Automatic retrieval generation method and device, computer equipment and storage medium
Through automated data retrieval and analysis methods, the preset interface of the agent is used to solve the problem of low data processing efficiency in daily environmental protection management, and efficient data retrieval and analysis result generation is achieved.
Patent Information
- Application Number
- CN202411983866.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-25
AI Technical Summary
In daily environmental protection management, when processing a large amount of unstructured environmental impact assessment related document data, it is necessary to sort out and retrieve a large amount of management requirements and basic standard data, resulting in large workload and time-consuming.
By obtaining target tasks and target data, inputting the initial model for searching and generation, adjusting model parameters to obtain the target model, and finally, based on the target tasks and target models, the preset interface of the agent is automatically retrieved and generated, reducing manual workload and time-consuming.
Automatic data retrieval and analysis result generation is realized, reducing manual workload and time-consuming, and improving processing efficiency.
Smart Images
Figure CN119884343A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of computer technology, and in particular to a method, an apparatus, a computer device, and a storage medium for automatic retrieval generation. Background Art
[0002] With the development of economy and society, the importance of environmental protection has gradually increased, and the environmental impact assessment data has continued to increase, and appropriate methods need to be used to process these data.
[0003] In response to the above problems, when relevant technologies process a large amount of unstructured environmental impact assessment-related document data in daily environmental protection management, they need to sort out and retrieve a large amount of management requirements and basic standard data, and make corresponding evaluation suggestions based on the retrieved data, which is labor-intensive and time-consuming.
[0004] Therefore, the related technologies have the problems of large workload and long time consumption when retrieving data and generating analysis results. Summary of the invention
[0005] In view of this, the present disclosure provides a method, apparatus, computer device, and storage medium for automatic retrieval and generation to solve the problems of large workload and long time consumption in related technologies when retrieving data and generating analysis results.
[0006] In a first aspect, the present disclosure provides a method for automatic retrieval generation, the method comprising:
[0007] Obtain target tasks and target data;
[0008] Input the target data into the initial model for retrieval and generation to obtain the initial results;
[0009] Adjust the parameters of the initial model according to the initial results to obtain the target model;
[0010] Based on the target task and target model, the preset interface of the intelligent agent is called to automatically retrieve and generate the target result of the target task.
[0011] In the disclosed embodiment, the target task and target data are obtained; the target data is input into the initial model for retrieval and generation to obtain the initial result; the parameters of the initial model are adjusted according to the initial result to obtain the target model; based on the target task and the target model, the preset interface of the intelligent agent is called to automatically retrieve and generate to obtain the target result of the target task. Since the disclosed embodiment is based on the target task and the target model, the preset interface of the intelligent agent is called to automatically perform a comprehensive search and generate the target result of the target task, which reduces the manual workload and saves time.
[0012] In an optional implementation, before inputting the target data into the initial model for retrieval and generation to obtain the initial result, the method further includes:
[0013] Performing data preprocessing on the target data to obtain preprocessed target data;
[0014] The preprocessed target data is input into the initial model for retrieval and generation.
[0015] In the embodiments of the present disclosure, by preprocessing the data, the quality and accuracy of the data can be improved, the consistency of the data can be enhanced, and it can be ensured that the initial model accurately understands the data information.
[0016] In an optional implementation, before inputting the target data into the initial model for retrieval and generation to obtain the initial result, the method further includes:
[0017] Select the corresponding large model according to the target task and build the generation module;
[0018] Construct a retrieval module based on the target task and target data;
[0019] Based on the generation module and the retrieval module, an initial model is constructed.
[0020] In the disclosed embodiment, by selecting a corresponding large model according to the target task and constructing a generation module, an initial result can be generated. By constructing a retrieval module according to the target task and the target data, a comprehensive search can be performed. By constructing an initial model according to the generation module and the retrieval module, a comprehensive search can be performed and an initial result can be generated.
