Simulation computing APP application process recommendation method, device and system

Abstract

The present invention relates to the technical field of large language models, and specifically discloses a simulation computing APP application process recommendation method, device and system, including: receiving question information of the simulation computing APP application process constructed by the user; inputting the question information of the simulation computing APP application process constructed into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; performing entity association according to the simulation computing application process recommendation result to obtain the simulation computing application flow chart, wherein the simulation computing application flow chart can complete the simulation computing and obtain the simulation computing result when it is run. The simulation computing APP application process recommendation method provided by the present invention can obtain the simulation computing APP application process efficiently and without being restricted by professional knowledge.

Description

Technical Field

[0001] The present invention relates to the technical field of large language models, and in particular to a method for recommending a simulation computing APP application process, a simulation computing APP application process recommendation device, and a simulation computing APP application process recommendation system. Background Art

[0002] When building a common application task, users may not know which steps to take to achieve their target tasks. In reality, they often need to be taught by experienced personnel. In the process, they may face some communication efficiency problems, and may need multiple exchanges to complete the task. It is necessary to collect the task process experience of each department to form experience knowledge about the task process.

[0003] The simulation computing APP required to complete the common model task can form a simulation computing APP application process. In the prior art, the simulation computing APP application process specifically refers to connecting multiple categories of simulation computing APPs to complete complex simulation computing tasks. There is a data flow between the front and rear connected APPs, starting from the initial data, passing through multiple APPs, to complete the calculation of intermediate data and process data. Because the output data of the previous APP is used as the input data of the next APP, its numerical correctness and even file format will affect the results of the APP calculation and whether the process can run correctly. Therefore, the application process needs to be constructed by multi-disciplinary professional application personnel. With the increase of APPs, the requirements for application process builders are getting higher and higher. It is necessary to continuously learn APP knowledge to complete the construction and operation of the application process, which is not only inefficient but also requires high professional knowledge of personnel.

[0004] Therefore, how to obtain the simulation computing APP application process efficiently and without being restricted by professional knowledge has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] The present invention provides a simulation computing APP application process recommendation method, a simulation computing APP application process recommendation device and a simulation computing APP application process recommendation system, which solve the problem in the related art that the simulation computing APP application process cannot be obtained efficiently and without being restricted by professional knowledge.

[0006] As a first aspect of the present invention, a method for recommending a simulation computing APP application process is provided, which includes:

[0007] Receive the question information entered by the user for constructing the simulation computing APP application process;

[0008] Input the question information constructed by the simulation computing APP application process into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; wherein the simulation computing application process recommendation large language model at least includes a large language model and a simulation computing domain knowledge vector library connected to the large language model, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain a vector retrieval result, and the large language model can output the simulation computing application process recommendation result according to the vector retrieval result;

[0009] Entity association is performed according to the simulation calculation application process recommendation result to obtain a simulation calculation application flow chart, wherein the simulation calculation application flow chart can complete the simulation calculation and obtain the simulation calculation result when it is executed.

[0010] Furthermore, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain vector search results, including:

[0011] Preprocessing the question information constructed by the simulation computing APP application process to obtain a question information vector;

[0012] Performing a similarity search based on the question information vector and obtaining a similarity search result;

[0013] If the similarity search result is greater than or equal to a preset similarity threshold, it is determined that the simulation computing domain knowledge vector library has a corresponding simulation computing application process;

[0014] If the similarity search result is less than the preset similarity threshold, it is determined that the simulation computing domain knowledge vector library does not have a corresponding simulation computing application process.

[0015] Furthermore, the large language model can output simulation computing application process recommendation results according to the vector search results, including:

[0016] When the vector search result indicates that the simulation computing domain knowledge vector library has a corresponding simulation computing application process, a large language model prompt word is generated according to the simulation computing application process and the question information constructed by the simulation computing APP application process;

[0017] The simulation computing application process recommendation result is obtained according to the large language model prompt word.

[0018] Furthermore, the large language model can output simulation computing application process recommendation results according to the vector search results, including:

[0019] When the vector search result is that the simulation computing field knowledge vector library does not have a corresponding simulation computing application process, the question information constructed by the simulation computing APP application process is input into the intelligent agent to obtain the intelligent agent recommended application process, wherein the intelligent agent can call the classification retrieval algorithm according to the question information constructed by the simulation computing APP application process to obtain the intelligent agent recommended application process;

[0020] Generate a large language model prompt word according to the question information constructed by the agent recommendation application process and the simulation computing APP application process;

[0021] The simulation computing application process recommendation result is obtained according to the large language model prompt word.

[0022] Furthermore, the agent can call a classification retrieval algorithm according to the question information constructed by the simulation computing APP application process to obtain the agent recommended application process, including:

[0023] Determine whether there is an application process execution step in the question information constructed by the simulation calculation APP application process;

[0024] If there is no application process execution step, prompt the user to enter the application process execution step;

[0025] If there are application process execution steps, the classification retrieval algorithm is called according to each application process execution step to obtain the classification retrieval result of each application process execution step;

[0026] Among them, the classification retrieval algorithm can retrieve the simulation computing field knowledge vector library according to the description information of each application process execution step to obtain multiple APPs to be grouped for each application process execution step, and group them according to the application field of each APP to obtain the group of each application process execution step and its corresponding APP.

[0027] Further, entity association is performed according to the simulation computing application process recommendation result to obtain a simulation computing application flow chart, including:

[0028] Determine the recommended APP according to the simulation calculation application process recommendation result;

[0029] Outputting the recommended APP to a pre-arranged process APP file;

[0030] Associating the recommended APP in the pre-arranged process APP file with the simulation computing APP in the simulation computing field knowledge vector library;

[0031] Generate a simulation calculation application flow chart based on the associated pre-arranged process APP file.

