Interaction flow chart processing method and device, storage medium and electronic equipment

Abstract

The invention discloses an interactive flow chart processing method and device, a storage medium and electronic equipment, and relates to the technical field of flow chart generation. The method comprises the following steps: receiving target information input by a target object; inputting target information input by the target object into the knowledge retrieval model for processing, and retrieving to obtain flow information related to the target information; determining prompt information of the target generation model based on the flow information related to the target information, inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting an initial interaction flow chart; and performing integrity verification on the initial interaction flow chart to obtain a verification result, and determining a target interaction flow chart according to the verification result. Through the method and the device, the problem that the effect of generating the interaction flow chart is poor due to the fact that the interaction flow chart related to the financial business depends on manual configuration in the related technology is solved.

Description

Technical Field

[0001] The present application relates to the technical field of flowchart generation, and in particular to a method and device for processing an interactive flowchart, a storage medium, and an electronic device. Background Art

[0002] As the capabilities of intelligent models gradually develop, people pay more attention to accurate and effective interactive experiences. Moreover, the intelligent interactive capabilities in related technologies usually rely on manual configuration, presetting the model interaction words and interaction processes in advance, and then generating interactive flowcharts. However, the method of generating interactive flowcharts in related technologies is affected by factors such as insufficient human understanding of the business or lack of configuration experience, and it takes a lot of time to approve and configure the interactive flowchart, which will result in low accuracy and efficiency of the generated interactive flowchart, and the difficulty of configuring the flowchart is also greater, which in turn leads to poor results in generating interactive flowcharts.

[0003] Currently, no effective solution has been proposed to address the problem that the interaction flowcharts related to financial services rely on manual configuration in related technologies, resulting in poor results in generating interaction flowcharts. Summary of the invention

[0004] The main purpose of the present application is to provide a method and device for processing an interactive flowchart, a storage medium and an electronic device, so as to solve the problem in the related art that interactive flowcharts related to financial services rely on manual configuration, resulting in poor results in generating interactive flowcharts.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for processing an interactive flowchart is provided. The method comprises: receiving target information input by a target object, wherein the target information is used to describe information of an interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; inputting the target information input by the target object into a knowledge retrieval model for processing, and retrieving process information related to the target information; based on the process information related to the target information, determining prompt information of a target generation model, and inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting an initial interactive flowchart, wherein the target generation model is a model obtained by adjusting the original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interactive flowchart; performing integrity check on the initial interactive flowchart to obtain a check result, and determining the target interactive flowchart based on the check result.

[0006] Furthermore, the target information input by the target object is input into the knowledge retrieval model for processing, and the process information related to the target information is retrieved, including: vectorizing the target information input by the target object through the knowledge retrieval model to obtain vector information corresponding to the target information; based on the vector information corresponding to the target information, retrieving vector data related to the vector information corresponding to the target information in a vector database, wherein the vector database is used to store multiple vector data, each vector data is vector data obtained by vectorizing the information used to describe each historical flowchart; based on the vector data related to the vector information corresponding to the target information, the process information related to the target information is obtained.

[0007] Furthermore, based on the process information related to the target information, determining the prompt information of the target generation model includes: based on the process information related to the target information, obtaining the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information; splicing the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information to obtain the prompt information of the target generation model.

[0008] Furthermore, the target information input by the target object and the prompt information of the target generation model are input into the target generation model for processing, and outputting the initial interaction flowchart includes: inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interaction flowchart to be constructed; based on the text information, determining the information set of the interaction flowchart to be constructed, wherein the information set includes at least the attribute information of each node of the interaction flowchart to be constructed and the attribute information of the connection line between every two nodes of the interaction flowchart to be constructed; rendering the information in the information set of the interaction flowchart to be constructed to obtain the initial interaction flowchart.

[0009] Furthermore, the target generation model is obtained in the following manner: obtaining multiple voice interaction data in a vertical field from a target database, and converting each voice interaction data to obtain multiple question-and-answer texts, wherein the vertical field includes at least the financial field; obtaining multiple target texts, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about the model; constructing a training data set based on the multiple question-and-answer texts and the multiple target texts; and obtaining the target generation model based on the training data set.

[0010] Furthermore, based on the training data set, obtaining the target generation model includes: based on the training data set, adjusting the original generation model to obtain an adjusted model; quantizing the adjusted model to obtain a processed model; and accelerating inference on the processed model to obtain the target generation model.

[0011] Furthermore, according to the verification result, determining the target interaction flowchart includes: if the verification result indicates that the initial interaction flowchart is a complete flowchart, using the initial interaction flowchart as the target interaction flowchart; if the verification result indicates that the initial interaction flowchart is an incomplete flowchart, determining abnormal information existing in the initial interaction flowchart; obtaining a description text for describing the abnormal information in the initial interaction flowchart; inputting the description text into the target generation model, performing correction processing on the initial interaction flowchart, and outputting the target interaction flowchart.

[0012] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a device for processing an interactive flowchart is provided. The device includes: a first receiving unit, used to receive target information input by a target object, wherein the target information is used to describe information of an interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; a first input unit, used to input the target information input by the target object into a knowledge retrieval model for processing, and retrieve the process information related to the target information; a first processing unit, used to determine the prompt information of the target generation model based on the process information related to the target information, and input the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and output an initial interactive flowchart, wherein the target generation model is a model obtained by adjusting the original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interactive flowchart; a second processing unit, used to perform integrity verification on the initial interactive flowchart, obtain a verification result, and determine the target interactive flowchart according to the verification result.