[0021] In an optional implementation, before automatically retrieving and generating based on the target task and the target model by calling the preset interface of the agent to obtain the target result of the target task, the method further includes:
[0022] Obtain the functional requirements of the target task;
[0023] Build multiple agents based on functional requirements;
[0024] A first preset interface is constructed according to the interaction relationship between the intelligent agent and the target model, a second preset interface is constructed according to the interaction relationship between the intelligent agent and the external environment, and a third preset interface is constructed according to the interaction relationship between the intelligent agent and the user.
[0025] In the disclosed embodiment, by constructing multiple agents according to the functional requirements of the acquired target task, specific tasks and functions can be assigned to each agent. By constructing a first preset interface according to the interactive relationship between the agent and the target model, constructing a second preset interface according to the interactive relationship between the agent and the external environment, and constructing a third preset interface according to the interactive relationship between the agent and the user, interactive collaboration between the agent and the target model, the external environment, and the user can be achieved to ensure the joint completion of the target task.
[0026] In an optional implementation, based on the target task and the target model, the preset interface of the agent is called to automatically retrieve and generate, and the target result of the target task is obtained, including:
[0027] Acquire the user's query information according to the third preset interface;
[0028] The first preset interface, the second preset interface and the third preset interface are called according to the query information to output the target result.
[0029] In the embodiment of the present disclosure, by calling the first preset interface, the second preset interface and the third preset interface according to the acquired query information of the user, the target result can be automatically output, the manual workload is reduced, and the time consumption is shortened.
[0030] In an optional implementation, calling the first preset interface, the second preset interface, and the third preset interface according to the query information to output the target result includes:
[0031] Calling a first preset interface according to the query information to obtain a retrieval module in the target model;
[0032] Calling a second preset interface according to the search module to obtain external resource information;
[0033] Calling a first preset interface according to the query information to obtain a generated module in the target model;
[0034] Transmitting the external resource information and the query information to the generation module through the first preset interface to generate the target result;
[0035] The target result is output through the third preset interface.
[0036] In the disclosed embodiment, by calling the first preset interface according to the query information to obtain the retrieval module in the target model, and then calling the second preset interface according to the retrieval module to obtain the external resource information, and then calling the first preset interface according to the query information to obtain the generation module in the target model, and then transmitting the external resource information and the query information to the generation module through the first preset interface to generate the target result, and finally outputting the target result through the third preset interface, it is possible to automatically perform a comprehensive search and generate the target result of the target task, thereby reducing the manual workload and shortening the time consumption.
[0037] In an optional implementation, after automatically searching and generating based on the target task and the target model by calling the preset interface of the agent and obtaining the target result of the target task, the method further includes:
[0038] Get feedback from users;
[0039] The parameters of the target model are adjusted according to the feedback information to obtain the optimized model.
[0040] In the embodiment of the present disclosure, by adjusting the parameters of the target model according to the acquired user feedback information, the accuracy and performance of the target model can be improved to obtain an optimized model.
[0041] In a second aspect, the present disclosure provides a device for automatic retrieval generation, the device comprising:
[0042] A first acquisition module, used to acquire target tasks and target data;
[0043] The first obtaining module is used to input the target data into the initial model for retrieval and generation to obtain the initial result;
[0044] The second module is used to adjust the parameters of the initial model according to the initial results to obtain the target model;
[0045] The third obtaining module is used to call the preset interface of the intelligent agent to automatically retrieve and generate based on the target task and the target model, and obtain the target result of the target task.
[0046] In a third aspect, the present disclosure provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the automatic retrieval and generation method of the above-mentioned first aspect or any corresponding embodiment thereof by executing the computer instructions.
[0047] In a fourth aspect, the present disclosure provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for automatic retrieval and generation of the above-mentioned first aspect or any corresponding embodiment thereof.