[0032] Furthermore, it also includes:

[0033] Determine the location information of the recommended APP according to the simulation calculation application process recommendation result;

[0034] The accuracy of the recommended APP is verified based on the location information of the recommended APP.

[0035] Furthermore, it also includes:

[0036] Construct a knowledge vector library in the field of simulation computing;

[0037] Constructing the simulation computing application process recommendation large language model according to the simulation computing field knowledge vector library and the large language model;

[0038] The construction of the simulation computing domain knowledge vector library includes:

[0039] Get the simulation computing APP application process data set;

[0040] Convert the simulation computing APP application process data set into text data;

[0041] Extracting structured process data from the text data to obtain structured text data;

[0042] Segmenting the unstructured process data in the text data to obtain segmented text data;

[0043] The structured text data and the segmented text data are used together as the original simulation computing APP application process data;

[0044] Convert the original simulation APP application process data into a simulation calculation APP application process vector;

[0045] The simulation computing domain knowledge vector library is constructed based on the simulation computing APP application process vector.

[0046] As another aspect of the present invention, a simulation computing APP application process recommendation device is provided, which is used to implement the simulation computing APP application process recommendation method described above, which includes:

[0047] A receiving module is used to receive the question information input by the user for constructing the simulation computing APP application process;

[0048] A recommendation module, used for inputting the question information constructed by the simulation computing APP application process into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; wherein, the simulation computing application process recommendation large language model at least includes a large language model and a simulation computing domain knowledge vector library connected to the large language model, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain a vector retrieval result, and the large language model can output the simulation computing application process recommendation result according to the vector retrieval result;

[0049] The association module is used to perform entity association according to the simulation calculation application process recommendation result to obtain a simulation calculation application flow chart, wherein the simulation calculation application flow chart can complete the simulation calculation and obtain the simulation calculation result when it is executed.

[0050] As another aspect of the present invention, a simulation computing APP application process recommendation system is provided, which includes: a simulation computing application process recommendation large language model and the simulation computing APP application process recommendation device mentioned above, and the simulation computing application process recommendation large language model is communicatively connected with the simulation computing APP application process recommendation device.

[0051] The simulation computing APP application process recommendation method provided by the present invention obtains the simulation computing application process recommendation result by inputting the question information of the simulation computing APP application process constructed by the user into the pre-built simulation computing application process recommendation large language model, and can generate a simulation computing application flow chart based on the simulation computing application process recommendation result, thereby achieving the goal of obtaining the simulation computing APP application process without being restricted by professional knowledge, and because the recommendation result can be directly obtained based on the simulation computing application process recommendation large language model, it also has the advantage of high efficiency. In addition, generating a simulation computing application flow chart based on the recommendation result can also directly and efficiently complete the simulation computing. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.

[0053] Figure 1 A flowchart of a method for recommending the application process of a simulation computing APP provided by the present invention.

[0054] Figure 2 A schematic diagram of the specific implementation process of the simulation computing APP application process recommendation method provided by the present invention.

[0055] Figure 3 A flowchart of the specific implementation process of entity association provided by the present invention.

[0056] Figure 4 A schematic diagram of the segmentation strategy provided by the present invention.

[0057] Figure 5 This is a framework diagram of the large language model for intelligent recommendation of simulation computing APP application processes provided by the present invention.

[0058] Figure 6 The present invention provides a data flow diagram for intelligent recommendation of the simulation computing APP application process based on the large language model.

[0059] Figure 7 This is a technical flow chart of intelligent recommendation of simulation computing APP application process based on large language model provided by the present invention.

[0060] Figure 8 This is a schematic diagram of the simulation computing APP application process recommendation interface service provided by the present invention.

[0061] Fig. 9 A schematic diagram of constructing a structured data set for simulation calculation provided by the present invention.

[0062] Fig.10 A schematic diagram of constructing an unstructured data set provided by the present invention.

[0063] Fig.11 This is a flow chart of the LangChain construction method provided by the present invention.

[0064] Fig.12 This is a schematic diagram of the document segmentation process provided by the present invention.

[0065] Fig.13 This is a schematic diagram of the interface after establishing the association provided by the present invention.

[0066] Fig.14 This is a structural block diagram of the simulation computing APP application process recommendation device provided by the present invention. DETAILED DESCRIPTION

[0067] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0068] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0069] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0070] In the traditional model, when using full-text search keyword matching to complete APP search and assist in application process construction, there are often problems such as "cannot find, slow to find, inaccurate to find", which seriously affects the efficiency of APP application process construction.

[0071] Based on this, a simulation computing APP application process recommendation method is provided in this embodiment. Figure 1 is a flowchart of a method for recommending a simulation computing APP application process according to an embodiment of the present invention. Figure 1 As shown, including:

[0072] S100, receiving the question information of the simulation computing APP application process construction input by the user;

[0073] The question information constructed by the simulation calculation APP application process in the embodiment of the present invention may specifically be "How to design a typical propeller strength process", or "I am a simulation software integration application engineer, please help me build an XXX equipment design and performance evaluation application process, including design and modeling, power evaluation, performance evaluation, comprehensive performance evaluation and other steps", etc. The specific question information may be described according to the needs of the user.