[0013] Furthermore, the first input unit includes: a first processing module, used to vectorize the target information input by the target object through the knowledge retrieval model to obtain vector information corresponding to the target information; a first retrieval module, used to retrieve vector data related to the vector information corresponding to the target information in a vector database based on the vector information corresponding to the target information, wherein the vector database is used to store multiple vector data, each vector data is vector data obtained by vectorizing the information used to describe each historical flowchart; a first determination module, used to obtain process information related to the target information based on the vector data related to the vector information corresponding to the target information.

[0014] Furthermore, the first processing unit includes: a first acquisition module, used to acquire the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information based on the process information related to the target information; a first splicing module, used to splice the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information to obtain the prompt information of the target generation model.

[0015] Furthermore, the first processing unit includes: a second processing module, used to input the target information input by the target object and the prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interactive flowchart to be constructed; a second determination module, used to determine the information set of the interactive flowchart to be constructed based on the text information, wherein the information set includes at least the attribute information of each node of the interactive flowchart to be constructed and the attribute information of the connecting line between every two nodes of the interactive flowchart to be constructed; a third processing module, used to render the information in the information set of the interactive flowchart to be constructed to obtain the initial interactive flowchart.

[0016] Furthermore, the target generation model is obtained through the following units: a third processing unit, used to obtain multiple voice interaction data in a vertical field from a target database, and convert each voice interaction data to obtain multiple question and answer texts, wherein the vertical field includes at least the financial field; a first acquisition unit, used to obtain multiple target texts, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about models; a first construction unit, used to construct a training data set based on the multiple question and answer texts and the multiple target texts; a first determination unit, used to obtain the target generation model based on the training data set.

[0017] Furthermore, the first determination unit includes: a first adjustment module, used to adjust the original generation model based on the training data set to obtain an adjusted model; a fourth processing module, used to quantize the adjusted model to obtain a processed model; a first acceleration module, used to accelerate the inference of the processed model to obtain the target generation model.

[0018] Furthermore, the second processing unit includes: a third determination module, used to use the initial interaction flowchart as the target interaction flowchart if the verification result indicates that the initial interaction flowchart is a complete flowchart; a fourth determination module, used to determine the abnormal information existing in the initial interaction flowchart if the verification result indicates that the initial interaction flowchart is an incomplete flowchart; a second acquisition module, used to acquire a description text for describing the abnormal information in the initial interaction flowchart; and a fifth processing module, used to input the description text into the target generation model, perform correction processing on the initial interaction flowchart, and output the target interaction flowchart.

[0019] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer-readable storage medium is provided, wherein the storage medium stores a program, wherein the program executes any one of the above-mentioned methods for processing an interactive flowchart.

[0020] In order to achieve the above-mentioned purpose, according to another aspect of the present application, an electronic device is provided, which includes one or more processors and a memory, and the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the processing method of the interactive flowchart described in any one of the above-mentioned items.

[0021] Through the present application, the following steps are adopted: receiving target information input by a target object, wherein the target information is used to describe information of an interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; inputting the target information input by the target object into a knowledge retrieval model for processing, and retrieving process information related to the target information; based on the process information related to the target information, determining prompt information of a target generation model, and inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting an initial interactive flowchart, wherein the target generation model is a model obtained by adjusting an original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interactive flowchart; performing an integrity check on the initial interactive flowchart to obtain a check result, and determining a target interactive flowchart based on the check result, thereby solving the problem in the related art of relying on manual configuration of interactive flowcharts related to financial business, resulting in poor effect of generating interactive flowcharts. By using the knowledge retrieval model to retrieve the corresponding process information according to the user's description content, and then determining the prompt words of the large model (i.e., the target generation model) based on the retrieved content, and inputting the prompt words and user questions into the fine-tuned large model (i.e., the target generation model), the initial interaction flowchart is output, and the fine-tuned large model is a model obtained by vertically fine-tuning the general large model. The initial interaction flowchart is then checked for integrity to obtain the target interaction flowchart, thereby eliminating the need to rely on manual configuration of interaction flowcharts related to financial services, and improving the quality of intelligent interaction expression information, strengthening the process canvas generation capability within the large model field, reducing the difficulty of configuring complex processes, and thereby improving the effect of generating interaction flowcharts. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 It is a flowchart of a method for processing an interactive flowchart provided in an embodiment of the present application;

[0024] Figure 2 is a flowchart of an optional method for processing an interactive flowchart provided in an embodiment of the present application;

[0025] Figure 3 is a flowchart of a large model generation interaction flowchart based on RAG retrieval enhancement capability in this embodiment;

[0026] Figure 4 is a schematic diagram of a processing device for an interaction flow chart provided in an embodiment of the present application;

[0027] Figure 5 is a schematic diagram of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0029] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising 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.

[0031] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. For example, an interface is set between this system and the relevant user or organization. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or organization.

[0032] The present invention is described below in conjunction with preferred implementation steps. Figure 1 is a flowchart of a method for processing an interactive flowchart according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0033] Step S101, receiving target information input by a target object, wherein the target information is used to describe information of an interaction flowchart to be constructed, and the interaction flowchart to be constructed is an interaction flowchart related to financial services.

[0034] For example, the flowchart description information (corresponding to the above target information) input by the user (corresponding to the above target object) is obtained, and the flowchart description information (corresponding to the above target information) can be used to describe the interactive flowchart to be constructed.

[0035] Step S102: input the target information input by the target object into the knowledge retrieval model for processing, and retrieve the process information related to the target information.

[0036] For example, according to the user description content (corresponding to the above-mentioned target information), the corresponding process information is retrieved using the knowledge retrieval model.

[0037] Step S103, based on the process information related to the target information, determine the prompt information of the target generation model, and input the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and output the initial interaction flowchart, wherein the target generation model is a model obtained by adjusting the original generation model based on the data of the vertical field, the data of the vertical field is data related to the financial field, and the original generation model is used to generate the interaction flowchart.