[0048] In a fifth aspect, the present disclosure provides a computer program product, including computer instructions, which are used to enable a computer to execute the automatic retrieval and generation method of the above-mentioned first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0050] Figure 1 is a flowchart of a method for automatic retrieval generation according to an embodiment of the present disclosure;
[0051] Figure 2 is a structural block diagram of a method for automatic retrieval generation according to an embodiment of the present disclosure;
[0052] Figure 3 is a structural block diagram of an automatic retrieval and generation device according to an embodiment of the present disclosure;
[0053] Figure 4 It is a schematic diagram of the hardware structure of the computer device of the embodiment of the present disclosure. DETAILED DESCRIPTION
[0054] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.
[0055] With the development of economy and society, the importance of environmental protection has gradually increased, and the environmental impact assessment data has continued to increase, and appropriate methods need to be used to process these data.
[0056] In response to the above problems, when relevant technologies process a large amount of unstructured environmental impact assessment-related document data in daily environmental protection management, they need to sort out and retrieve a large amount of management requirements and basic standard data, and make corresponding evaluation suggestions based on the retrieved data, which is labor-intensive and time-consuming.
[0057] Therefore, the related technologies have the problems of large workload and long time consumption when retrieving data and generating analysis results.
[0058] In order to solve the above problems, according to an embodiment of the present disclosure, a method embodiment of automatic retrieval generation is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0059] In this embodiment, a method for automatic retrieval generation is provided, such as Figure 1 As shown, Figure 1 : is a flow chart of the automatic retrieval generation method according to an embodiment of the present disclosure, which can be applied to a server and includes the following steps:
[0060] Step S101, obtaining target tasks and target data.
[0061] Optionally, in the disclosed embodiment, the target task is a core task in a preset field, such as a core task in the field of environmental protection, including environmental pollution source identification, environmental impact assessment, resource utilization optimization, etc. The target data is data related to the target task, such as environmental protection reports submitted by enterprises, environmental monitoring data, emission records, environmental protection requirements documents of the industry, etc.
[0062] Specifically, after obtaining the target task, the server obtains target data from the relevant platform according to the target task. For example, the server can use crawler technology to collect air quality data (such as the concentration of pollutants such as PM2.5, PM10, sulfur dioxide, nitrogen dioxide, etc.) provided by environmental monitoring stations, water quality data (such as dissolved oxygen, pH value, ammonia nitrogen, total phosphorus, etc.) provided by water quality monitoring stations, and industry environmental protection requirements documents and other relevant data from the environmental protection information platform to obtain the target data.
[0063] Step S102, input the target data into the initial model for retrieval and generation to obtain initial results.
[0064] Optionally, in the disclosed embodiment, the retrieval adopts the retrieval augmented generation technology (RAG), and the generation adopts the large model (LLM). The initial model combines the two technologies of information retrieval and language generation. The relevant data is retrieved from the relevant platform through RAG and input into the LLM to generate the relevant evaluation report (such as environmental protection evaluation). The initial result is used to represent the performance evaluation result of the initial model.
[0065] Specifically, the server inputs the target data into the initial model, retrieves relevant information from the relevant field database through the initial model, and generates a relevant evaluation report based on the retrieved relevant information, and then compares the target data with the retrieved relevant information and the generated relevant evaluation report to obtain the initial result, that is, the performance evaluation result of the initial model.
[0066] Step S103, adjusting the parameters of the initial model according to the initial results to obtain the target model.
[0067] Optionally, in the embodiment of the present disclosure, the target model refers to a model obtained by optimizing the initial model.
[0068] Specifically, based on the initial results, the server can adjust the parameters of the initial model through hyperparameter search, gradient descent and other methods to obtain the target model, that is, the model obtained after optimizing the initial model.
[0069] Step S104, based on the target task and the target model, the preset interface of the intelligent agent is called to automatically search and generate, and the target result of the target task is obtained.
[0070] Optionally, in the disclosed embodiments, an agent is a concept in the field of artificial intelligence. It is an intelligent entity with environmental perception, autonomous understanding, decision-making and action execution capabilities. It can autonomously perceive information in its environment and make decisions based on this information to achieve specific goals or tasks. The preset interface of the agent is used to implement specific interactive functions. The target result refers to the evaluation report obtained after the target model calls the preset interface of the agent according to the target task, automatically retrieves and generates the result, wherein the evaluation report contains corresponding improvement suggestions, etc.
