be denoted as s1_result_w, The secondary recall weight can be denoted as s2_result_w,

[0100] Based on the primary recall weight, the secondary recall weight, and the preset recall result contribution coefficient, the target contribution of the secondary recall information is calculated. Among them, the preset recall result contribution coefficient can be denoted as contri_index, which can represent the target expectation of the secondary recall relative to the primary recall, and this value is generally given in advance by the application engineer according to the comprehensive retrieval effect. Thus, the target contribution can be denoted as contri_rate, and contri_rate = s1_result_w + contri_index * s2_result_w.

[0101] Finally, based on the target contribution, the target information retrieval result can be obtained. In some optional embodiments, as Figure 4 shown, Figure 4 is a flowchart of the steps for obtaining the target information retrieval result provided by the embodiment of the present application, including:

[0102] Step 402, obtain at least one optimized contribution.

[0103] Among them, the optimized contribution is obtained by optimizing at least one initial parameter in the information retrieval process according to at least one preset correction strategy in the preset correction strategy set. The initial parameters may include a preset structured input sequence, a preset recall result contribution coefficient, and a preset maximum number of secondary recall loops. In some optional embodiments, as Figure 5 shown, Figure 5 is a flowchart of the steps for obtaining the optimized contribution provided by the embodiment of the present application, including:

[0104] Step 502, obtain at least one preset correction strategy in the preset correction strategy set.

[0105] Step 504: Optimize at least one initial parameter corresponding to the information retrieval process using at least one preset correction strategy to obtain optimized parameters.

[0106] Step 506: Perform secondary information retrieval based on the optimized parameters and the summary information to obtain secondary recall information returned by the preset knowledge base, and calculate the optimized contribution degree corresponding to the secondary recall information.

[0107] Among them, the preset correction strategy set can include multiple preset correction strategies, mainly for correcting initial parameters such as the preset structured input sequence, the preset recall result contribution degree coefficient, and the preset maximum number of secondary recall loops.

[0108] Optionally, at least one preset correction strategy in the preset correction strategy set can be randomly obtained, so that at least one preset correction strategy can be used to optimize at least one initial parameter corresponding to the information retrieval process to obtain optimized parameters. Exemplarily, three preset correction strategies can be randomly obtained from the preset correction strategy set. Then, for the parameter values corresponding to these three preset correction strategies, at least one initial parameter corresponding to the information retrieval process is optimized to obtain the corresponding optimized parameters.

[0109] Perform secondary information retrieval based on the optimized parameters and the summary information to obtain secondary recall information returned by the preset knowledge base, and calculate the optimized contribution degree corresponding to the secondary recall information. The specific calculation process can refer to the process of the above embodiment and will not be elaborated here. By appropriately correcting the initial parameters, the recall effect is comprehensively improved, and the accuracy of information retrieval is also improved.

[0110] Based on each of the above optimized parameters, a corresponding optimized contribution degree can be obtained, which can be respectively denoted as contri_rate_1, contri_rate_2, and contri_rate_3.

[0111] Step 404: Obtain the target information retrieval result based on the optimized contribution degree and the target contribution degree.

[0112] Among them, in some optional embodiments, as Figure 6 shown, Figure 6 is a flowchart of the steps for generating the target information retrieval result provided by the embodiment of the present application, including:

[0113] Step 602: Obtain the difference between the maximum value and the minimum value in the optimized contribution degree and the target contribution degree.

[0114] Step 604: Determine the target information retrieval result based on the difference and the preset difference threshold.

[0115] Among them, continuing with the above three optimization contribution degrees as an example, the maximum value among the optimization contribution degree and the target contribution degree can be denoted as max(contri_rate, contri_rate_1, contri_rate_2, contri_rate_3), and the minimum value among the optimization contribution degree and the target contribution degree can be denoted as min(contri_rate, contri_rate_1, contri_rate_2, contri_rate_3). Thus, the difference between the maximum and minimum values of the optimization contribution degree and the target contribution degree can be calculated, and this difference can be denoted as cc_limit_current.

[0116] Next, based on the difference and a preset difference threshold, the target information retrieval result can be determined. The preset difference threshold can be custom-set in advance and is denoted as cc_limit. In some optional embodiments, as Figure 7 shown, Figure 7 is a flowchart of the steps for determining the target information retrieval result provided by an embodiment of the present application, including:

[0117] Step 702: If the difference is less than the preset difference threshold, then use the secondary recall information corresponding to the target contribution degree as the target information retrieval result.