[0074] S200, inputting the question information constructed by the simulation computing APP application process into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; wherein, the simulation computing application process recommendation large language model at least includes a large language model and a simulation computing domain knowledge vector library connected to the large language model, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain a vector retrieval result, and the large language model can output the simulation computing application process recommendation result according to the vector retrieval result;

[0075] In an embodiment of the present invention, the above-mentioned question information is input into a pre-built large language model for recommending simulation computing application processes, which includes at least a large language model and a simulation computing domain knowledge vector library. The large language model for recommending simulation computing application processes can output the simulation computing application process civil engineering results for the input question information. Specifically, similarity search can be performed through the simulation computing domain knowledge vector library to obtain vector search results, and the large language model can output the simulation computing application process recommendation results based on the vector search results.

[0076] S300, performing entity association according to the simulation calculation application process recommendation result to obtain a simulation calculation application flow chart, wherein the simulation calculation application flow chart can complete the simulation calculation and obtain the simulation calculation result when it is executed.

[0077] In an embodiment of the present invention, after obtaining the above-mentioned simulation computing application process recommendation results, since the simulation computing application process recommendation results include the simulation computing application process and the recommended APP required in the process, it is possible to establish an association between the recommended APP and the APP in the simulation computing field knowledge vector library to generate a simulation computing application flow chart. Since the simulation computing application flow chart has established an entity association, it can directly call the corresponding APP and execute the corresponding process when it is run to automatically complete the simulation calculation and obtain the simulation results.

[0078] Therefore, the simulation computing APP application process recommendation method provided by the present invention obtains the simulation computing application process recommendation result by inputting the question information of the simulation computing APP application process constructed by the user into the pre-built simulation computing application process recommendation large language model, and can generate a simulation computing application flow chart based on the simulation computing application process recommendation result, thereby achieving the goal of obtaining the simulation computing APP application process without being restricted by professional knowledge, and because the recommendation result can be directly obtained based on the simulation computing application process recommendation large language model, it also has the advantage of high efficiency. In addition, generating a simulation computing application flow chart based on the recommendation result can also directly and efficiently complete the simulation computing.

[0079] A large language model can be pre-built for recommending simulation computing application processes. Specifically, in an embodiment of the present invention, it also includes:

[0080] a1. Build a knowledge vector library in the field of simulation computing;

[0081] a2. Constructing the simulation computing application process recommendation large language model according to the simulation computing domain knowledge vector library and the large language model;

[0082] The construction of the simulation computing domain knowledge vector library includes:

[0083] a11. Obtain the simulation calculation APP application process data set;

[0084] In an embodiment of the present invention, the simulation computing APP application process data set includes at least: simulation computing application process information, simulation computing APP information, and simulation computing physical or virtual test reports.

[0085] The simulation calculation application process information at least includes a simulation calculation process user manual and a simulation calculation process information table;

[0086] The simulation calculation APP information at least includes APP basic information, APP parameter information and APP user manual.

[0087] a12. Convert the simulation calculation APP application process data set into text data;

[0088] In an embodiment of the present invention, relevant texts in simulation calculation files are extracted, and files of different formats are converted into txt format, including file types such as simulation calculation APP user manual, simulation calculation application process description document, simulation calculation APP structured information, simulation calculation task information, etc.

[0089] a13. Extracting structured process data from the text data to obtain structured text data;

[0090] a14, segmenting the unstructured process data in the text data to obtain segmented text data;

[0091] In the embodiment of the present invention, the text loaded by the simulation calculation file loader is segmented into different text blocks, such as Fig.12 shown.

[0092] It should be noted that there are different ways to construct data sets for simulation calculations of structured data and unstructured data.

[0093] The simulation calculation structured data is stored in the database. Therefore, for the simulation calculation structured data, the corresponding attribute values ​​are extracted according to the database attribute names and converted into simulation calculation structured data sets in text form, such as Fig. 9 shown.

[0094] The unstructured datasets for simulation calculations are stored in the file system in different file formats. When loading data in batches, it is necessary to read files of various formats and convert them into text data in text format. The text-based datasets are then combined into unstructured datasets for simulation calculations. Fig.10 shown.

[0095] a15. Use the structured text data and the segmented text data together as the original simulation calculation APP application process data;

[0096] a16. Convert the original simulation APP application process data into a simulation calculation APP application process vector;

[0097] a17. Construct the simulation computing domain knowledge vector library based on the simulation computing APP application process vector.

[0098] It should be understood that the simulation computing field data set serves as the data support for the big language model for recommending the simulation computing APP application process, and is the source of the big language model's thinking and answering. The data set used for the simulation computing APP application process includes simulation computing structured data (stored in the database), simulation computing unstructured data (from the simulation computing application process user manual, the simulation computing APP user manual, the simulation computing physical / virtual test report, and the unstructured information of the specified file path of the simulation computing structured data storage). Among them, the simulation computing APP structured data also includes structured data stored in the database, including the server storage address of the APP user manual and the server storage address of the application process structured file. Among them, the simulation computing application process user manual is the method of using the simulation computing application process and related introduction compiled for developers. It is the detailed information of the simulation computing application process, including the preface (software overview, scope of use), software installation guide (system requirements, software installation), software usage instructions, input and output instructions of each APP, opening the interface, parameter input, submitting calculations, and viewing result information.

[0099] In the embodiment of the present invention, there are different file reading loaders for different types of files. The loader converts data from a specific format or source into a format that can be used by the simulation computing APP application process intelligent recommendation large language model, where the target format is called a document.

[0100] The data set includes simulation computing application process information, simulation computing APP information, simulation computing physical / virtual test reports, etc. Among them, simulation computing application process information includes simulation computing process user manual and simulation computing process information, and simulation computing APP information includes APP basic information, APP parameter information, and APP user manual. The diversity of data sources determines the diversity of loaders used, including CSV file loader, file directory loader, HTML loader, JSON loader, Markdown loader, and PDF loader.

