Financial RPA method and system based on artificial intelligence large language model

Abstract

The invention relates to robot process automation, in particular to a financial RPA method and system based on an artificial intelligence large language model, and the method comprises the following steps: obtaining a natural language instruction and a file inputted by a user; converting into a structured instruction set through a large language model; calling a corresponding API interface based on the structured instruction set to trigger the RPA, and transmitting parameters to the RPA; acquiring data for generating a report based on an RPA preset process; sorting and cleaning the data, and transmitting the data to a large language model through an API interface; the large language model carries out report generation based on the data template; and outputting the generated report to the user. Compared with the prior art, the method has the advantages that the obvious defects and deficiencies of the process automation technology in adaptability and natural language processing capacity in the prior art are overcome, the advantages of the artificial intelligence large language model are utilized, the natural language input is accurately understood and processed, and the flexibility and adaptability of the automatic process are improved.

Description

Technical Field

[0001] The present invention relates to a robotic process automation, and in particular to a financial RPA method and system based on an artificial intelligence large language model. Background Art

[0002] With the continuous development of information technology and the in-depth promotion of enterprise informatization, Robotic Process Automation (RPA) software has become an important means to improve enterprise operational efficiency and reduce operating costs. However, the existing process automation technology still has many problems and defects in practice.

[0003] First, existing process automation tools mostly rely on fixed rules and templates. These rules and templates are often designed based on specific business processes, and their adaptability to complex and changing business scenarios is limited. When business processes change, it is usually necessary to manually adjust rules and templates, which not only increases the workload and complexity, but may also affect the accuracy and stability of the automated process.

[0004] Secondly, existing process automation tools have obvious deficiencies when dealing with links involving natural language understanding. Natural language is the main way for humans to communicate, but in computer processing, the complexity and diversity of natural language make it very difficult to understand and parse. Existing process automation tools often fail to accurately capture user intentions and needs, resulting in deviations or errors in the execution of automated processes.

[0005] In addition, existing process automation tools have obvious deficiencies in document generation. In daily work, document writing is an important task. These documents usually have a fixed format and need to be compiled from data from multiple channels. Although process automation tools are excellent in data collection, their automation level is difficult to further improve due to the lack of document writing capabilities. This defect limits the application of tools in comprehensive automation processes, and users still need to manually complete document writing, which affects the improvement of overall work efficiency.

[0006] To sum up, the existing process automation technology has obvious defects and deficiencies in adaptability and natural language processing capabilities, and it is necessary to further improve the flexibility and adaptability of the automated process. Summary of the invention

[0007] The purpose of the present invention is to provide a financial RPA method and system based on an artificial intelligence big language model in order to solve at least one of the above problems, so as to solve the obvious defects and deficiencies in the adaptability and natural language processing capabilities of the process automation technology in the prior art. This solution utilizes the advantages of the artificial intelligence big language model to achieve accurate understanding and processing of natural language input, and improve the flexibility and adaptability of the automation process.

[0008] The purpose of the present invention is achieved through the following technical solutions:

[0009] The first aspect of the present invention discloses a financial RPA method based on an artificial intelligence large language model, comprising the following steps:

[0010] Step 101: Obtain natural language instructions and files input by a user;

[0011] Step 102: converting the natural language instructions and files into a structured instruction set through a large language model;

[0012] Step 103: Call the corresponding API interface based on the structured instruction set to trigger the RPA, and transmit parameters to the RPA;

[0013] Step 104: Acquire data for generating a report based on the RPA preset process;

[0014] Step 105: Arrange and clean the data, and transmit it to the large language model through the API interface;

[0015] Step 106: The large language model generates a report based on the data template;

[0016] Step 107: Output the generated report to the user.

[0017] Preferably, in step 102, the large language model parses natural language instructions and files through deep learning and / or natural language processing, and generates a structured instruction set.

[0018] Preferably, the parsing includes identifying key entities and relationships and extracting keywords and parameters.