[0118] Step 704: If the difference is greater than or equal to the preset difference threshold, then optimize the corresponding parameters in the information retrieval based on the preset correction strategy corresponding to the optimization contribution degree to obtain new secondary recall information.

[0119] Step 706: Use the new secondary recall information as the target information retrieval result.

[0120] Among them, if the difference cc_limit_current is less than the preset difference threshold cc_limit, it is considered that the correction effect is not obvious, and the secondary recall information corresponding to the target contribution degree is used as the target information retrieval result.

[0121] If the difference cc_limit_current is greater than or equal to the preset difference threshold cc_limit, it is considered that the correction effect is obvious, and the corresponding parameters in the information retrieval can be optimized based on the preset correction strategy corresponding to the optimization contribution degree to obtain new secondary recall information. Optionally, continuing with the above three optimization contribution degrees as an example, the preset correction strategy corresponding to the maximum optimization contribution degree among contri_rate_1, contri_rate_2, and contri_rate_3 can be used to optimize the corresponding parameters in the information retrieval and then perform secondary information retrieval again to obtain new secondary recall information, and finally use this new secondary recall information as the final target information retrieval result.

[0122] The present application provides an information retrieval method, which includes: obtaining the initial recall information returned by a preset knowledge base, performing summary processing on the initial recall information by using a preset large language model to obtain summary information, where the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate an output result expected by the user, performing secondary information retrieval based on the summary information to obtain the secondary recall information returned by the preset knowledge base, and generating a target information retrieval result based on the initial recall information and the secondary recall information. By performing secondary information retrieval based on the summary information, the accuracy of information retrieval is improved in this solution; in addition, by using the preset large language model to perform summary processing on the initial recall information, the obtained summary information is more accurate, thereby further improving the accuracy of information retrieval.

[0123] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0124] Figure 8 It is a structural block diagram of an information retrieval device provided by an embodiment of the present application.

[0125] As Figure 8 shown, the information retrieval device 800 includes:

[0126] An acquisition module 802, configured to acquire the initial recall information returned by a preset knowledge base.

[0127] A processing module 804, configured to perform summary processing on the initial recall information by using a preset large language model to obtain summary information; where the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate an output result expected by the user.

[0128] A retrieval module 806, configured to perform secondary information retrieval based on the summary information to obtain the secondary recall information returned by a preset knowledge base.

[0129] A generation module 808, configured to generate a target information retrieval result based on the initial recall information and the secondary recall information.

[0130] Regarding the device in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here. Each module in the above information retrieval device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations of each of the above modules.

[0131] In an embodiment of the present application, a computer device is provided. The computer device includes a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the following steps are implemented:

[0132] Obtain the initial recall information returned by the preset knowledge base;

[0133] Use a preset large language model to perform abstract processing on the initial recall information to obtain abstract information; wherein, the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate the output result expected by the user;

[0134] Perform secondary information retrieval based on the abstract information to obtain the secondary recall information returned by the preset knowledge base;

[0135] Generate a target information retrieval result based on the initial recall information and the secondary recall information.

[0136] In an embodiment of the present application, when the processor executes the computer program, the following steps are also implemented:

[0137] Obtain the first recall similarity corresponding to the initial recall information and the second recall similarity corresponding to the secondary recall information respectively;

[0138] Calculate the initial recall weight based on the first recall similarity and the preset similarity threshold, and calculate the secondary recall weight based on the second recall similarity and the preset similarity threshold;

[0139] Calculate the target contribution degree corresponding to the secondary recall information based on the initial recall weight, the secondary recall weight, and the preset recall result contribution coefficient;

[0140] Obtain the target information retrieval result based on the target contribution degree.

[0141] In an embodiment of the present application, when the processor executes the computer program, the following steps are also implemented:

[0142] Obtain at least one optimization contribution degree; wherein, the optimization contribution degree is obtained after optimizing at least one initial parameter corresponding in the information retrieval process according to at least one preset correction strategy in the preset correction strategy set;

[0143] Based on the optimization contribution degree and the target contribution degree, obtain the target information retrieval result.

[0144] In an embodiment of the present application, when the processor executes the computer program, the following steps are further implemented:

[0145] Obtain the difference between the maximum value and the minimum value in the optimization contribution degree and the target contribution degree;

[0146] Based on the difference and the preset difference threshold, determine the target information retrieval result.

[0147] In an embodiment of the present application, when the processor executes the computer program, the following steps are further implemented:

[0148] If the difference is less than the preset difference threshold, use the secondary recall information corresponding to the target contribution degree as the target information retrieval result;

[0149] If the difference is greater than or equal to the preset difference threshold, optimize the corresponding parameters in the information retrieval based on the preset correction strategy corresponding to the optimization contribution degree to obtain new secondary recall information;

[0150] Use the new secondary recall information as the target information retrieval result.

