Self-service terminal large model task processing system and method

Abstract

The invention discloses a self-service terminal large model task processing system and method, and the system comprises a multi-mode perception fusion unit which is integrated with an infrared multi-point touch screen, an MEMS microphone array, and an industrial grade code scanning module. A medical diagnosis knowledge graph and a financial transaction ontology industry exclusive graph are built in the multi-field intention classification model; the dynamic demand matching engine is used for analyzing a cross-domain task link by adopting a Transform-XL long sequence modeling technology; a Petri net simulator and a conflict detection algorithm are built in the task arrangement middleware, and the task arrangement middleware is used for guaranteeing the consistency of execution time sequences of multiple sub-tasks; the self-adaptive response generation module is used for supporting automatic conversion from a natural language instruction to a PLC control code; according to the method, the task identification accuracy can be improved based on multi-modal fusion; the dynamic demand matching engine supports processing of a large number of concurrent requests per second, and response delay is low.

Description

Technical Field

[0001] The present invention relates to the technical field of self-service terminals, and in particular to a large-scale model task processing system and method for self-service terminals. Background Art

[0002] Existing self-service terminals mostly use single-modal input (such as only supporting touch screens or scanning codes), which makes it difficult to meet the diverse needs in complex business scenarios; traditional task processing systems rely on preset rule libraries, and their response accuracy when facing personalized user needs is less than 65%, and they are unable to dynamically adjust service priorities; the task decomposition method based on rule engines has the risk of combinatorial explosion. When user needs involve more than three sub-tasks, the number of execution paths increases exponentially; the entity recognition accuracy of existing natural language understanding (NLU) models in vertical fields (such as medical care and finance) is less than 82%, resulting in frequent instruction conversion errors. Summary of the Invention

[0003] The purpose of the present invention is to provide a large-scale model task processing system for a self-service terminal in response to the above-mentioned problems in the prior art, thereby solving all or one of the above-mentioned problems in the prior art.

[0004] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: In one aspect, the present invention provides a self-service terminal large model task processing system, comprising: A multimodal sensing fusion unit, which integrates an infrared multi-touch screen, a MEMS microphone array, and an industrial-grade code scanning module; A multi-domain intent classification model based on the RoBERTa architecture, which includes a built-in medical diagnosis knowledge graph and a financial transaction ontology industry-specific graph; A dynamic demand matching engine, which uses Transformer-XL long sequence modeling technology to analyze cross-domain task links; Task scheduling middleware with built-in Petri net simulator and conflict detection algorithm to ensure the timing consistency of multiple subtask execution; An adaptive response generation module is used to support the automatic conversion of natural language instructions into PLC control codes.

[0005] As an improved solution, the multimodal perception fusion unit is also used to: achieve spatiotemporal alignment of the screen click area and the voice command based on the Yolov5 target detection algorithm; use the attention mechanism to dynamically allocate confidence weights for touch, voice, and code scanning; and eliminate abnormal operations based on the isolation forest algorithm.

[0006] As an improved solution, the intent classification model includes: Domain-adaptive pre-training module for fine-tuning using industry-specific corpora; Multi-hop reasoning mechanism for enhancing implicit demand understanding through knowledge graph retrieval; The real-time feedback correction module is used to dynamically adjust the initial intention determination results based on the user's subsequent operations.

[0007] As an improved solution, the dynamic demand matching engine is used to: use local sensitive hash index to accelerate similar demand retrieval; A resource scheduling strategy based on reinforcement learning optimizes the load balancing of CPU, memory, and GPU; a timing dependency parser is called to support time window constraints between tasks.

[0008] As an improved solution, the task orchestration middleware is also used to: support engineers in real-time debugging through a color-coded state machine visualization interface; automatically trigger compensation transactions; and utilize edge computing nodes to parallelize subtasks.

[0009] As an improved solution, the adaptive response generation module is also used to: cover the industrial control instruction set and API call instructions; verify the matching degree between instructions and requirements through the BERT model; and support the generation of voice broadcast and dynamic visual guidance composite feedback.