[0019] Preferably, in step 104, the RPA acquires data for generating reports through element picking and / or computer vision.

[0020] Preferably, in step 105, organizing and cleaning the data includes data conversion, data verification and data merging.

[0021] The second aspect of the present invention discloses a financial RPA system based on an artificial intelligence large language model, which adopts any of the financial RPA methods described above;

[0022] The financial RPA system includes:

[0023] A front-end interaction module, used to receive user input and output reports to the user;

[0024] Backend server, including large language model engine module and RPA execution module;

[0025] Large language model engine module, used to process natural language instructions and files input by users and generate output reports;

[0026] RPA execution module, used to automatically acquire and process data.

[0027] Preferably, the front-end interaction module and the back-end server communicate asynchronously via Ajax or WebSocket.

[0028] Preferably, the front-end interaction module interacts with the user through a front-end interaction interface constructed by a front-end technology stack.

[0029] Preferably, the front-end technology stack includes HTML, CSS and JavaScript.

[0030] Preferably, the front-end interactive interface is a responsive design.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention proposes a robotic process automation method based on an artificial intelligence large language model, which has important practical significance and application value. This method can take advantage of the artificial intelligence large language model to accurately understand and process natural language input, and improve the flexibility and adaptability of the automation process.

[0033] Specific:

[0034] Dynamic process construction: Traditional RPA process automation can only replace manual execution of a series of tasks according to preset logic. The present invention uses a large language model to decompose business processes, and can dynamically identify automated processes that match user descriptions, directly give users answers as needed, or give answers after collecting relevant data through RPA, without pre-defining fixed templates or rules. Compared with the previous use of RPA processes, it is innovative in the interactive mode, effectively improving the convenience and applicability of use.

[0035] Structured data extraction: By using a large language model to automatically extract business data from various systems for processing and analysis in the RPA process, key structured data can be automatically extracted to provide an accurate data source for document generation, effectively reducing the shortcomings of reports generated by large language models in the past, which had no data source and were prone to large model illusions.

[0036] Intelligent filling of templates: Based on the document template selected by the user and the extracted structured data, the large language model can intelligently fill in the various fields in the template and generate a document that meets the requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a schematic diagram of the front-end interactive interface of Example 1.

[0038] Figure 2 This is a flowchart of the financial RPA method of Example 2. DETAILED DESCRIPTION

[0039] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] Example 1

[0041] A financial RPA system based on artificial intelligence large language model, such as Figure 1 As shown, the financial RPA method described in Example 1 above is adopted;

[0042] The financial RPA system includes:

[0043] A front-end interaction module, used to receive user input and output reports to the user;

[0044] Backend server, including large language model engine module and RPA execution module;

[0045] Large language model engine module, used to process natural language instructions and files input by users and generate output reports;

[0046] RPA execution module, used to automatically acquire and process data.

[0047] More specifically, in this embodiment:

[0048] A robotic process automation system based on artificial intelligence large language model, including the following modules:

[0049] A. Front-end interactive module (web page interactive interface);

[0050] B. Robotic process automation (hereinafter referred to as RPA) execution module;

[0051] C. Large language model engine module;

[0052] The RPA execution module and the large language model engine module are included in the backend server.

[0053] (1) Front-end interaction module

[0054] As a bridge between users and software, the web interface is designed with full consideration of user experience and technical implementation. The interface adopts responsive design to adapt to different devices and screen sizes, ensuring that users can get a consistent experience on different platforms.

[0055] In terms of function implementation, the front-end interactive interface uses front-end technology stacks such as HTML, CSS and JavaScript. The interface provides users with a friendly interactive method through graphical elements, form controls and buttons. When users enter relevant information or requirements of business processes through the interface, the front-end code will monitor the user's input behavior in real time and verify and format the data.