[0071] Specifically, the server automatically retrieves and generates the target result of the target task, that is, the generated evaluation report, by calling the preset interface of the intelligent agent through the target model according to the target task.
[0072] In the disclosed embodiment, the target task and target data are obtained; the target data is input into the initial model for retrieval and generation to obtain the initial result; the parameters of the initial model are adjusted according to the initial result to obtain the target model; based on the target task and the target model, the preset interface of the intelligent agent is called to automatically retrieve and generate to obtain the target result of the target task. Since the disclosed embodiment is based on the target task and the target model, the preset interface of the intelligent agent is called to automatically perform a comprehensive search and generate the target result of the target task, which reduces the manual workload and saves time.
[0073] In some optional implementations, before inputting the target data into the initial model for retrieval and generation to obtain the initial results, the method further includes:
[0074] Performing data preprocessing on the target data to obtain preprocessed target data;
[0075] The preprocessed target data is input into the initial model for retrieval and generation.
[0076] Optionally, in the embodiment of the present disclosure, the server may perform data preprocessing on the target data before inputting the target data into the initial model for retrieval and generation, including operations such as data cleaning, data enhancement, data missing value filling, data normalization, word segmentation, and data labeling.
[0077] It should be noted that data cleaning refers to the removal of noise, duplicates and irrelevant information from the target data; data enhancement refers to text replacement, synonym replacement and other operations on the target data; missing data value filling can be done by filling in the missing target data using methods such as mean filling, median filling, mode filling and previous and next data filling; data normalization refers to scaling the target data so that it falls into a small specific interval, usually [0, 1] or [-1, 1]; word segmentation refers to decomposing long text target data such as sentences, paragraphs, and articles into a data structure based on words for subsequent processing and analysis. For example, the server can use the International Chemical Dictionary as a word segmentation dictionary to perform word segmentation on the text target data and extract key information such as pollutant types, emissions, emission standards, etc.; data labeling refers to labeling the target data that has undergone the above preprocessing operations to form training sets and test sets, where the labeling content can be environmental evaluation levels, existing problems, and improvement suggestions.
[0078] Specifically, the server can perform data preprocessing operations such as data cleaning, data enhancement, data missing value filling, data normalization, word segmentation, etc. on the target data, and then label the target data that has undergone the above preprocessing operations to obtain training sets and test sets, and input the training sets and test sets into the initial model for retrieval and generation.
[0079] In the embodiments of the present disclosure, by preprocessing the data, the quality and accuracy of the data can be improved, the consistency of the data can be enhanced, and it can be ensured that the initial model accurately understands the data information.
[0080] In some optional implementations, before inputting the target data into the initial model for retrieval and generation to obtain the initial results, the method further includes:
[0081] Select the corresponding large model according to the target task and build the generation module;
[0082] Construct a retrieval module based on the target task and target data;
[0083] Based on the generation module and the retrieval module, an initial model is constructed.
[0084] Optionally, in the disclosed embodiment, the retrieval module is used to retrieve relevant information from the relevant platform according to the target task, and the generation module is used to generate a relevant evaluation report according to the retrieved relevant information. The initial model includes a retrieval module and a generation module.
[0085] First, the server selects the corresponding large model according to the target task, such as the Generative Pre-trained Transformer (GPT) series model, and builds a generation module.
[0086] Then, the server plans the content of the search engine according to the target task, including determining the target user group of the search engine (such as professionals, students, researchers in the petrochemical industry, etc.), the content scope that the search engine needs to cover (such as petrochemical technical documents, academic papers, product information, etc.), and the functional requirements of the search engine (such as keyword search, advanced search, result sorting, related recommendations, etc.); and plans the architecture of the search engine, including an index module, a query module, a result display module, etc., among which the index module is used to index the pre-processed target data and establish data structures such as inverted indexes; the query module is used to support users to enter keywords for search and return relevant results, and the module can be constructed using search algorithms such as word segmentation technology, word frequency statistics, and weight calculation; the result display module is used to display search results to users.