[0038] For example, based on the corresponding process information retrieved, the prompt words of the large model (corresponding to the above-mentioned target generation model) are determined, and the prompt words and user questions are input into the fine-tuned large model (corresponding to the above-mentioned target generation model), and then the initial interaction flowchart is output.

[0039] Step S104, performing integrity check on the initial interaction flowchart to obtain a check result, and determining a target interaction flowchart based on the check result.

[0040] For example, the integrity of the graph structure (corresponding to the above-mentioned initial interaction flowchart) is checked to determine whether there is any missing or redundant connection information, and then based on the integrity of the graph structure (corresponding to the above-mentioned initial interaction flowchart), the final interaction flowchart (corresponding to the above-mentioned target interaction flowchart) is obtained.

[0041] Through the above steps S101 to S104, the corresponding process information is retrieved by using the knowledge retrieval model according to the user description content, and then the prompt words of the large model (i.e., the target generation model) are determined based on the retrieved content, and the prompt words and user questions are input into the fine-tuned large model (i.e., the target generation model), and the initial interaction flowchart is output. The fine-tuned large model is a model obtained by vertically fine-tuning the general large model. The initial interaction flowchart is then checked for integrity to obtain the target interaction flowchart, thereby eliminating the need to rely on manual configuration of interaction flowcharts related to financial services, and improving the quality of intelligent interaction expression information, enhancing the process canvas generation capability within the large model field, reducing the difficulty of configuring complex processes, and thereby improving the effect of generating interaction flowcharts.

[0042] Optionally, in the interactive flowchart processing method provided in the embodiment of the present application, the target information input by the target object is input into the knowledge retrieval model for processing, and the process information related to the target information is retrieved, including: vectorizing the target information input by the target object through the knowledge retrieval model to obtain vector information corresponding to the target information; based on the vector information corresponding to the target information, retrieving vector data related to the vector information corresponding to the target information from a vector database, wherein the vector database is used to store multiple vector data, each vector data being vector data obtained by vectorizing the information used to describe each historical flowchart; based on the vector data related to the vector information corresponding to the target information, obtaining the process information related to the target information.

[0043] For example, a process description text (corresponding to the target information mentioned above) is generated according to user input needs, the process description text (corresponding to the target information mentioned above) is embedded and vectorized through the above knowledge retrieval model, and the corresponding knowledge content is retrieved in the vector retrieval knowledge base through vector retrieval, thereby realizing the retrieval of historical related process knowledge according to the user input topic or flowchart description.

[0044] Through the above solution, the process information related to the user description text can be obtained quickly and accurately.

[0045] Optionally, in the interactive flowchart processing method provided in the embodiment of the present application, based on the process information related to the target information, determining the prompt information of the target generation model includes: based on the process information related to the target information, obtaining the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information; splicing the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information to obtain the prompt information of the target generation model.

[0046] For example, based on the retrieved knowledge base content, the flowchart description and the flowchart structure are extracted, and the flowchart description and the flowchart structure are spliced ​​to obtain the prompt words of the large model (corresponding to the above-mentioned prompt information).

[0047] Through the above solution, the prompt information of the large model can be obtained quickly and accurately.

[0048] Optionally, in the method for processing an interactive flowchart provided in an embodiment of the present application, target information input by the target object and prompt information of the target generation model are input into the target generation model for processing, and outputting an initial interactive flowchart includes: inputting the target information input by the target object and prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interactive flowchart to be constructed; determining an information set of the interactive flowchart to be constructed based on the text information, wherein the information set includes at least attribute information of each node of the interactive flowchart to be constructed and attribute information of the connection line between every two nodes of the interactive flowchart to be constructed; rendering the information in the information set of the interactive flowchart to be constructed to obtain an initial interactive flowchart.

[0049] For example, when building an information extraction module, we first remove irrelevant text information, extract nodes, node descriptions, node connection properties and other information from the flowchart Markdown, and use NetworkX to combine the nodes and connection information into a graph structure. Then, by passing the graph structure information to the front end, the front-end process canvas driver renders the graph structure and displays the real process structure to the user.

[0050] Through the above scheme, the initial interaction flowchart can be generated quickly and accurately according to the target generation model.

[0051] Optionally, in the processing method of the interaction flowchart provided in the embodiment of the present application, the target generation model is obtained in the following manner: obtaining multiple voice interaction data in a vertical field from a target database, and converting each voice interaction data to obtain multiple question and answer texts, wherein the vertical field includes at least the financial field; obtaining multiple target texts, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about the model; constructing a training data set based on the multiple question and answer texts and the multiple target texts; and obtaining the target generation model based on the training data set.

[0052] For example, a large model vertical field data set is constructed, and the large model vertical field data set contains business data, cognitive data, and compliance constraint data. Then, the final generation model to be used is determined based on the constructed vertical field data set.

[0053] In summary, by using the training set to train the model, the final generation model can be easily obtained.

[0054] Optionally, in the processing method of the interactive flowchart provided in the embodiment of the present application, obtaining the target generation model based on the training data set includes: adjusting the original generation model based on the training data set to obtain an adjusted model; quantizing the adjusted model to obtain a processed model; and accelerating inference on the processed model to obtain the target generation model.

[0055] For example, when fine-tuning large model quantization and inference acceleration deployment, you can use the AWQ quantization method based on fine-tuning the large model weights to reduce the weight volume of the large model and lower the threshold for server GPU requirements. At the same time, you can optionally set the corresponding int4 / int8 precision to reduce the model storage size, improve the large model throughput through vllm inference acceleration, optimize the response rate, and build a large model streaming interactive response based on fastapi.