[0151] In an embodiment of the present application, when the processor executes the computer program, the following steps are further implemented:

[0152] Obtain at least one preset correction strategy in the preset correction strategy set;

[0153] Use at least one preset correction strategy to optimize at least one initial parameter corresponding in the information retrieval process to obtain optimized parameters; wherein, the initial parameters include a preset structured input sequence, a preset recall result contribution degree coefficient, and a preset maximum number of secondary recall loops;

[0154] Based on the optimized parameters and the abstract information, perform secondary information retrieval to obtain the secondary recall information returned by the preset knowledge base, and calculate the optimization contribution degree corresponding to the secondary recall information.

[0155] In an embodiment of the present application, when the processor executes the computer program, the following steps are further implemented:

[0156] If the second recall similarity is greater than the preset similarity threshold, then repeat the step of performing secondary information retrieval based on the summary information based on the preset maximum number of secondary recall loops, obtain the secondary recall information returned by the preset knowledge base, and generate the target information retrieval result based on the primary recall information and the secondary recall information;

[0157] If the second recall similarity is less than or equal to the preset similarity threshold, then continue to perform the step of generating the target information retrieval result based on the primary recall information and the secondary recall information.

[0158] The computer device provided by the embodiments of the present application has a similar implementation principle and technical effect to the above method embodiments, and will not be elaborated here.

[0159] In an embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0160] Obtain the primary recall information returned by the preset knowledge base;

[0161] Perform summary processing on the primary recall information using a preset large language model to obtain summary information; wherein, the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate the output result expected by the user;

[0162] Perform secondary information retrieval based on the summary information to obtain the secondary recall information returned by the preset knowledge base;

[0163] Generate the target information retrieval result based on the primary recall information and the secondary recall information.

[0164] In an embodiment of the present application, when the computer program is executed by a processor, the following steps are also implemented:

[0165] Obtain the first recall similarity corresponding to the primary recall information and the second recall similarity corresponding to the secondary recall information respectively;

[0166] Calculate the primary recall weight based on the first recall similarity and the preset similarity threshold, and calculate the secondary recall weight based on the second recall similarity and the preset similarity threshold;

[0167] Calculate the target contribution degree corresponding to the secondary recall information based on the primary recall weight, the secondary recall weight and the preset recall result contribution coefficient;

[0168] Obtain the target information retrieval result based on the target contribution degree.

[0169] In an embodiment of the present application, when the computer program is executed by a processor, the following steps are also implemented:

[0170] Obtain at least one optimization contribution degree; wherein, the optimization contribution degree is obtained by optimizing at least one initial parameter corresponding in the information retrieval process according to at least one preset correction strategy in a preset correction strategy set;

[0171] Based on the optimization contribution degree and the target contribution degree, obtain a target information retrieval result.

[0172] In an embodiment of the present application, when the computer program is executed by a processor, the following steps are further implemented:

[0173] Obtain the difference between the maximum value and the minimum value in the optimization contribution degree and the target contribution degree;

[0174] Based on the difference and a preset difference threshold, determine a target information retrieval result.

[0175] In an embodiment of the present application, when the computer program is executed by a processor, the following steps are further implemented:

[0176] If the difference is less than the preset difference threshold, use the secondary recall information corresponding to the target contribution degree as the target information retrieval result;

[0177] If the difference is greater than or equal to the preset difference threshold, optimize the corresponding parameters in the information retrieval based on the preset correction strategy corresponding to the optimization contribution degree to obtain new secondary recall information;

[0178] Use the new secondary recall information as the target information retrieval result.

[0179] In an embodiment of the present application, when the computer program is executed by a processor, the following steps are further implemented:

[0180] Obtain at least one preset correction strategy in the preset correction strategy set;

[0181] Use at least one preset correction strategy to optimize at least one initial parameter corresponding in the information retrieval process to obtain optimized parameters; wherein, the initial parameters include a preset structured input sequence, a preset recall result contribution degree coefficient, and a preset maximum number of secondary recall loops;

[0182] Based on the optimized parameters and summary information, perform secondary information retrieval to obtain secondary recall information returned by a preset knowledge base, and calculate the optimization contribution degree corresponding to the secondary recall information.