[0010] On the other hand, the present invention also provides a method for processing a large model task of a self-service terminal, comprising the following steps: The heterogeneous data stream input by the user is collected through the multimodal perception fusion unit; Input the data into the multi-domain intent classification model and output the top-5 candidate intents and their confidence levels. When the highest confidence level is less than 0.9, the active clarification process is triggered: additional information is collected through multiple rounds of dialogue; Use the dynamic demand matching engine to build task dependency graphs and identify critical paths; The task orchestration middleware generates executable intermediate representation code; The adaptive response generation module converts IR codes into device control commands; The execution results are simulated and verified through the digital twin engine; Establish user behavior profiles and continuously optimize intent classification model parameters; Regularly generate task processing quality reports, including KPIs such as response time and instruction conversion rate; Through the federated learning mechanism, global model parameters are updated while protecting user privacy.

[0011] As an improved solution, the active clarification process includes: The questioning strategy based on reinforcement learning prioritizes asking about the attributes with the highest information gain; The speech synthesis engine uses affective computing technology, and the speaking speed and pitch are adaptive to the user's emotional state; Allow users to interrupt the clarification process with gestures.

[0012] As an improved solution, the task dependency graph is constructed, including: Time constraint modeling: predicting subtask duration through LSTM; Resource constraint modeling: dynamically adjust task priorities based on device load; Conflict resolution mechanism: Use the minimum conflict heuristic algorithm to resolve resource contention.

[0013] As an improved solution, the digital twin verification includes: The device interaction process is simulated based on Unity3D's physical engine; Abnormal scenario injection testing: simulating network delay and sensor noise interference; Generative adversarial networks generate extreme test cases.

[0014] The beneficial effects of the technical solution of the present invention are: 1. The self-service terminal large-model task processing system described in the present invention can improve the task recognition accuracy based on multimodal fusion through the mutual cooperation of system modules; the dynamic demand matching engine supports processing a large number of concurrent requests per second with low response latency; the task orchestration middleware shortens the development cycle of complex business processes; the adaptive response generation module supports the automatic conversion of multiple natural languages to device instructions; digital twin verification reduces the failure rate of the system after it is launched; the federated learning mechanism realizes model parameter updates while reducing the risk of user privacy leakage; it supports horizontal expansion and meets the requirements of regulations such as GDPR and HIPAA, making up for the shortcomings of existing technologies and having high application value.

[0015] 2. The self-service terminal large model task processing method described in the present invention can call the system modules in an orderly manner, thereby realizing the system logic of the self-service terminal large model task processing system described in the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 This is a schematic diagram of the architecture of the self-service terminal large model task processing system according to Example 1 of the present invention; Figure 2 It is a flowchart of the self-service terminal large model task processing method described in Example 2 of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more precise definition of the protection scope of the present invention.

[0019] In the description of the present invention, it should be noted that the embodiments described in the present invention are only part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of the present invention.

[0020] The terms "first," "second," and the like in the specification and claims herein and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments of the present invention described herein can be implemented in orders other than those illustrated or described herein. In addition, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device. Example 1

[0021] This embodiment provides a self-service terminal large model task processing system, such as Figure 1 As shown, including: A multimodal sensing fusion unit, which integrates an infrared multi-touch screen, a MEMS microphone array, and an industrial-grade code scanning module; A multi-domain intent classification model based on the RoBERTa architecture, which includes a built-in medical diagnosis knowledge graph and a financial transaction ontology industry-specific graph; A dynamic demand matching engine, which uses Transformer-XL long sequence modeling technology to analyze cross-domain task links; Task scheduling middleware with built-in Petri net simulator and conflict detection algorithm to ensure the timing consistency of multiple subtask execution; An adaptive response generation module is used to support the automatic conversion of natural language instructions into PLC control codes.

[0022] As an improved solution, the multimodal perception fusion unit is also used to: achieve spatiotemporal alignment of the screen click area and the voice command based on the Yolov5 target detection algorithm; use the attention mechanism to dynamically allocate confidence weights for touch, voice, and code scanning; and eliminate abnormal operations based on the isolation forest algorithm.