[0087] Next, the server organizes dictionaries in related fields according to the target data (such as organizing a large chemical dictionary, which includes basic terms and professional terms in the petrochemical field), extracts thesaurus in related fields (such as analyzing petrochemical materials, extracting high-frequency words and phrases, and forming a specific thesaurus in the petrochemical field), and establishes databases in related fields (such as integrating the large chemical dictionary and high-frequency words into the search engine's thesaurus and establishing an environmental protection information platform database).
[0088] Finally, the server builds a retrieval module based on the search engine and related domain databases, and builds an initial model based on the generation module and the retrieval module.
[0089] In the disclosed embodiment, by selecting a corresponding large model according to the target task and constructing a generation module, an initial result can be generated. By constructing a retrieval module according to the target task and the target data, a comprehensive search can be performed. By constructing an initial model according to the generation module and the retrieval module, a comprehensive search can be performed and an initial result can be generated.
[0090] In some optional implementations, before automatically retrieving and generating based on the target task and the target model by calling the preset interface of the agent to obtain the target result of the target task, the method further includes:
[0091] Obtain the functional requirements of the target task;
[0092] Build multiple agents based on functional requirements;
[0093] A first preset interface is constructed according to the interaction relationship between the intelligent agent and the target model, a second preset interface is constructed according to the interaction relationship between the intelligent agent and the external environment, and a third preset interface is constructed according to the interaction relationship between the intelligent agent and the user.
[0094] Optionally, in the embodiment of the present disclosure, the functional requirements of the target task include data collection, analysis and prediction, requirement specification, etc. The first preset interface is used to realize the interaction between the intelligent agent and the target model, the second preset interface is used to realize the interaction between the intelligent agent and the external environment, and the third preset interface is used to realize the interaction between the intelligent agent and the user.
[0095] Specifically, the server constructs multiple agents (such as data collection agents, analysis and prediction agents, requirement-specified agents, etc.) according to the functional requirements of the acquired target tasks, and then constructs a first preset interface according to the interaction relationship between the agent and the target model, a second preset interface according to the interaction relationship between the agent and the external environment, and a third preset interface according to the interaction relationship between the agent and the user.
[0096] It should be noted that the interaction between the agent and the target model includes natural language communication and structured data communication, where natural language refers to a language that evolves naturally with culture, such as Chinese and English; structured data is data organized according to certain formats and rules, usually with a clear structure. Common forms of structured data include tables in databases, where data is organized into rows and columns, each column represents a specific attribute or field, and each row represents a record. The interaction between the agent and the external environment includes the interaction between the agent and the database, application programming interface, etc. The interaction between the agent and the user includes the user inputting instructions and the user obtaining results.
[0097] In the disclosed embodiment, by constructing multiple agents according to the functional requirements of the acquired target task, specific tasks and functions can be assigned to each agent. By constructing a first preset interface according to the interactive relationship between the agent and the target model, constructing a second preset interface according to the interactive relationship between the agent and the external environment, and constructing a third preset interface according to the interactive relationship between the agent and the user, interactive collaboration between the agent and the target model, the external environment, and the user can be achieved to ensure the joint completion of the target task.
[0098] In some optional implementations, based on the target task and the target model, the preset interface of the agent is called to automatically retrieve and generate, and the target result of the target task is obtained, including:
[0099] Acquire the user's query information according to the third preset interface;
[0100] The first preset interface, the second preset interface and the third preset interface are called according to the query information to output the target result.
[0101] Optionally, in the embodiment of the present disclosure, the query information may be instructions or questions in natural language form, used to enable the agent to perform specific operations or provide specific information.
[0102] Specifically, the server calls the first preset interface, the second preset interface, and the third preset interface according to the query information of the user obtained by the third preset interface, and outputs the target result.