[0056] In summary, by accelerating model inference, the throughput of large models can be improved and the response rate can be optimized.

[0057] Optionally, in the method for processing an interaction flowchart provided in an embodiment of the present application, determining a target interaction flowchart based on a verification result includes: if the verification result indicates that the initial interaction flowchart is a complete flowchart, taking the initial interaction flowchart as the target interaction flowchart; if the verification result indicates that the initial interaction flowchart is an incomplete flowchart, determining abnormal information present in the initial interaction flowchart; obtaining a description text for describing the abnormal information in the initial interaction flowchart; inputting the description text into a target generation model, performing correction processing on the initial interaction flowchart, and outputting the target interaction flowchart.

[0058] For example, NetworkX can be used to check the integrity of the graph structure to determine whether there is missing or redundant connection information that leads to independent nodes or loops, and preset the corresponding problem description text based on the abnormal information, input it into the fine-tuned large model for optimization and correction, and then render the optimized generation result of the large model to the front-end page, so that the interactive flowchart generation can meet complex needs.

[0059] Through the above solution, an interactive flow chart that meets the requirements can be easily obtained.

[0060] In this embodiment, in order to improve the quality of intelligent interaction expression information, enhance the process canvas generation capability in the large model field, and reduce the difficulty of complex process configuration, a method for generating intelligent interaction processes based on fine-tuning the vertical domain large model is proposed.

[0061] In this embodiment, a method for generating an intelligent interactive process based on fine-tuning a vertical domain big model is provided. By fine-tuning the general big model in the vertical field, building a task knowledge base, and strengthening the big model parsing and generating capabilities for field-specific tasks, the model can master the field knowledge, and the generated results can be more in line with business needs. The interactive process page rendering is achieved by realizing the automated parsing capability, and the relevant interactive node information is actively filled. According to the method in this embodiment, the big model capability in the vertical field is realized, and the task knowledge base is built to realize the intelligent interactive process generation capability, thereby providing intelligent reference examples and reducing the cost of manual configuration operations.

[0062] For example, this embodiment provides a method for generating intelligent interactive processes based on fine-tuning the vertical domain big model. Through the big model fine-tuning solution to learn domain knowledge, integrate RAG (retrieval enhanced generation) technology, build a flowchart generation knowledge base, deploy Prompt (big model prompt word) tasks to achieve the big model to generate interactive flowcharts for service business effects, and use information extraction strategies to achieve automatic filling of node numbers, node information, interactive processes, link intent and other information in the process canvas. For example, Figure 2 is a flowchart of an optional interactive flowchart processing method provided in an embodiment of the present application, such as Figure 2 As shown, the optional processing method of the interactive flowchart includes the following contents:

[0063] (1) Construction of large model vertical domain dataset

[0064] Step 1: Integrate business data: First, integrate the business knowledge base data, the real interaction data of online and voice interactions, and unify them into a multi-round question and answer text format;

[0065] Step 2: Construct cognitive data: In order to enable the big model to have self-cognition ability, it is necessary to add self-question and answer data, such as the construction of question and answer data related to "Who are you" to achieve self-cognition of the big model;

[0066] Step 3: Build compliance constraint data: Add risk prevention and rigorous Q&A knowledge to prevent compliance risks in the responses generated by large models;

[0067] Step 4: Standardization of data set training: For large models with multiple rounds of training data, you can set the prompt to describe the business domain knowledge, such as "You are an expert in financial management business in the financial field. The following is business domain question and answer information", and the constructed question and answer information will correspond to multiple rounds of user questions and assistant replies.

[0068] (2) Fine-tuning of the general large model in vertical fields

[0069] Step 1: Selection of general large model: Use open source large model to generate effect call of interactive flowchart, and manually evaluate the model with better general effect.

[0070] Step 2: Fine-tuning of general large models: In order to improve the domain knowledge learning ability of large models and reduce the hardware resource requirements for model training, and to meet the long text interaction requirements of the interactive process, the dataset constructed in step (1) is fine-tuned using the LongLora method. The optimal rank size is selected based on the scale of the large model, and the weights after fine-tuning are merged to form deployment weights.

[0071] (3) Fine-tuning large model quantization and inference acceleration deployment

[0072] Based on the fine-tuning of the large model weights in step (2), the AWQ quantization method is used to reduce the weight volume of the large model, lower the threshold for the server GPU requirement, and optionally set the corresponding int4 / int8 precision to reduce the model storage size. The throughput of the large model is improved through vllm inference acceleration, the response rate is optimized, and the large model streaming interactive response is built based on fastapi.

[0073] (4) Construct an interactive process canvas generation method based on fine-tuning the vertical domain large model, and Figure 3 is a flowchart of a large model generation interaction flowchart based on RAG retrieval enhancement capability in this embodiment, such as Figure 3 As shown, the specific construction process is as follows:

[0074] Step 1: First, you need to add text descriptions to the historical flowcharts, clarify the specific business scenario theme overview based on the corresponding business process, design triples (theme overview-flowchart description-flowchart structure) to build an interactive process knowledge base, so that you can retrieve the corresponding process information based on the user description content, use the triple information as the prompt information for fine-tuning the large model as the model input, and output the corresponding flowchart.

[0075] Step 2: Build RAG search enhancement capabilities to assist in the generation of large models. Users can improve the engineering effect of large model prompt words. By introducing existing knowledge structures and constraining the randomness of large model generation, the generated results are more in line with task requirements. The specific technical process of building RAG capabilities in the large model generation process canvas is as follows:

[0076] 1) Interaction process knowledge base storage: According to the above triple process knowledge base, the topic overview and flowchart description in the triple are extracted as the standard description data set, and the standard description data is used to construct similar questions with similar pronunciation.