[0183] In an embodiment of the present application, when the computer program is executed by a processor, the following steps are further implemented:

[0184] If the second recall similarity is greater than the preset similarity threshold, then repeat the step of performing secondary information retrieval based on the summary information based on the preset maximum number of secondary recall loops, obtain the secondary recall information returned by the preset knowledge base, and generate the target information retrieval result based on the primary recall information and the secondary recall information;

[0185] If the second recall similarity is less than or equal to the preset similarity threshold, then continue to perform the step of generating the target information retrieval result based on the primary recall information and the secondary recall information.

[0186] The computer-readable storage medium provided in this embodiment has the same implementation principle and technical effects as the above method embodiment, and will not be elaborated here.

[0187] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. This computer program can be stored in a non-volatile computer-readable storage medium. When this computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in this application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0188] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure, which follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of the present disclosure are pointed out by the following claims.

[0189] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An information retrieval method, characterized in that: The method comprises: Obtain the initial recall information returned by the preset knowledge base; The initial recall information is summarized by using a preset large language model to obtain summary information; wherein the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate an output result expected by the user; Perform secondary information retrieval based on the summary information to obtain secondary recall information returned by the preset knowledge base; Based on the initial recall information and the secondary recall information, a target information retrieval result is generated.

2. The method according to claim 1, characterized in that The generating target information retrieval result based on the initial recall information and the secondary recall information includes: Respectively obtaining a first recall similarity corresponding to the initial recall information and a second recall similarity corresponding to the secondary recall information; Calculating a first recall weight based on the first recall similarity and a preset similarity threshold, and calculating a second recall weight based on the second recall similarity and the preset similarity threshold; Based on the initial recall weight, the secondary recall weight and the preset recall result contribution coefficient, calculating the target contribution corresponding to the secondary recall information; Based on the target contribution, the target information retrieval result is obtained.

3. The method according to claim 2, characterized in that The step of obtaining the target information retrieval result based on the target contribution includes: Obtaining at least one optimization contribution; wherein the optimization contribution is obtained by optimizing at least one initial parameter corresponding to the information retrieval process according to at least one preset correction strategy in the preset correction strategy set; Based on the optimization contribution and the target contribution, the target information retrieval result is obtained.

4. The method according to claim 3, characterized in that The obtaining of the target information retrieval result based on the optimization contribution and the target contribution includes: Obtaining a difference between a maximum value and a minimum value in the optimization contribution and the target contribution; Based on the difference and a preset difference threshold, the target information retrieval result is determined.

5. The method according to claim 4, characterized in that The step of determining the target information retrieval result based on the difference and a preset difference threshold value includes: If the difference is less than the preset difference threshold, the secondary recall information corresponding to the target contribution is used as the target information retrieval result; If the difference is greater than or equal to the preset difference threshold, then optimizing the corresponding parameters in the information retrieval based on the preset correction strategy corresponding to the optimization contribution to obtain new secondary recall information; The new secondary recall information is used as the target information retrieval result.

6. The method according to any one of claims 3 to 5, characterized in that: The obtaining of at least one optimization contribution comprises: Obtaining at least one preset correction strategy from a preset correction strategy set; The at least one preset correction strategy is used to optimize at least one initial parameter corresponding to the information retrieval process to obtain an optimized parameter; wherein the initial parameter includes a preset structured input sequence, a preset recall result contribution coefficient, and a preset maximum number of secondary recall cycles; A secondary information search is performed based on the optimization parameters and the summary information to obtain the secondary recall information returned by the preset knowledge base, and the optimization contribution corresponding to the secondary recall information is calculated.

7. The method according to any one of claims 2 to 5, characterized in that The method further comprises: If the second recall similarity is greater than the preset similarity threshold, then repeatedly performing the steps of performing secondary information retrieval based on the summary information based on a preset secondary recall maximum number of cycles, obtaining secondary recall information returned by the preset knowledge base, and generating a target information retrieval result based on the initial recall information and the secondary recall information; If the second recall similarity is less than or equal to the preset similarity threshold, then continue to execute the step of generating a target information retrieval result based on the initial recall information and the secondary recall information.

8. An information retrieval device, characterized in that: The device comprises: An acquisition module, used to obtain the initial recall information returned by the preset knowledge base; A processing module, configured to perform summary processing on the initial recall information using a preset large language model to obtain summary information; wherein the input information of the preset large language model includes a preset structured input sequence, and the preset structured input sequence is used to guide the preset large language model to generate an output result expected by the user; A retrieval module, used to perform secondary information retrieval based on the summary information to obtain secondary recall information returned by the preset knowledge base; A generation module is used to generate a target information retrieval result based on the initial recall information and the secondary recall information.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the information retrieval method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the information retrieval method as described in any one of claims 1-7.

Images