[0056] After the user submits the input, the front-end interaction module will use technologies such as Ajax or WebSocket to communicate asynchronously with the back-end server. By sending HTTP requests, the front-end interaction module passes the user's input data to the back-end large language model engine module. At the same time, the front-end interaction module will also process the files uploaded by the user, such as converting the files into binary streams or performing necessary format conversions to meet the needs of back-end processing.

[0057] After receiving the data transmitted by the front end, the large language model engine module will use deep learning algorithms and natural language processing technology to deeply analyze the user's intentions and business processes. The large language model will analyze the semantics, grammar and contextual information in the text, identify key entities and relationships, and generate corresponding instructions or data.

[0058] These instructions or data will then trigger the start of the RPA process. RPA realizes the automation of business processes by integrating various APIs and automation tools. Each step in the process is carefully designed and optimized to ensure efficient and accurate completion of tasks.

[0059] During the process execution, the front-end interaction module will monitor the data returned by the back-end in real time. Through the WebSocket polling mechanism, the front-end can obtain the execution status and intermediate results of the process in real time. Once the process execution is completed, the back-end will return the final processing results or generated documents to the front-end.

[0060] After receiving the data, the front end will use JavaScript to parse and process the data. For the processing results, the front end will choose the appropriate display method, such as table, chart or text, according to the data type and format. For the generated documents, the front end will use document rendering technology, such as PDF.js or Canvas API, to render and display the documents on the interface.

[0061] (2)RPA execution module

[0062] In the current wave of digitalization, enterprises are facing various technical challenges in their daily business operations. Especially in scenarios with incompatible software interfaces, complex operation processes and high repetitiveness, these problems not only consume a lot of time and human resources of enterprises, but also seriously affect the efficiency and accuracy of business processing. RPA, with its efficient and flexible characteristics, has become an ideal choice for solving such problems.

[0063] In this solution, the RPA execution module is responsible for accurately and automatically executing the corresponding business processes based on the parsing results of the large language model engine. To achieve this goal, we selected the most advanced RPA development software and RPA robot control platform on the market, which are highly integrated and scalable, and integrate a variety of business process processing components, including but not limited to data collection, data processing, task scheduling and other functions. These components work together to cover all aspects of the business process and achieve comprehensive automated processing.

[0064] According to the user's business demand scenarios, we have pre-developed corresponding RPA processes for collecting data from multiple systems. RPA element picking technology is based on the command interface provided by the operating system or the structural analysis of the web page. For some software that cannot be obtained through the operating system command interface, the RPA execution module can use computer vision (CV) technology to parse the software UI image to locate and identify the various elements in the software and perform a series of information capture operations.

[0065] In terms of data collection, the RPA execution module supports automatic capture of multiple data sources and formats. Whether it is a relational database, API interface, web page, document, etc., RPA can automatically extract the required data according to the preset rules and logic, and perform necessary cleaning and formatting.

[0066] Data processing is a key link in the RPA process. The RPA execution module in this solution provides a wealth of data processing functions, including data conversion, data verification, data merging, etc. These functions can ensure the accuracy and consistency of data and provide reliable data support for subsequent business processes.

[0067] In addition, the RPA robot control platform provides a variety of API interfaces, allowing users to remotely control the RPA execution module through programming. These API interfaces support multiple programming languages ​​and development frameworks, making it convenient for users to integrate and customize according to their needs. By calling these interfaces, users can obtain real-time information such as the running status and task progress of the RPA execution module, and can also monitor and adjust its behavior in real time.

[0068] (3) Large language model engine module

[0069] The big language model is an advanced technology in the field of natural language processing. It is based on deep learning algorithms and has mastered rich language knowledge and context understanding capabilities by learning from large amounts of text data. In this solution, the big language model engine is used to parse the natural language conversations of front-end users, thereby accurately obtaining the instructions that users expect to execute, and can intelligently call the corresponding API interface based on these instructions, thereby starting the RPA execution module and realizing the automated execution of business processes.