[0077] 2) Vector retrieval knowledge base storage requires the construction of a knowledge retrieval model. This embodiment adopts the Cosent knowledge retrieval model based on the BGE pre-trained weights, performs text matching training on the standard description data, vectorizes the process knowledge base based on the model training results, and stores the vectors in the vector database.

[0078] 3) Generate process description text according to user input needs, embed and vectorize the process description text through the above knowledge retrieval model, and retrieve the corresponding knowledge content in the vector retrieval knowledge base through vector retrieval, so as to realize the retrieval of historical related process knowledge according to the user input topic or flowchart description.

[0079] Step 3: Based on the information supplement of the triple process knowledge base prompt, a large model business interaction process generation method is implemented. The specific technical process is as follows:

[0080] 1) Construction of large model prompt words: Based on the knowledge base content retrieved in step (4), the flowchart description and the flowchart structure are extracted and spliced ​​as the large model prompt words. At the same time, the business domain scenario and flowchart format requirements are spliced ​​to guide the large model to generate results that meet the business domain description.

[0081] 2) Large model generation: Input the above prompt words and user questions into the fine-tuned large model deployed in step (3), and output the large model generation text information and flowchart Markdown information.

[0082] 3) Effective information extraction: Build an information extraction module, first remove irrelevant text information, extract nodes, node descriptions, node connection properties and other information from the flowchart Markdown, and use NetworkX to organize the nodes and connection information into a graph structure.

[0083] 4) Flowchart display: By passing the graph structure information to the front end, the front-end process canvas driver renders the graph structure and displays the real process structure to the user.

[0084] (5) Complete process verification and intelligent filling: Use NetworkX to verify the integrity of the graph structure to determine whether there is missing or redundant connection information that leads to independent nodes or loops. Preset the corresponding problem description text based on the abnormal information and input it into the fine-tuned large model for optimization and correction. Repeat the step three process of step (4) and render the optimized generation result of the large model to the front-end page, so that the interactive flowchart generation can meet complex needs.

[0085] In this embodiment, large model training data is constructed through existing knowledge sets, and large model fine-tuning is achieved through the longlora method to have the required vertical domain knowledge response capability, and the response efficiency is improved through AWQ quantization and vllm reasoning acceleration methods; RAG retrieval enhancement capabilities are constructed to implement the Cosent knowledge retrieval model based on BGE pre-trained weights, and auxiliary knowledge content screening is achieved through vectorized retrieval; RAG recall results are assembled into Prompt words, combined with user input content, and input into the large model to achieve the interactive flowchart text generation effect, and an information extraction module is constructed to obtain flowchart-related information; a flowchart is generated based on the large model for integrity verification, abnormal information is supplemented and corrected, and complex interactive flowchart generation tasks are adapted.

[0086] The intelligent interactive process generation method based on fine-tuning the vertical domain big model in this embodiment, through longlora fine-tuning training of the general big model capabilities, enables the big model to have the ability to learn vertical domain knowledge; through quantization of the fine-tuning model, the storage space occupied by the big model is reduced; the vllm reasoning acceleration capability is used to reduce the response time of the big model; the RAG retrieval enhancement capability is constructed to realize the reference of existing knowledge, which can guide the big model to perform complex process tasks; the retrieved knowledge is used as the input of the big model to generate the interactive flowchart result, and the information extraction module is used to extract the relevant information of the valid nodes, and the process verification and filling capabilities are added, so that the big model generation effect can better meet the process interaction diagram of the real business scenario and improve the efficiency of personnel allocation.

[0087] In summary, the method for processing an interactive flowchart provided in an embodiment of the present application receives target information input by a target object, wherein the target information is used to describe information of an interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; the target information input by the target object is input into a knowledge retrieval model for processing, and process information related to the target information is retrieved; based on the process information related to the target information, prompt information of a target generation model is determined, and the target information input by the target object and the prompt information of the target generation model are input into the target generation model for processing, and an initial interactive flowchart is output, wherein the target generation model is a model obtained by adjusting an original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interactive flowchart; an integrity check is performed on the initial interactive flowchart to obtain a check result, and based on the check result, a target interactive flowchart is determined, thereby solving the problem in the related art of relying on manual configuration of interactive flowcharts related to financial business, resulting in poor effect of generating interactive flowcharts. By using the knowledge retrieval model to retrieve the corresponding process information according to the user's description content, and then determining the prompt words of the large model (i.e., the target generation model) based on the retrieved content, and inputting the prompt words and user questions into the fine-tuned large model (i.e., the target generation model), the initial interaction flowchart is output, and the fine-tuned large model is a model obtained by vertically fine-tuning the general large model. The initial interaction flowchart is then checked for integrity to obtain the target interaction flowchart, thereby eliminating the need to rely on manual configuration of interaction flowcharts related to financial services, and improving the quality of intelligent interaction expression information, strengthening the process canvas generation capability within the large model field, reducing the difficulty of configuring complex processes, and thereby improving the effect of generating interaction flowcharts.

[0088] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0089] The embodiment of the present application also provides a processing device for an interactive flowchart. It should be noted that the processing device for an interactive flowchart in the embodiment of the present application can be used to execute the processing method for an interactive flowchart provided in the embodiment of the present application. The following introduces the processing device for an interactive flowchart provided in the embodiment of the present application.

[0090] Figure 4 Schematic diagram of a processing device for an interactive flow chart according to an embodiment of the present application. Figure 4 As shown, the device includes: a first receiving unit 401, a first input unit 402, a first processing unit 403 and a second processing unit 404.