[0103] In the embodiment of the present disclosure, by calling the first preset interface, the second preset interface and the third preset interface according to the acquired query information of the user, the target result can be automatically output, the manual workload is reduced, and the time consumption is shortened.
[0104] In some optional implementations, calling the first preset interface, the second preset interface, and the third preset interface according to the query information to output the target result includes:
[0105] Calling a first preset interface according to the query information to obtain a retrieval module in the target model;
[0106] Calling a second preset interface according to the search module to obtain external resource information;
[0107] Calling a first preset interface according to the query information to obtain a generated module in the target model;
[0108] Transmitting the external resource information and the query information to the generation module through the first preset interface to generate the target result;
[0109] The target result is output through the third preset interface.
[0110] Optionally, in the embodiment of the present disclosure, after calling the third preset interface to obtain query information, the server first performs internal processing through the intelligent agent, including understanding the task intent, analyzing the task requirements, and determining the steps and resources required to perform the task, and can use its own relevant domain knowledge base, reasoning capabilities and target models to generate a preliminary plan for performing the task.
[0111] Then, the server takes specific actions to execute the task through the intelligent agent according to the preliminary plan. During the execution stage, the intelligent agent will monitor the execution of the task in real time and adjust the execution plan as needed, including calling the first preset interface according to the query information to obtain the generation module in the target model, calling the second preset interface according to the retrieval module to obtain external resource information, calling the first preset interface according to the query information to obtain the generation module in the target model, transmitting the external resource information and query information to the generation module through the first preset interface to generate the target result.
[0112] Finally, the server observes and analyzes the target result through the intelligent agent. If the result meets expectations, the server outputs the target result through a third preset interface and may provide additional explanations or suggestions. If the result does not meet expectations, the server re-executes the task, adjusts the execution plan, or reports the error to the user through the intelligent agent.
[0113] In the disclosed embodiment, by calling the first preset interface according to the query information to obtain the retrieval module in the target model, and then calling the second preset interface according to the retrieval module to obtain the external resource information, and then calling the first preset interface according to the query information to obtain the generation module in the target model, and then transmitting the external resource information and the query information to the generation module through the first preset interface to generate the target result, and finally outputting the target result through the third preset interface, it is possible to automatically perform a comprehensive search and generate the target result of the target task, thereby reducing the manual workload and shortening the time consumption.
[0114] In some optional implementations, after automatically retrieving and generating based on the target task and the target model by calling the preset interface of the agent and obtaining the target result of the target task, the method further includes:
[0115] Get feedback from users;
[0116] The parameters of the target model are adjusted according to the feedback information to obtain the optimized model.
[0117] Optionally, in the embodiment of the present disclosure, the feedback information refers to the user's suggestions and opinions on the target result generated by the target model.
[0118] Specifically, the server can obtain user feedback information through questionnaires, online comments, customer service feedback, etc., and then organize and analyze the collected feedback information, identify hot issues and potential needs of users, use natural language processing technology to judge the sentiment tendency of feedback information, and understand the user's satisfaction and preference for the target results generated by the target model. Then, the server adjusts the parameters and structure of the target model according to the feedback information, and inputs the feedback information as new training data into the target model for incremental learning or transfer learning, so that the target model can better adapt to actual needs.
[0119] For example, when the server predicts air quality through the target model, it finds that the PM2.5 concentration in a certain area continues to exceed the standard. Through file mining, the server finds that the area has not implemented effective emission reduction measures recently. At the same time, the user feedback information obtained by the server also mentioned that the industrial emissions in the area are serious and hopes to strengthen supervision. In response to the above problems, the server can adjust the parameters of the target model based on the above information, add industrial emissions as a characteristic variable, and increase the prediction weight of emission reduction measures.
[0120] In the embodiment of the present disclosure, by adjusting the parameters of the target model according to the acquired user feedback information, the accuracy and performance of the target model can be improved to obtain an optimized model.