[0070] When users input business processes described in natural language through the front-end interaction module, the large language model engine module uses natural language processing capabilities to parse the conversation. This capability is based on deep learning technology. Through the training of large-scale corpora, the large language model engine can deeply parse text content and accurately capture user intentions and needs. After a complex text analysis and semantic understanding process, the engine converts the user's natural language input into a structured instruction set. These instruction sets describe in detail the operations that the user wants to perform, the required parameters, and specific conditions, providing clear input parameters for subsequent RPA process execution.

[0071] Once the instruction set is parsed, the large language model engine module intelligently calls the corresponding API interface according to the instruction content. These API interfaces serve as a key bridge between the RPA execution module and the large language model engine module, ensuring the accurate transmission of data and the effective execution of commands. The RPA execution module obtains the data required to execute subsequent processes and specific execution commands through the API interface, and provides real-time feedback on the execution results.

[0072] In the data collection stage, RPA can go deep into various business systems and databases, accurately capture the required data, and integrate it into a unified data platform. The data platform then further hands over the data to the large language model for processing. When processing data, the large language model can use its powerful language generation capabilities to structure the data into text reports. These reports not only contain the original content of the data, but also provide in-depth interpretation and analysis of the data in natural language. By pre-training the large language model with previous documents of the same type, various types of documents can be generated according to business needs, such as corporate financial flash reports, financial data analysis reports, feasibility studies, etc.

[0073] Example 2

[0074] A financial RPA method based on artificial intelligence large language model, such as Figure 2 As shown, the following steps are included:

[0075] Step 101: Obtain natural language instructions and files input by a user;

[0076] Step 102: converting the natural language instructions and files into a structured instruction set through a large language model;

[0077] Step 103: Call the corresponding API interface based on the structured instruction set to trigger the RPA, and transmit parameters to the RPA;

[0078] Step 104: Acquire data for generating a report based on the RPA preset process;

[0079] Step 105: Arrange and clean the data, and transmit it to the large language model through the API interface;

[0080] Step 106: The large language model generates a report based on the data template;

[0081] Step 107: Output the generated report to the user.

[0082] More specifically, in this embodiment:

[0083] Figure 2 The flowchart of this embodiment is shown. The process steps of the method are described in detail below:

[0084] In step 101, the user uses an intuitive and easy-to-use front-end page (front-end interaction module, see Figure 1 The front-end interactive interface shown in the figure) conducts a natural conversation with the digital assistant (the system of Example 1). The user can enter questions to be queried or instructions to perform specific tasks. After receiving this information, the front-end page will immediately pass this information to the back-end server. The back-end server then calls the pre-integrated large language model engine, which uses advanced natural language processing technology to parse and understand the user input and prepares to execute the subsequent steps, namely step 102.

[0085] In step 102, after receiving the user's conversation content, the large language model engine uses its powerful semantic understanding ability to extract keywords and parameters. These keywords and parameters are the key to the subsequent process execution, and they will be used to guide the specific operation of the RPA process.

[0086] In step 103, the large language model engine calls the corresponding API interface based on the extracted keywords and parameters. These API interfaces are responsible for starting and managing the RPA process. By passing in the process startup parameters, the large language model engine triggers the start of the RPA process. The RPA process then automatically performs a series of operations according to the preset rules and logic.

[0087] In step 104, the RPA process starts to simulate manual operation and logs into the specified financial software system. Once the login is successful, RPA can automatically obtain the report data required by the user. This data may include financial data, sales data, etc., which are crucial for the user's subsequent analysis and decision-making.

[0088] In step 105, after completing the acquisition of data, the RPA software will initially organize and clean the data. By removing redundant information, correcting erroneous data, etc., the accuracy and availability of the data are ensured. Subsequently, the RPA execution module sends the data to the receiving end through the API interface. After receiving the data, the receiving end will pass it to the large language model for further analysis and processing.

[0089] In step 106, after receiving the data, the large language model starts to generate a report according to the pre-trained data template. These data templates contain the format, structure and content requirements of the report, which provide guidance and constraints for the large language model to generate reports. The large language model generates the report required by the user by analyzing and understanding the data and combining the requirements of the template.