[0091] Specifically, the first receiving unit 401 is used to receive target information input by the target object, wherein the target information is used to describe information of the interaction flowchart to be constructed, and the interaction flowchart to be constructed is an interaction flowchart related to financial services;

[0092] The first input unit 402 is used to input the target information input by the target object into the knowledge retrieval model for processing, and retrieve the process information related to the target information;

[0093] The first processing unit 403 is used to determine prompt information of the target generation model based on process information related to the target information, and input the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and output an initial interaction flowchart, wherein the target generation model is a model obtained by adjusting the original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate the interaction flowchart;

[0094] The second processing unit 404 is used to perform integrity check on the initial interaction flowchart, obtain a check result, and determine a target interaction flowchart according to the check result.

[0095] In summary, the interactive flowchart processing device provided in the embodiment of the present application receives the target information input by the target object through the first receiving unit 401, wherein the target information is used to describe the information of the interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; the first input unit 402 inputs the target information input by the target object into the knowledge retrieval model for processing, and retrieves the process information related to the target information; the first processing unit 403 determines the prompt information of the target generation model based on the process information related to the target information, and inputs the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputs the initial interactive flowchart, wherein the target generation model is a model obtained by adjusting the original generation model based on the data of the vertical field, the data of the vertical field is data related to the financial field, and the original generation model is used to generate the interactive flowchart; the second processing unit 404 performs integrity check on the initial interactive flowchart to obtain the check result, and determines the target interactive flowchart based on the check result, thereby solving the problem that the interactive flowchart related to financial business in the related technology relies on manual configuration, resulting in poor effect of generating the interactive flowchart. By using the knowledge retrieval model to retrieve the corresponding process information according to the user's description content, and then determining the prompt words of the large model (i.e., the target generation model) based on the retrieved content, and inputting the prompt words and user questions into the fine-tuned large model (i.e., the target generation model), the initial interaction flowchart is output, and the fine-tuned large model is a model obtained by vertically fine-tuning the general large model. The initial interaction flowchart is then checked for integrity to obtain the target interaction flowchart, thereby eliminating the need to rely on manual configuration of interaction flowcharts related to financial services, and improving the quality of intelligent interaction expression information, strengthening the process canvas generation capability within the large model field, reducing the difficulty of configuring complex processes, and thereby improving the effect of generating interaction flowcharts.

[0096] Optionally, in the processing device of the interactive flowchart provided in the embodiment of the present application, the first input unit includes: a first processing module, which is used to vectorize the target information input by the target object through a knowledge retrieval model to obtain vector information corresponding to the target information; a first retrieval module, which is used to retrieve vector data related to the vector information corresponding to the target information in a vector database based on the vector information corresponding to the target information, wherein the vector database is used to store multiple vector data, each vector data is vector data obtained by vectorizing the information used to describe each historical flowchart; a first determination module, which is used to obtain process information related to the target information based on the vector data related to the vector information corresponding to the target information.

[0097] Optionally, in the interactive flowchart processing device provided in the embodiment of the present application, the first processing unit includes: a first acquisition module, used to obtain the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information based on the process information related to the target information; a first splicing module, used to splice the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information to obtain prompt information of the target generation model.

[0098] Optionally, in the processing device of the interactive flowchart provided in the embodiment of the present application, the first processing unit includes: a second processing module, which is used to input the target information input by the target object and the prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interactive flowchart to be constructed; a second determination module, which is used to determine the information set of the interactive flowchart to be constructed based on the text information, wherein the information set includes at least the attribute information of each node of the interactive flowchart to be constructed and the attribute information of the connection line between every two nodes of the interactive flowchart to be constructed; a third processing module, which is used to render the information in the information set of the interactive flowchart to be constructed to obtain an initial interactive flowchart.

[0099] Optionally, in the processing device of the interaction flowchart provided in the embodiment of the present application, the target generation model is obtained by the following units: a third processing unit, used to obtain multiple voice interaction data in a vertical field from a target database, and convert each voice interaction data to obtain multiple question and answer texts, wherein the vertical field includes at least the financial field; a first acquisition unit, used to obtain multiple target texts, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about the model; a first construction unit, used to construct a training data set based on multiple question and answer texts and multiple target texts; a first determination unit, used to obtain the target generation model based on the training data set.

[0100] Optionally, in the processing device of the interactive flowchart provided in the embodiment of the present application, the first determination unit includes: a first adjustment module, used to adjust the original generation model based on the training data set to obtain an adjusted model; a fourth processing module, used to quantize the adjusted model to obtain a processed model; a first acceleration module, used to accelerate the inference of the processed model to obtain a target generation model.

[0101] Optionally, in the processing device for the interaction flowchart provided in the embodiment of the present application, the second processing unit includes: a third determination module, which is used to use the initial interaction flowchart as the target interaction flowchart if the verification result indicates that the initial interaction flowchart is a complete flowchart; a fourth determination module, which is used to determine the abnormal information existing in the initial interaction flowchart if the verification result indicates that the initial interaction flowchart is an incomplete flowchart; a second acquisition module, which is used to acquire a description text for describing the abnormal information in the initial interaction flowchart; and a fifth processing module, which is used to input the description text into a target generation model, perform correction processing on the initial interaction flowchart, and output the target interaction flowchart.

[0102] The processing device of the interactive flowchart includes a processor and a memory. The above-mentioned first receiving unit 401, first input unit 402, first processing unit 403 and second processing unit 404 are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0103] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the effect of generating the interactive flowchart can be improved by adjusting the kernel parameters.

[0104] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0105] An embodiment of the present invention provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, the processing method of the interaction flowchart is implemented.

[0106] An embodiment of the present invention provides a processor, which is used to run a program, wherein the processing method of the interaction flowchart is executed when the program is running.