[0101] In an embodiment of the present invention, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain vector search results, including:

[0102] b1. Preprocess the question information constructed by the simulation calculation APP application process to obtain a question information vector;

[0103] b2. performing a similarity search based on the question information vector and obtaining a similarity search result;

[0104] b3. If the similarity search result is greater than or equal to a preset similarity threshold, it is determined that the simulation computing domain knowledge vector library has a corresponding simulation computing application process;

[0105] b4. If the similarity search result is less than the preset similarity threshold, it is determined that the simulation computing domain knowledge vector library does not have a corresponding simulation computing application process.

[0106] It should be understood that, in the embodiment of the present invention, after the data is segmented to obtain different text blocks, the text blocks or the user's query can be converted into vector information to facilitate subsequent similarity analysis.

[0107] When performing vector similarity analysis, we can specifically retrieve text blocks in the knowledge base based on the user's query vector, obtain the most similar text blocks based on the similarity calculation algorithm, and then construct prompt words for the simulation computing APP application process intelligent recommendation large model based on the text blocks matched by vector similarity to complete specific tasks; finally, we use the matched knowledge base related text blocks and the text processing capabilities of the large model to answer the user's query.

[0108] It should be noted that users need to sort out various local files related to simulation calculations before use, and simply clean them after reading. Due to the length limit of the maximum input word block of the large language model, it is necessary to perform segmentation processing according to the actual situation of the simulation calculation text, and store the vectorized simulation calculation data into the local simulation calculation vector database. The principle of langChain+chatGLM is to encode the content of the simulation calculation knowledge base into the vector knowledge base, and then each question asked by the user will be encoded, and the vector correlation algorithm will be used to find the most matching knowledge base fragments. These knowledge base fragments are used as context and submitted to chatGLM as prompt templates together with user questions. The large model of intelligent recommendation of simulation calculation APP application process returns the answer to the simulation calculation application process.

[0109] In an embodiment of the present invention, the large language model can output a simulation computing application process recommendation result according to the vector search result, including:

[0110] When the vector search result indicates that the simulation computing domain knowledge vector library has a corresponding simulation computing application process, a large language model prompt word is generated according to the simulation computing application process and the question information constructed by the simulation computing APP application process;

[0111] The simulation computing application process recommendation result is obtained according to the large language model prompt word.

[0112] It should be understood that when performing knowledge vector retrieval, a vector similarity retrieval is first performed in the knowledge vector library in the simulation computing field. When it is determined that there is a corresponding simulation computing application process in the knowledge vector library in the simulation computing field, the large language model can generate large language model prompt words based on the simulation computing application process and the question information constructed by the simulation computing APP application process, and then output the simulation computing application process recommendation results based on the large language model prompt words.

[0113] In an embodiment of the present invention, the large language model can output a simulation computing application process recommendation result according to the vector search result, including:

[0114] When the vector search result is that the simulation computing field knowledge vector library does not have a corresponding simulation computing application process, the question information constructed by the simulation computing APP application process is input into the intelligent agent to obtain the intelligent agent recommended application process, wherein the intelligent agent can call the classification retrieval algorithm according to the question information constructed by the simulation computing APP application process to obtain the intelligent agent recommended application process;

[0115] Generate a large language model prompt word according to the question information constructed by the agent recommendation application process and the simulation computing APP application process;

[0116] The simulation computing application process recommendation result is obtained according to the large language model prompt word.

[0117] It should be understood that when there is no corresponding simulation computing application process in the knowledge vector library of the simulation computing field, the intelligent agent processes the question information constructed by the simulation computing APP application process to obtain the intelligent agent recommended application process, and then the intelligent agent recommended application process and the question information constructed by the simulation computing APP application process generate a large language model prompt word, and the large language model then obtains the simulation computing application process recommendation result based on the large language model prompt word.

[0118] Specifically, the agent can call a classification retrieval algorithm according to the question information constructed by the simulation computing APP application process to obtain the agent recommended application process, including:

[0119] Determine whether there is an application process execution step in the question information constructed by the simulation calculation APP application process;

[0120] If there is no application process execution step, prompt the user to enter the application process execution step;

[0121] If there are application process execution steps, the classification retrieval algorithm is called according to each application process execution step to obtain the classification retrieval result of each application process execution step;

[0122] Among them, the classification retrieval algorithm can retrieve the simulation computing field knowledge vector library according to the description information of each application process execution step to obtain multiple APPs to be grouped for each application process execution step, and group them according to the application field of each APP to obtain the group of each application process execution step and its corresponding APP.

[0123] In the embodiment of the present invention, the relevant contents in the simulation computing domain knowledge vector library are retrieved based on the user's question; the retrieval results from the simulation computing domain knowledge vector library are processed in two cases, such as Figure 2 As shown:

[0124] (1) The first case is: if the existing simulation computing application process can be directly obtained from the vector knowledge base in the simulation computing field, it will be used as the vector prompt word Prompt input, and the large language model is recommended for polishing through the simulation computing application process.

[0125] For example, if a user asks "How to design a typical propeller strength process", the big model will retrieve the corresponding recommended application process based on the user's question, and input it into the big model as a prompt word instruction. The big model replies to the question according to the above constructed instructions, and its answer may be "According to the known information, the software contains 5 application modules, namely: XXX calculation, YYY forecast, ZZZ forecast, TTT forecast, FFF component quick selection, SSS forecast." The big model's reply statement describes in detail the APP included in the simulation calculation process. The user can select and edit the name of the simulation calculation APP based on the reply content so that the simulation calculation APP application process meets the needs.

[0126] (2) The second situation is: if it is impossible to directly obtain the ready-made simulation computing application process from the vector knowledge base in the simulation computing field, the answer will be given by the intelligent agent.