[0121] In some optional embodiments, such as Figure 2 As shown, Figure 2 It is a structural flowchart of the method for automatic retrieval generation according to an embodiment of the present disclosure, including the following modules: demand assessment (including business direction size, data volume assessment, basic configuration), data collection (including data planning, internal data extraction, external data crawling, knowledge object collection), data preprocessing (including data cleaning, referencing word segmentation dictionaries, word segmenter training, and determining required tools), model pre-training (including model architecture implementation, training analysis, experimental strategies, verification and optimization), business association (including establishing business relationships and building interactive methods) and model deployment (including model compression, model iterative optimization, deployment, evaluation and adjustment).
[0122] In the present embodiment, a device for automatic retrieval generation is also provided, and the device is used to implement the above-mentioned embodiment and preferred implementation mode, and the description has been made no further. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented with software, the implementation of hardware, or a combination of software and hardware is also possible and conceived.
[0123] This embodiment provides a device for automatically searching and generating. Figure 3 As shown, including:
[0124] The first acquisition module 301 is used to acquire target tasks and target data;
[0125] The first obtaining module 302 is used to input the target data into the initial model for retrieval and generation to obtain the initial result;
[0126] The second obtaining module 303 is used to adjust the parameters of the initial model according to the initial results to obtain the target model;
[0127] The third obtaining module 304 is used to call the preset interface of the intelligent agent to automatically search and generate based on the target task and the target model, so as to obtain the target result of the target task.
[0128] In the disclosed embodiment, the target task and target data are obtained; the target data is input into the initial model for retrieval and generation to obtain the initial result; the parameters of the initial model are adjusted according to the initial result to obtain the target model; based on the target task and the target model, the preset interface of the intelligent agent is called to automatically retrieve and generate to obtain the target result of the target task. Since the disclosed embodiment is based on the target task and the target model, the preset interface of the intelligent agent is called to automatically perform a comprehensive search and generate the target result of the target task, which reduces the manual workload and saves time.
[0129] In some optional embodiments, the device further comprises:
[0130] The fourth obtaining module is used to perform data preprocessing on the target data to obtain the preprocessed target data;
[0131] The input module is used to input the preprocessed target data into the initial model for retrieval and generation.
[0132] In some optional embodiments, the device further comprises:
[0133] The first construction module is used to select the corresponding large model according to the target task and construct the generation module;
[0134] The second construction module is used to construct a retrieval module according to the target task and target data;
[0135] The third building module is used to build an initial model according to the generation module and the retrieval module.
[0136] In some optional embodiments, the device further comprises:
[0137] The second acquisition module is used to obtain the functional requirements of the target task;
[0138] The fourth building module is used to build multiple agents according to functional requirements;
[0139] The fifth construction module is used to construct a first preset interface according to the interaction relationship between the intelligent agent and the target model, a second preset interface according to the interaction relationship between the intelligent agent and the external environment, and a third preset interface according to the interaction relationship between the intelligent agent and the user.
[0140] In some optional implementations, the third obtaining module 304 includes:
[0141] An acquisition unit, used for acquiring the user's query information according to a third preset interface;
[0142] The output unit is used to call the first preset interface, the second preset interface and the third preset interface according to the query information to output the target result.
[0143] In some optional embodiments, the output unit includes:
[0144] A first acquisition submodule is used to call a first preset interface according to the query information to acquire a retrieval module in the target model;
[0145] The obtaining submodule is used to call the second preset interface according to the retrieval module to obtain external resource information;
[0146] A second acquisition submodule is used to call the first preset interface according to the query information to obtain the generated module in the target model;
[0147] A generation submodule, used to transmit the external resource information and the query information to the generation module through the first preset interface to generate a target result;
[0148] The output submodule is used to output the target result through a third preset interface.
[0149] In some optional embodiments, the device further comprises:
[0150] The third acquisition module is used to obtain user feedback information;
[0151] The fifth obtaining module is used to adjust the parameters of the target model according to the feedback information to obtain the optimized model.