[0107] like Figure 5As shown, an embodiment of the present invention provides an electronic device, the device includes a processor, a memory and a program stored in the memory and executable on the processor, and the processor implements the following steps when executing the program: receiving target information input by a target object, wherein the target information is used to describe information of an interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; inputting the target information input by the target object into a knowledge retrieval model for processing, and retrieving process information related to the target information; based on the process information related to the target information, determining prompt information of a target generation model, and inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting an initial interactive flowchart, wherein the target generation model is a model obtained by adjusting an original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interactive flowchart; performing an integrity check on the initial interactive flowchart to obtain a check result, and determining a target interactive flowchart based on the check result.

[0108] When the processor executes the program, the following steps are also implemented: inputting the target information input by the target object into the knowledge retrieval model for processing, and retrieving process information related to the target information, including: vectorizing the target information input by the target object through the knowledge retrieval model to obtain vector information corresponding to the target information; based on the vector information corresponding to the target information, retrieving vector data related to the vector information corresponding to the target information in the vector database, wherein the vector database is used to store multiple vector data, each vector data is vector data obtained by vectorizing the information used to describe each historical flowchart; based on the vector data related to the vector information corresponding to the target information, obtaining process information related to the target information.

[0109] When the processor executes the program, the following steps are also implemented: based on the process information related to the target information, the prompt information of the target generation model is determined, including: based on the process information related to the target information, the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information are obtained; the description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information are spliced ​​to obtain the prompt information of the target generation model.

[0110] When the processor executes the program, the following steps are also implemented: inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting the initial interaction flowchart includes: inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interaction flowchart to be constructed; determining the information set of the interaction flowchart to be constructed based on the text information, wherein the information set at least includes the attribute information of each node of the interaction flowchart to be constructed and the attribute information of the connection line between every two nodes of the interaction flowchart to be constructed; rendering the information in the information set of the interaction flowchart to be constructed to obtain the initial interaction flowchart.

[0111] When the processor executes the program, the following steps are also implemented: the target generation model is obtained in the following manner: a plurality of voice interaction data in a vertical field is obtained from a target database, and each voice interaction data is converted and processed to obtain a plurality of question and answer texts, wherein the vertical field includes at least the financial field; a plurality of target texts are obtained, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about the model; a training data set is constructed based on the plurality of question and answer texts and the plurality of target texts; and a target generation model is obtained based on the training data set.

[0112] When the processor executes the program, the following steps are also implemented: based on the training data set, obtaining the target generation model includes: based on the training data set, adjusting the original generation model to obtain an adjusted model; quantizing the adjusted model to obtain a processed model; and accelerating the inference of the processed model to obtain the target generation model.

[0113] When the processor executes the program, the following steps are also implemented: according to the verification result, the target interaction flowchart is determined, including: if the verification result indicates that the initial interaction flowchart is a complete flowchart, the initial interaction flowchart is used as the target interaction flowchart; if the verification result indicates that the initial interaction flowchart is an incomplete flowchart, the abnormal information existing in the initial interaction flowchart is determined; a description text for describing the abnormal information in the initial interaction flowchart is obtained; the description text is input into the target generation model, the initial interaction flowchart is corrected, and the target interaction flowchart is output.

[0114] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0115] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having the following method steps: receiving target information input by a target object, wherein the target information is used to describe information of an interactive flowchart to be constructed, and the interactive flowchart to be constructed is an interactive flowchart related to financial business; inputting the target information input by the target object into a knowledge retrieval model for processing, and retrieving process information related to the target information; based on the process information related to the target information, determining prompt information of a target generation model, and inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting an initial interactive flowchart, wherein the target generation model is a model obtained by adjusting the original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interactive flowchart; performing an integrity check on the initial interactive flowchart to obtain a check result, and determining the target interactive flowchart based on the check result.

[0116] When executed on a data processing device, it is also suitable for executing an initialized program having the following method steps: inputting the target information input by the target object into the knowledge retrieval model for processing, and retrieving process information related to the target information, including: vectorizing the target information input by the target object through the knowledge retrieval model to obtain vector information corresponding to the target information; based on the vector information corresponding to the target information, retrieving vector data related to the vector information corresponding to the target information in a vector database, wherein the vector database is used to store multiple vector data, each vector data being vector data obtained by vectorizing the information used to describe each historical flowchart; based on the vector data related to the vector information corresponding to the target information, obtaining process information related to the target information.

[0117] When executed on a data processing device, it is also suitable for executing an initialized program having the following method steps: based on process information related to target information, determining prompt information of a target generation model includes: based on process information related to target information, obtaining description information of historical flowcharts related to target information and structural information of historical flowcharts related to target information; splicing the description information of historical flowcharts related to target information and structural information of historical flowcharts related to target information to obtain prompt information of the target generation model.

[0118] When executed on a data processing device, it is also suitable for executing an initialization program having the following method steps: inputting the target information of the target object input and the prompt information of the target generation model into the target generation model for processing, and outputting the initial interaction flowchart including: inputting the target information of the target object input and the prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interaction flowchart to be constructed; determining the information set of the interaction flowchart to be constructed based on the text information, wherein the information set includes at least the attribute information of each node of the interaction flowchart to be constructed and the attribute information of the connection line between every two nodes of the interaction flowchart to be constructed; rendering the information in the information set of the interaction flowchart to be constructed to obtain the initial interaction flowchart.

[0119] When executed on a data processing device, it is also suitable for executing a program initialized with the following method steps: the target generation model is obtained in the following manner: a plurality of voice interaction data in a vertical field is obtained from a target database, and each voice interaction data is converted and processed to obtain a plurality of question and answer texts, wherein the vertical field includes at least the financial field; a plurality of target texts are obtained, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about the model; a training data set is constructed based on the plurality of question and answer texts and the plurality of target texts; and a target generation model is obtained based on the training data set.