[0023] As an improved solution, the intent classification model includes: Domain-adaptive pre-training module for fine-tuning using industry-specific corpora; Multi-hop reasoning mechanism for enhancing implicit demand understanding through knowledge graph retrieval; The real-time feedback correction module is used to dynamically adjust the initial intention determination results based on the user's subsequent operations.

[0024] As an improved solution, the dynamic demand matching engine is used to: use local sensitive hash index to accelerate similar demand retrieval; A resource scheduling strategy based on reinforcement learning optimizes the load balancing of CPU, memory, and GPU; a timing dependency parser is called to support time window constraints between tasks.

[0025] As an improved solution, the task orchestration middleware is also used to: support engineers in real-time debugging through a color-coded state machine visualization interface; automatically trigger compensation transactions; and utilize edge computing nodes to parallelize subtasks.

[0026] As an improved solution, the adaptive response generation module is also used to: cover the industrial control instruction set and API call instructions; verify the matching degree between instructions and requirements through the BERT model; and support the generation of voice broadcast and dynamic visual guidance composite feedback.

[0027] It should be noted that the examples herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention. Example 2

[0028] This embodiment is based on the same inventive concept as the self-service terminal large model task processing system described in Example 1, and provides a self-service terminal large model task processing method, such as Figure 2 As shown, the following steps are included: The heterogeneous data stream input by the user is collected through the multimodal perception fusion unit; Input the data into the multi-domain intent classification model and output the top-5 candidate intents and their confidence levels. When the highest confidence level is less than 0.9, the active clarification process is triggered: additional information is collected through multiple rounds of dialogue; Use the dynamic demand matching engine to build task dependency graphs and identify critical paths; The task orchestration middleware generates executable intermediate representation code; The adaptive response generation module converts IR codes into device control commands; The execution results are simulated and verified through the digital twin engine; Establish user behavior profiles and continuously optimize intent classification model parameters; Regularly generate task processing quality reports, including KPIs such as response time and instruction conversion rate; Through the federated learning mechanism, global model parameters are updated while protecting user privacy.

[0029] As an improved solution, the active clarification process includes: The questioning strategy based on reinforcement learning prioritizes asking about the attributes with the highest information gain; The speech synthesis engine uses affective computing technology, and the speaking speed and pitch are adaptive to the user's emotional state; Allow users to interrupt the clarification process with gestures.

[0030] As an improved solution, the task dependency graph is constructed, including: Time constraint modeling: predicting subtask duration through LSTM; Resource constraint modeling: dynamically adjust task priorities based on device load; Conflict resolution mechanism: Use the minimum conflict heuristic algorithm to resolve resource contention.

[0031] As an improved solution, the digital twin verification includes: The device interaction process is simulated based on Unity3D's physical engine; Abnormal scenario injection testing: simulating network delay and sensor noise interference; Generative adversarial networks generate extreme test cases.

[0032] It should be understood that in the various embodiments of this document, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0033] It should also be understood that in the embodiments herein, the term "and / or" merely describes an association between associated objects, indicating that three possible relationships exist. For example, "A and / or B" could represent: A alone, A and B simultaneously, or B alone. Furthermore, the character " / " in this document generally indicates an "or" relationship between the associated objects.

[0034] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the composition and steps of each example according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this document.

[0035] Those skilled in the art will clearly understand that, for the sake of convenience and brevity of description, the specific logical process of the method described above can refer to the corresponding working processes of the systems, devices and units in the aforementioned method embodiments, and will not be repeated here.

[0036] In the several embodiments provided herein, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices, or units, or can be an electrical, mechanical, or other form of connection.

[0037] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments herein.