[0090] Finally, in step 107, the front-end interactive page displays the report generated by the large language model in an intuitive form in the dialogue. The user can view the content of the report through the front-end page and choose to download the report for subsequent use.

[0091] In summary,

[0092] By deeply integrating RPA and AI large language model technology, the present invention significantly expands the coverage of processes that can be automated, greatly reduces the tediousness and intervention requirements of manual operations, and thus achieves a significant improvement in work efficiency.

[0093] In terms of technical details, RPA can accurately simulate and execute a series of repetitive tasks in business processes through preset automation scripts, including but not limited to data capture, file processing, system operations, etc. The AI ​​big language model uses its powerful natural language processing capabilities to deeply analyze and intelligently process the text information involved in the process. The combination of the two can not only process structured data, but also understand and process unstructured texts such as contracts and reports, thereby further broadening the scope of automation applications.

[0094] In addition, automated processes can quickly complete a series of tasks through efficient algorithms and optimized execution paths, significantly shortening the business process cycle and improving the company's response speed. This enables companies to respond more quickly to market changes or customer needs, thereby enhancing their competitiveness.

[0095] The present invention is also innovative in terms of human-computer interaction. Users only need to interact with process automation robot software through natural language, without having to deeply understand the complex operation procedures and setting rules of previous RPA software. This natural and intuitive interaction method greatly reduces the user's learning cost and improves the user experience. At the same time, the execution of the automated process reduces errors and deviations caused by human factors, ensures the consistency and accuracy of the data, and effectively avoids errors in the data entry and processing process.

[0096] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.

Claims

1. A financial RPA method based on artificial intelligence large language model, characterized in that: The steps include: Step 101: Obtain natural language instructions and files input by a user; Step 102: converting the natural language instructions and files into a structured instruction set through a large language model; Step 103: Call the corresponding API interface based on the structured instruction set to trigger the RPA, and transmit parameters to the RPA; Step 104: Acquire data for generating a report based on the RPA preset process; Step 105: Arrange and clean the data, and transmit it to the large language model through the API interface; Step 106: The large language model generates a report based on the data template; Step 107: Output the generated report to the user.

2. According to claim 1, a financial RPA method based on an artificial intelligence large language model is characterized in that: In step 102, the large language model parses natural language instructions and files through deep learning and / or natural language processing, and generates a structured instruction set.

3. A financial RPA method based on artificial intelligence large language model according to claim 2, characterized in that: The parsing includes identifying key entities and relationships and extracting keywords and parameters.

4. According to claim 1, a financial RPA method based on an artificial intelligence large language model is characterized in that: In step 104, the RPA obtains data for generating reports through element picking and / or computer vision.

5. According to claim 1, a financial RPA method based on an artificial intelligence large language model is characterized in that: In step 105, the data is sorted and cleaned including data conversion, data verification and data merging.

6. A financial RPA system based on artificial intelligence large language model, characterized in that: Adopting the financial RPA method as described in any one of claims 1 to 5; The financial RPA system includes: A front-end interaction module, used to receive user input and output reports to the user; Backend server, including large language model engine module and RPA execution module; Large language model engine module, used to process natural language instructions and files input by users and generate output reports; RPA execution module, used to automatically acquire and process data.

7. A financial RPA system based on artificial intelligence large language model according to claim 6, characterized in that: The front-end interaction module and the back-end server communicate asynchronously via Ajax or WebSocket.

8. A financial RPA system based on artificial intelligence large language model according to claim 6, characterized in that: The front-end interaction module interacts with the user through the front-end interaction interface built by the front-end technology stack.

9. A financial RPA system based on artificial intelligence large language model according to claim 8, characterized in that: The front-end technology stack includes HTML, CSS and JavaScript.

10. A financial RPA system based on artificial intelligence large language model according to claim 8, characterized in that: The front-end interactive interface is a responsive design.

Images