[0127] When the user asks a question, the agent is provided with the description information of the previous and next steps of the process composition, and groups and APPs are generated. An example of a user's question is as follows: "I am a simulation software integrated application engineer. Please help me build an XXX equipment design and performance evaluation application process, including design and modeling, power evaluation, performance evaluation, comprehensive performance evaluation and other steps." The trained agent can call the classification retrieval algorithm (get_flows algorithm) according to the question description, retrieve APPs from the vector knowledge base in the simulation calculation field, and then group them according to the APPs and hang the groups under the steps. For example: Under the "Design and Modeling" step, there are three groups: "Theoretical Design", "Strength Design", and "3D Modeling", and under "Theoretical Design" there are two APPs: "QT Test APP" and "DAS".

[0128] In the embodiment of the present invention, the classification retrieval algorithm, namely the get_flows algorithm, is described as follows:

[0129] 1) The simulation computing domain vector knowledge base stores detailed information about the APP, including its functions, application fields, and other information that can be used to distinguish the steps of the application process.

[0130] 2) According to the description information of each step, the vector knowledge base of the simulation computing field is retrieved, and multiple APPs to be grouped are retrieved in each step.

[0131] 3) Group apps according to their application areas.

[0132] 4) The grouped results are returned to the big model, and the language is reorganized and displayed to the user in a friendly manner.

[0133] In the embodiment of the present invention, entity association is performed according to the simulation computing application process recommendation result to obtain a simulation computing application flow chart, such as Figure 3 As shown, including:

[0134] S310, determining a recommended APP according to the simulation calculation application process recommendation result;

[0135] S320, outputting the recommended APP to a pre-arranged process APP file;

[0136] S330, associating the recommended APP in the pre-arranged process APP file with the simulation computing APP in the simulation computing field knowledge vector library;

[0137] S340, generating a simulation calculation application flow chart according to the associated pre-arranged process APP file.

[0138] In the embodiment of the present invention, based on the generated recommended process APP, it is output to the pre-arranged process APP table, and the recommended process APP can be associated with the APP in the simulation computing APP library. The table supports editing. If there is an undetected simulation computing APP, the associated simulation computing APP can be selected according to the name of the undetected simulation computing APP. The recommended simulation computing APP supports multiple selections. When all simulation computing APPs exist in the simulation computing APP library, a simulation computing application process can be directly generated for simulation computing tasks. Fig.13 The following is a schematic diagram of the interface after the association is established.

[0139] In an embodiment of the present invention, it also includes:

[0140] Determine whether there is a simulation computing APP corresponding to the recommended APP in the simulation computing field vector library;

[0141] If there is no simulation computing APP corresponding to the recommended APP in the simulation computing domain vector library, at least one simulation computing APP to be selected that is associated with the recommended APP is selected in the simulation computing domain vector library for association.

[0142] It should be understood that even if there is no simulation computing APP corresponding to the recommended APP in the simulation computing field vector library, a simulation computing APP with a relatively high similarity to the recommended APP can be selected from the simulation computing field vector library as a candidate simulation computing APP. For example, the similarity between the simulation computing APP in the current simulation computing vector library and the recommended APP is calculated, and they are sorted from high to low according to the similarity. Therefore, the candidate simulation computing APP for the seat with the highest similarity can be associated with the recommended APP.

[0143] In the embodiment of the present invention, the recommended APP can also be traced.

[0144] Specifically, the simulation computing APP application process recommendation method of the present invention also includes:

[0145] Determine the location information of the recommended APP according to the simulation calculation application process recommendation result;

[0146] The accuracy of the recommended APP is verified based on the location information of the recommended APP.

[0147] It should be understood that when the response result of the simulation calculation APP application process intelligent recommendation model is returned, the location of the answer source will also be returned synchronously. Users can open the source to check whether the answer of the intelligent recommendation model is true and reliable. The name of the extracted APP is returned and manually confirmed. Users can manually open the source document and conduct online verification based on expert experience to ensure that the recommended APP is reliable and feasible.

[0148] In the embodiment of the present invention, when recommending the construction of a large language model for a simulation computing application process, it is also necessary to set up an environment.

[0149] Specifically, in order to build a large language model for simulation computing APP application process recommendation, you need to install software such as Text2Vec, ChatGLM, Faiss, and Langchain. Among them:

[0150] Text2Vec converts text data into vector form so that the simulation computing APP application process intelligent recommendation model can process and analyze simulation computing text information. Its goal is to embed simulation computing text information into a vector space, so that various simulation computing text analysis tasks can be performed, such as text classification, sentiment analysis, topic modeling, and similarity calculation. After mapping the words in the text to a high-dimensional vector space, the model can better understand and process text data, and at the same time can capture the semantic relationship between words, and the model can better understand the meaning of the text;

[0151] ChatGLM generates the subsequent simulation calculation text content based on the input simulation calculation prompt context, and the generated text content is more in line with human natural language habits;

[0152] The Faiss database is used to process efficient similarity search and vector retrieval of large-scale simulation computing vector data sets, such as images, text, audio, etc. The database's high-performance vector indexing and similarity search operations can be used to build applications in the field of simulation computing. It uses its various vector index structures to accelerate vector similarity searches. When it comes to large-scale text data sets, it uses multi-threading and GPU acceleration to simplify the complexity of similarity searches.

[0153] like Fig.11 As shown in the figure, LangChain creates a large language model for intelligent recommendation of simulation computing APP application processes, which is used to handle various questions and answers and interactions related to simulation computing APP application processes. It helps developers use language models to build end-to-end applications, and also integrates additional resources to provide interfaces and databases to the outside world.