[0038] In addition, the functional units in the various embodiments herein may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0039] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this article is essentially or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of this article. The aforementioned storage medium includes: various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0040] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A large-scale model task processing system for a self-service terminal, characterized in that: include: A multimodal sensing fusion unit, which integrates an infrared multi-touch screen, a MEMS microphone array, and an industrial-grade code scanning module; A multi-domain intent classification model based on the RoBERTa architecture, which includes a built-in medical diagnosis knowledge graph and a financial transaction ontology industry-specific graph; A dynamic demand matching engine, which uses Transformer-XL long sequence modeling technology to analyze cross-domain task links; Task scheduling middleware with built-in Petri net simulator and conflict detection algorithm to ensure the timing consistency of multiple subtask execution; An adaptive response generation module is used to support the automatic conversion of natural language instructions into PLC control codes.

2. The self-service terminal large model task processing system according to claim 1, characterized in that: The multimodal perception fusion unit is also used to: achieve spatiotemporal alignment of the screen click area and the voice command based on the Yolov5 target detection algorithm; dynamically allocate confidence weights for touch, voice, and code scanning using the attention mechanism; and eliminate abnormal operations based on the isolation forest algorithm.

3. The self-service terminal large model task processing system according to claim 1, characterized in that: The intent classification model includes: Domain-adaptive pre-training module for fine-tuning using industry-specific corpora; Multi-hop reasoning mechanism for enhancing implicit demand understanding through knowledge graph retrieval; The real-time feedback correction module is used to dynamically adjust the initial intention determination results based on the user's subsequent operations.

4. The self-service terminal large model task processing system according to claim 1, characterized in that: The dynamic demand matching engine is used to: use local sensitive hash index to accelerate similar demand retrieval; A resource scheduling strategy based on reinforcement learning optimizes the load balancing of CPU, memory, and GPU; a timing dependency parser is called to support time window constraints between tasks.

5. The self-service terminal large model task processing system according to claim 1, characterized in that: The task orchestration middleware is also used to: support engineers in real-time debugging through a color-coded state machine visualization interface; automatically trigger compensation transactions; and utilize edge computing nodes to parallelize subtasks.

6. The self-service terminal large model task processing system according to claim 1, characterized in that: The adaptive response generation module is also used to: cover industrial control instruction sets and API call instructions; verify the matching degree between instructions and requirements through the BERT model; and support the generation of voice broadcasts and dynamic visual guidance composite feedback.

7. A self-service terminal large model task processing method, characterized in that: The following steps are involved: The heterogeneous data stream input by the user is collected through the multimodal perception fusion unit; Input the data into the multi-domain intent classification model and output the top-5 candidate intents and their confidence levels. When the highest confidence level is less than 0.9, the active clarification process is triggered: additional information is collected through multiple rounds of dialogue; Use the dynamic demand matching engine to build task dependency graphs and identify critical paths; The task orchestration middleware generates executable intermediate representation code; The adaptive response generation module converts IR codes into device control commands; The execution results are simulated and verified through the digital twin engine; Establish user behavior profiles and continuously optimize intent classification model parameters; Regularly generate task processing quality reports, including KPIs such as response time and instruction conversion rate; Through the federated learning mechanism, global model parameters are updated while protecting user privacy.

8. The self-service terminal large model task processing method according to claim 7, characterized in that: The proactive clarification process includes: The questioning strategy based on reinforcement learning prioritizes asking about the attributes with the highest information gain; The speech synthesis engine uses affective computing technology, and the speaking speed and pitch are adaptive to the user's emotional state; Allow users to interrupt the clarification process with gestures.

9. The self-service terminal large model task processing method according to claim 7, characterized in that: The task relies on graph construction, including: Time constraint modeling: predicting subtask duration through LSTM; Resource constraint modeling: dynamically adjust task priorities based on device load; Conflict resolution mechanism: Use the minimum conflict heuristic algorithm to resolve resource contention.

10. The self-service terminal large model task processing method according to claim 7, characterized in that: The digital twin verification includes: The device interaction process is simulated based on Unity3D's physical engine; Abnormal scenario injection testing: simulating network delay and sensor noise interference; Generative adversarial networks generate extreme test cases.

Images