[0152] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.
[0153] The automatic retrieval generation device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0154] The present disclosure also provides a computer device having the above Figure 3 The automatic retrieval generation device shown.
[0155] See also Figure 4 , Figure 4 is a schematic diagram of a computer device provided by an optional embodiment of the present disclosure, such as Figure 4As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.
[0156] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0157] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.
[0158] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0159] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0160] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.
[0161] The embodiments of the present disclosure also provide a computer-readable storage medium. The above-mentioned method according to the embodiments of the present disclosure can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium and downloaded through a network, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0162] A part of the present disclosure may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present disclosure through the operation of the computer. Those skilled in the art should understand that the existence of computer program instructions in computer-readable media includes, but is not limited to, source files, executable files, installation package files, etc., and accordingly, the way in which computer program instructions are executed by a computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.
[0163] Although the embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A method for automatic retrieval generation, characterized in that: The method comprises: Obtain target tasks and target data; Inputting the target data into the initial model for retrieval and generation to obtain an initial result; Adjusting the parameters of the initial model according to the initial results to obtain a target model; Based on the target task and the target model, the preset interface of the intelligent agent is called to automatically search and generate, so as to obtain the target result of the target task.
2. The method according to claim 1, characterized in that: Before inputting the target data into the initial model for retrieval and generation to obtain the initial result, the method further includes: Performing data preprocessing on the target data to obtain preprocessed target data; The preprocessed target data is input into the initial model for retrieval and generation.
3. The method according to claim 2, characterized in that Before inputting the target data into the initial model for retrieval and generation to obtain the initial result, the method further includes: Select a corresponding large model according to the target task and construct a generation module; Constructing a retrieval module according to the target task and the target data; The initial model is constructed according to the generation module and the retrieval module.
4. The method according to claim 3, characterized in that Before automatically searching and generating based on the target task and the target model by calling the preset interface of the intelligent agent to obtain the target result of the target task, the method further includes: Obtaining functional requirements of the target task; Construct a plurality of the intelligent agents according to the functional requirements; A first preset interface is constructed according to the interaction relationship between the intelligent agent and the target model, a second preset interface is constructed according to the interaction relationship between the intelligent agent and the external environment, and a third preset interface is constructed according to the interaction relationship between the intelligent agent and the user.
5. The method according to claim 4, characterized in that Based on the target task and the target model, calling the preset interface of the intelligent agent to automatically search and generate, and obtaining the target result of the target task, includes: Acquiring the user's query information according to the third preset interface; The first preset interface, the second preset interface and the third preset interface are called according to the query information to output the target result.
6. The method according to claim 5, characterized in that The calling the first preset interface, the second preset interface, and the third preset interface according to the query information to output the target result includes: Calling the first preset interface according to the query information to obtain the retrieval module in the target model; Calling the second preset interface according to the search module to obtain external resource information; Calling the first preset interface according to the query information to obtain the generation module in the target model; Transmitting the external resource information and the query information to the generation module through the first preset interface to generate the target result; The target result is output through the third preset interface.
7. The method according to any one of claims 1 to 6, characterized in that: After automatically searching and generating based on the target task and the target model by calling the preset interface of the agent to obtain the target result of the target task, the method further includes: Get feedback from users; The parameters of the target model are adjusted according to the feedback information to obtain an optimized model.
8. An automatic retrieval generation device, characterized in that: The device comprises: A first acquisition module, used to acquire target tasks and target data; The first obtaining module is used to input the target data into the initial model for retrieval and generation to obtain an initial result; A second obtaining module is used to adjust the parameters of the initial model according to the initial result to obtain a target model; The third obtaining module is used to call the preset interface of the intelligent agent to automatically retrieve and generate based on the target task and the target model, so as to obtain the target result of the target task.
9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the automatic retrieval generation method according to any one of claims 1 to 7 by executing the computer instructions.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the automatic retrieval generation method according to any one of claims 1 to 7.