[0154] The dataset text is first preprocessed by breaking it down into smaller parts or summaries, and then the summaries are embedded into a vector space. It provides a rich set of tools and libraries to facilitate processing text data, training models, generating dialogues, and more.

[0155] After the environment is set up, a general large language model is constructed. Specifically, the embodiment of the present invention uses the ChatGLM2-6B model to provide appropriate responses and support for user questions and requirements. This model has been trained for dialogue, can cope with continuous dialogue scenarios, and further optimizes video memory usage. The model uses the hybrid objective function of GLM, and has been pre-trained with 1.4T Chinese and English identifiers and trained with human preference alignment. Compared with the first-generation model, ChatGLM2-6B has achieved significant performance improvements on data sets such as MMLU (+23%), CEval (+33%), GSM8K (+571%), and BBH (+60%), and is highly competitive among open source models of the same size.

[0156] Finally, the simulation domain knowledge base is constructed. Specifically, after the simulation data is loaded, the simulation data set is preprocessed and stored in the simulation vector library. Since the amount of external simulation data is relatively large, such as PDF and text documents, the simulation external data documents need to be divided into blocks before they can be stored in the simulation vector database.

[0157] After collecting enough simulation calculation documents, they are loaded and cleaned, and the processed documents that are easy for vector matching are stored in the vector library.

[0158] RecursiveCharacterTextSplitter is used to split the unstructured documents and structured data in the simulation computing field into different modules for the subsequent construction of the knowledge vector library in the simulation computing field. This segmentation method splits the text by string, recursively tries to split the text by different delimiters, and takes into account the length and overlapping characters of the segmented text. Figure 4 This reflects the data set segmentation strategy adopted by the present invention in different simulation calculation data sets.

[0159] The RecursiveCharacterTextSplitter segmentation strategy is used for unstructured data such as simulation computing APP user manuals, simulation computing application process user manuals, and simulation computing physical / virtual test reports. Since structured data such as APP structured data and application process structured data already have sufficient information, the segmentation method is not used for processing. The extracted structured text is directly used as the original data for the subsequent knowledge vector library construction.

[0160] According to different application scenarios, the Chinese-BERT-large model is used to establish three sub-vector index libraries for the simulation computing domain knowledge vector library. The model is used to convert the segmented unstructured data and structured data into vectors and store them in the FAISS vector library. They are:

[0161] 1) Vector index library of APP information in the field of simulation computing: contains APP information in the field of simulation computing.

[0162] 2) Vector index library of application process information in the field of simulation computing: contains application process information in the field of simulation computing.

[0163] 3) Simulation and computational physics\virtual experiment report vector index library: contains simulation and computational physics\virtual experiment report information.

[0164] The above three types of vector index libraries are used to implement document vector indexing and similarity matching.

[0165] Before using the Faiss vector database to perform similarity search on the question query vector, the original vector set is encapsulated into an index file and cached in memory for real-time query calculations in simulation. When the index file is built for the first time, two processes are required: training and adding. If new vectors need to be added to the index file later, there can also be an addition operation to achieve incremental indexing. When retrieving vector indexes, the indexFlatL2 index type retrieval method is used to perform a brute force 2-norm distance search on the vector.

[0166] The construction of a large language model for intelligent recommendation of APP application process simulation Figure 5 shown.

[0167] 1) The user describes the application process of the simulation computing APP that needs to be built as a query statement for the knowledge vector library in the simulation computing field;

[0168] 2) Based on the knowledge in the field of simulation computing, construct a knowledge vector library in the field of simulation computing;

[0169] 3) Query similar simulation computing domain knowledge based on user query statements;

[0170] 4) Use highly similar simulation domain knowledge and user input to construct instructions;

[0171] 5) Use the instructions as the input of the large model of intelligent recommendation of the simulation computing APP application process, and return the question and answer results to form the actual content of the intelligent recommendation.

[0172] The simulation computing APP application process recommendation is based on the current large-scale pre-trained model. It uses the accuracy of the large-scale data pre-trained model in document question-answering tasks to realize the intelligent recommendation business process of the simulation computing APP application process. Figure 6 and Figure 7 As shown, the specific construction process includes the following steps:

[0173] (1) Upload the user manual of the simulation calculation APP application process;

[0174] The simulation computing APP application process user manual contains an introduction and detailed information about the simulation computing APP application process, as well as the association information between the APPs in the process. The detailed information of the simulation computing APP application process serves as the main reference information for building recommended business processes.

[0175] (2) Convert the user manual of the simulation calculation APP application process into text data in txt format;

[0176] The user manual of the simulation calculation APP application process is based on the data loader, which reads the contents in the table into the text content of txt. The text content after reading is in the form of "Number module 1 1-XXX calculation 2 2-YYY forecast 3 3-ZZZ forecast 4 4-TTT forecast 5 5-FFF component quick selection 6 6-SSS forecast". This content will be used as an important content for the subsequent recommendation process generation.

[0177] (3) Extract structured process information;

[0178] The APP information library and the simulation computing APP application process library contain some process data stored in the database. For example, "desc" contains the APP name contained in the simulation computing APP application process. Therefore, a regular expression-based method is used to extract the simulation computing APP name information contained in the structured process information.

[0179] (4) Segment text data into multiple fixed-length text segments;

[0180] The RecursiveCharacterTextSplitter method is used to split the unstructured process data, and the split text data and structured text data are used as the original simulation calculation APP application process data.

[0181] (5) Use the vector model to convert text into vectors;

[0182] Use the Chinese-BERT-Large model to convert the raw process data into vectors.