[0120] When executed on a data processing device, it is also suitable for executing a program that is initialized with the following method steps: based on a training data set, obtaining a target generation model includes: based on the training data set, adjusting the original generation model to obtain an adjusted model; quantizing the adjusted model to obtain a processed model; and accelerating inference on the processed model to obtain a target generation model.

[0121] When executed on a data processing device, it is also suitable for executing an initialization program having the following method steps: determining a target interaction flowchart based on a verification result, including: if the verification result indicates that the initial interaction flowchart is a complete flowchart, then using the initial interaction flowchart as the target interaction flowchart; if the verification result indicates that the initial interaction flowchart is an incomplete flowchart, then determining abnormal information existing in the initial interaction flowchart; obtaining a description text for describing the abnormal information in the initial interaction flowchart; inputting the description text into a target generation model, performing correction processing on the initial interaction flowchart, and outputting the target interaction flowchart.

[0122] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0123] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0124] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0126] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0127] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0128] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0129] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0130] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0131] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for processing an interactive flowchart, characterized in that: include: Receive target information input by a target object, wherein the target information is used to describe information of an interaction flowchart to be constructed, and the interaction flowchart to be constructed is an interaction flowchart related to financial services; Inputting the target information input by the target object into the knowledge retrieval model for processing, and retrieving process information related to the target information; Based on the process information related to the target information, the prompt information of the target generation model is determined, and the target information input by the target object and the prompt information of the target generation model are input into the target generation model for processing, and an initial interaction flowchart is output, wherein the target generation model is a model obtained by adjusting the original generation model based on the data of the vertical field, the data of the vertical field is data related to the financial field, and the original generation model is used to generate the interaction flowchart; An integrity check is performed on the initial interaction flowchart to obtain a check result, and a target interaction flowchart is determined based on the check result.

2. The method according to claim 1, characterized in that The target information input by the target object is input into the knowledge retrieval model for processing, and the process information related to the target information is retrieved, including: Vectorizing the target information input by the target object through the knowledge retrieval model to obtain vector information corresponding to the target information; According to the vector information corresponding to the target information, vector data related to the vector information corresponding to the target information is retrieved from a vector database, wherein the vector database is used to store a plurality of vector data, each of which is vector data obtained by vectorizing information used to describe each historical flow chart; Based on the vector data related to the vector information corresponding to the target information, process information related to the target information is obtained.

3. The method according to claim 1, characterized in that Based on the process information related to the target information, the prompt information for determining the target generation model includes: Based on the process information related to the target information, obtaining description information of the historical flowchart related to the target information and structure information of the historical flowchart related to the target information; The description information of the historical flowchart related to the target information and the structural information of the historical flowchart related to the target information are spliced ​​to obtain the prompt information of the target generation model.

4. The method according to claim 1, characterized in that: Inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and outputting the initial interaction flow chart includes: Inputting the target information input by the target object and the prompt information of the target generation model into the target generation model for processing to obtain text information, wherein the text information is used to describe the process information of the interactive flowchart to be constructed; Determine, based on the text information, an information set of the interactive flowchart to be constructed, wherein the information set at least includes attribute information of each node of the interactive flowchart to be constructed and attribute information of a connection line between every two nodes of the interactive flowchart to be constructed; Rendering is performed on the information in the information set of the interaction flowchart to be constructed to obtain the initial interaction flowchart.

5. The method according to claim 1, characterized in that The target generation model is obtained in the following way: Acquire multiple voice interaction data in vertical fields from a target database, and convert each voice interaction data to obtain multiple question-answer texts, wherein the vertical fields at least include the financial field; Acquire a plurality of target texts, wherein the target texts are used to answer target type questions, and the target type questions include at least one of the following: risky questions and questions about models; Constructing a training data set based on the multiple question-answer texts and the multiple target texts; Based on the training data set, the target generation model is obtained.

6. The method according to claim 5, characterized in that Based on the training data set, obtaining the target generation model includes: Based on the training data set, adjusting the original generation model to obtain an adjusted model; Performing quantization processing on the adjusted model to obtain a processed model; The processed model is subjected to inference acceleration to obtain the target generation model.

7. The method according to claim 1, characterized in that According to the verification result, determining the target interaction flow chart includes: If the verification result indicates that the initial interaction flowchart is a complete flowchart, the initial interaction flowchart is used as the target interaction flowchart; If the verification result indicates that the initial interaction flowchart is an incomplete flowchart, determining abnormal information existing in the initial interaction flowchart; Obtaining a description text for describing the abnormal information in the initial interaction flowchart; The description text is input into the target generation model, the initial interaction flowchart is corrected, and the target interaction flowchart is output.

8. A processing device for an interactive flow chart, characterized in that: include: A first receiving unit, configured to receive target information input by a target object, wherein the target information is used to describe information of an interaction flowchart to be constructed, and the interaction flowchart to be constructed is an interaction flowchart related to financial services; A first input unit, used to input the target information input by the target object into the knowledge retrieval model for processing, and retrieve the process information related to the target information; A first processing unit is used to determine prompt information of a target generation model based on process information related to the target information, and input the target information input by the target object and the prompt information of the target generation model into the target generation model for processing, and output an initial interaction flowchart, wherein the target generation model is a model obtained by adjusting an original generation model based on data in a vertical field, the data in the vertical field is data related to the financial field, and the original generation model is used to generate an interaction flowchart; The second processing unit is used to perform integrity check on the initial interaction flowchart to obtain a check result, and determine a target interaction flowchart according to the check result.

9. A computer-readable storage medium, characterized in that: The storage medium stores a program, wherein the program executes the interactive flowchart processing method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the processing method of the interactive flowchart described in any one of claims 1 to 7.

Images