[0183] (6) Construct a knowledge vector library in the field of simulation computing;

[0184] Use simulation calculation APP application process vectors to build a knowledge vector library.

[0185] (7) Retrieve relevant content from the knowledge vector library in the simulation computing field based on the user’s questions.

[0186] The specific implementation process of retrieving relevant content in the knowledge vector library in the simulation computing field based on user questions can be seen in the description of the two retrieval results mentioned above.

[0187] The above is a specific implementation process of the simulation computing APP application process recommendation method of the present invention.

[0188] In addition, after obtaining user input, the system retrieves relevant simulation domain knowledge through searching the vector knowledge base in the field of simulation computing, forms instructions based on the user's questions and answers, initiates instruction requests to the big model that intelligently recommends the simulation computing APP application process, outputs the answers to the questions, and returns them to the user, making it convenient for the user to edit the application process based on the answers of the big model.

[0189] As another embodiment of the present invention, a simulation computing APP application process recommendation device is provided, which is used to implement the simulation computing APP application process recommendation method described above, wherein, Fig.14 As shown, including:

[0190] The receiving module 100 is used to receive the question information input by the user for constructing the simulation computing APP application process;

[0191] The recommendation module 200 is used to input the question information constructed by the simulation computing APP application process into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; wherein, the simulation computing application process recommendation large language model at least includes a large language model and a simulation computing domain knowledge vector library connected to the large language model, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain a vector retrieval result, and the large language model can output the simulation computing application process recommendation result according to the vector retrieval result;

[0192] The association module 300 is used to perform entity association according to the simulation calculation application process recommendation result to obtain a simulation calculation application flow chart, wherein the simulation calculation application flow chart can complete the simulation calculation and obtain the simulation calculation result when it is executed.

[0193] The simulation computing APP application process recommendation device provided by the present invention obtains the simulation computing application process recommendation result by inputting the question information of the simulation computing APP application process constructed by the user into the pre-built simulation computing application process recommendation large language model, and can generate a simulation computing application flow chart based on the simulation computing application process recommendation result, thereby achieving the goal of obtaining the simulation computing APP application process without being restricted by professional knowledge, and because the recommendation result can be directly obtained based on the simulation computing application process recommendation large language model, it also has the advantage of high efficiency. In addition, generating a simulation computing application flow chart based on the recommendation result can also directly and efficiently complete the simulation computing.

[0194] Regarding the specific implementation process of the simulation computing APP application process recommendation device provided by the present invention, reference can be made to the description of the simulation computing APP application process recommendation method in the previous text, which will not be repeated here.

[0195] As another embodiment of the present invention, a simulation computing APP application process recommendation system is provided, which includes: a simulation computing application process recommendation large language model and the simulation computing APP application process recommendation device mentioned above, and the simulation computing application process recommendation large language model is communicatively connected with the simulation computing APP application process recommendation device.

[0196] like Figure 2 and Figure 8 As shown in the figure, the application process recommendation is implemented based on the LangChain+LLM+embedding principle for the simulation computing domain knowledge. First, the simulation computing domain knowledge is loaded into documents that can be used for simulation computing through different loaders, and then the text is converted into vectors through the Text2vec method and stored in the vector library. When there is external interaction with the large language model, the simulation computing text content is generated through ChatGLM, and the large language model LangChain returns the interaction data. Finally, based on the returned results, an APP about the simulation computing APP application process is generated, and a flowchart related to the application process is generated to help users find the implementation method of the simulation computing task application process.

[0197] When returning the reply result of the simulation computing APP application process intelligent recommendation large model, the location of the answer source will also be returned synchronously. Users can open the source to check whether the answer of the intelligent recommendation large model is true and reliable. The name of the extracted APP is returned for manual confirmation. The returned results are automatically constructed into the recommended APP table. The table supports addition, deletion and modification. If there are undetected simulation computing APPs, the related simulation computing APPs can be selected according to the names of the undetected simulation computing APPs. The recommended simulation computing APPs support multiple selection. When all simulation computing APPs exist in the simulation computing APP library, the simulation computing application process can be directly generated for simulation computing tasks.

[0198] The recommended interfaces for the simulation computing APP application process may include a knowledge base question and answer interface, a page data saving interface, a page data acquisition interface, a simulation computing process publishing interface, a data determination interface for determining whether there is a published data interface, a data search simulation computing APP interface, a simulation computing APP interface for checking whether the simulation computing APP is a correct interface, and a template data acquisition interface.

[0199] based on Figure 2As shown, the simulation computing APP application process recommendation system in the embodiment of the present invention, the simulation computing APP application process recommendation device included therein can be formed as a simulation computing application common model process recommendation service application when specifically implemented, and the simulation computing APP application process recommendation system can also include a simulation computing application process can only recommend a large language model construction module, the module is used to provide simulation computing process knowledge, so as to facilitate the construction of a simulation computing application process recommendation large language model.

[0200] Regarding the specific implementation process of the simulation computing APP application process recommendation system provided by the present invention, reference can be made to the description of the simulation computing APP application process recommendation method in the previous text, which will not be repeated here.

[0201] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for recommending application flow of simulation computing APP, characterized in that: include: Receive the question information entered by the user for constructing the simulation computing APP application process; Input the question information constructed by the simulation computing APP application process into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; wherein the simulation computing application process recommendation large language model at least includes a large language model and a simulation computing domain knowledge vector library connected to the large language model, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain a vector retrieval result, and the large language model can output the simulation computing application process recommendation result according to the vector retrieval result; Perform entity association according to the simulation calculation application process recommendation result to obtain a simulation calculation application flow chart, wherein the simulation calculation application flow chart can complete the simulation calculation and obtain the simulation calculation result when it is executed; Among them, it also includes: Construct a knowledge vector library in the field of simulation computing; Constructing the simulation computing application process recommendation large language model according to the simulation computing field knowledge vector library and the large language model; The construction of the simulation computing domain knowledge vector library includes: Get the simulation computing APP application process data set; Convert the simulation computing APP application process data set into text data; Extracting structured process data from the text data to obtain structured text data; Segmenting the unstructured process data in the text data to obtain segmented text data; The structured text data and the segmented text data are used together as the original simulation computing APP application process data; Convert the original simulation APP application process data into a simulation calculation APP application process vector; Construct the simulation computing domain knowledge vector library according to the simulation computing APP application process vector; Among them, the simulation computing application process recommendation large language model can perform segmented processing according to the actual situation of the simulation computing text, and store the vectorized simulation computing data into the local simulation computing vector database. Through the principle of langChain+chatGLM, the simulation computing knowledge base content is encoded and stored in the vector knowledge base. Then, every question asked by the user will be encoded, and the vector correlation algorithm will be used to find the most matching knowledge base fragments. These knowledge base fragments are used as context and submitted to chatGLM as prompt templates together with the user's questions. The simulation computing APP application process intelligent recommendation large model returns the answer to the simulation computing application process.

2. The method for recommending application flow of simulation computing APP according to claim 1, characterized in that: The simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain vector search results, including: Preprocessing the question information constructed by the simulation computing APP application process to obtain a question information vector; Performing a similarity search based on the question information vector and obtaining a similarity search result; If the similarity search result is greater than or equal to a preset similarity threshold, it is determined that the simulation computing domain knowledge vector library has a corresponding simulation computing application process; If the similarity search result is less than the preset similarity threshold, it is determined that the simulation computing domain knowledge vector library does not have a corresponding simulation computing application process.

3. The method for recommending application flow of simulation computing APP according to claim 2, characterized in that: The large language model can output simulation computing application process recommendation results according to the vector search results, including: When the vector search result indicates that the simulation computing domain knowledge vector library has a corresponding simulation computing application process, a large language model prompt word is generated according to the simulation computing application process and the question information constructed by the simulation computing APP application process; The simulation computing application process recommendation result is obtained according to the large language model prompt word.

4. The method for recommending application flow of simulation computing APP according to claim 2, characterized in that: The large language model can output simulation computing application process recommendation results according to the vector search results, including: When the vector search result is that the simulation computing field knowledge vector library does not have a corresponding simulation computing application process, the question information constructed by the simulation computing APP application process is input into the intelligent agent to obtain the intelligent agent recommended application process, wherein the intelligent agent can call the classification retrieval algorithm according to the question information constructed by the simulation computing APP application process to obtain the intelligent agent recommended application process; Generate a large language model prompt word according to the question information constructed by the agent recommendation application process and the simulation computing APP application process; The simulation computing application process recommendation result is obtained according to the large language model prompt word.

5. The method for recommending application flow of simulation computing APP according to claim 4 is characterized in that: The intelligent agent can call a classification retrieval algorithm according to the question information constructed by the simulation computing APP application process to obtain the intelligent agent recommended application process, including: Determine whether there is an application process execution step in the question information constructed by the simulation calculation APP application process; If there is no application process execution step, prompt the user to enter the application process execution step; If there are application process execution steps, the classification retrieval algorithm is called according to each application process execution step to obtain the classification retrieval result of each application process execution step; Among them, the classification retrieval algorithm can retrieve the simulation computing field knowledge vector library according to the description information of each application process execution step to obtain multiple APPs to be grouped for each application process execution step, and group them according to the application field of each APP to obtain the group of each application process execution step and its corresponding APP.

6. The method for recommending application flow of simulation computing APP according to any one of claims 1 to 5, characterized in that: Performing entity association according to the simulation computing application process recommendation result to obtain a simulation computing application flow chart includes: Determine the recommended APP according to the simulation calculation application process recommendation result; Outputting the recommended APP to a pre-arranged process APP file; Associating the recommended APP in the pre-arranged process APP file with the simulation computing APP in the simulation computing field knowledge vector library; Generate a simulation calculation application flow chart based on the associated pre-arranged process APP file.

7. The method for recommending application flow of simulation computing APP according to any one of claims 1 to 5, characterized in that: Also includes: Determine the location information of the recommended APP according to the simulation calculation application process recommendation result; The accuracy of the recommended APP is verified based on the location information of the recommended APP.

8. A simulation computing APP application process recommendation device, used to implement the simulation computing APP application process recommendation method according to any one of claims 1 to 7, characterized in that: include: A receiving module is used to receive the question information input by the user for constructing the simulation computing APP application process; A recommendation module, used for inputting the question information constructed by the simulation computing APP application process into the simulation computing application process recommendation large language model to obtain the simulation computing application process recommendation result; wherein, the simulation computing application process recommendation large language model at least includes a large language model and a simulation computing domain knowledge vector library connected to the large language model, the simulation computing domain knowledge vector library can retrieve the input question information constructed by the simulation computing APP application process to obtain a vector retrieval result, and the large language model can output the simulation computing application process recommendation result according to the vector retrieval result; The association module is used to perform entity association according to the simulation calculation application process recommendation result to obtain a simulation calculation application flow chart, wherein the simulation calculation application flow chart can complete the simulation calculation and obtain the simulation calculation result when it is executed.

9. A simulation computing APP application process recommendation system, characterized in that: include: A large language model for recommending a simulation computing application process and the simulation computing APP application process recommendation device described in claim 8, wherein the large language model for recommending a simulation computing application process is communicatively connected to the simulation computing APP application process recommendation device.

Images