Automated question answering method, system, device and medium for device support

Abstract

This invention relates to the field of information processing, specifically to an automatic question-answering method, system, device, and medium for device support. The method includes: responding to received input information, initiating a current session; performing multimodal parsing on the input information to extract session parameters; determining, based on service credentials, whether the current session meets the conditions for automatic service handover; if so, generating a remote request ticket based on the session parameters to initiate a manual service process; if not, retrieving matching knowledge units from a pre-built knowledge base based on target device information and dialogue intent, as a response to the input information. This invention enables knowledge retrieval and matching based on precise device context and genuine user intent, thereby solving the problems of intent comprehension bias and inaccurate answers caused by relying on superficial keyword matching in existing technologies, significantly improving the accuracy of identifying and resolving complex practical problems.

Description

Technical Field

[0001] This invention relates to the field of information processing, and more specifically to an automatic question-and-answer method, system, device, and medium for device support. Background Technology

[0002] In specific areas of medical equipment technical support, the widespread adoption and iteration of digital imaging equipment such as dental CBCTs have significantly increased the complexity of equipment operation, the diversity of failure modes, and the professionalism of maintenance. Traditional technical support relies heavily on human operators, which not only limits response efficiency but also poses a severe challenge to engineers' experience and knowledge. Automated question-and-answer systems, as a key application of artificial intelligence technology in this scenario, aim to automatically understand and respond to users' natural language questions by simulating human dialogue, thereby providing immediate and accurate operational guidance, troubleshooting solutions, or maintenance suggestions. The core value of such systems lies in their ability to provide uninterrupted service 24 / 7, significantly reducing labor costs, improving service response speed and accessibility, and promoting the efficient reuse and transfer of knowledge. They are an important technical means to improve the quality of after-sales service for medical equipment and ensure the continuity of clinical diagnosis and treatment activities.

[0003] In existing technologies, automated question-answering systems in the equipment support field mostly rely on keyword-based retrieval mechanisms. These systems typically compare the user's natural language query with pre-defined knowledge base entries using keywords and push pre-defined answers with high matching scores. However, keyword matching lacks a deep semantic understanding of the user's true intent behind the query and cannot effectively incorporate the contextual information of the current dialogue. This leads to the system often mechanically matching local words in the user's statement, failing to accurately understand the complex, implicit, or context-dependent nature of the user's specific problem. Consequently, the pushed answers are often disconnected from the user's actual problem situation and cannot accurately solve the specific equipment operation or malfunction problems encountered by customers. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides an automatic question-and-answer method, system, device, and medium for device support.

[0005] The first aspect of this invention discloses an automatic question-answering method for device support, comprising: In response to the received input information, the current session is started, the input information is parsed in a multimodal manner, and session parameters are extracted, including target device information, dialogue intent and service credentials; Based on the service credentials, determine whether the current session meets the automatic service handover conditions; If satisfied, a remote request ticket is generated based on the session parameters to initiate the manual service process; If the conditions are not met, then based on the target device information and the dialogue intent, a matching knowledge unit is retrieved from the pre-built knowledge base as a response to the input information.

[0006] Furthermore, the knowledge base is pre-built according to the following steps: Obtain equipment manuals and historical work order data files as raw knowledge data, and retain their original format; The organization structure of the original knowledge data is analyzed to obtain multiple knowledge fragments, and the logical relationships between the knowledge fragments are extracted and identified. Based on the logical relationship, the knowledge fragments are structured to form knowledge units with related relationships; The knowledge units are semantically vectorized to generate their vectorized representations; The knowledge units and their vectorized representations are associated according to the logical relationship, and stored in different knowledge partitions according to the device model to which the knowledge units belong, thus completing the construction of the knowledge base.

[0007] Furthermore, the step of performing multimodal parsing on the input information to extract session parameters includes: Identify the data type of the input information; If an image type is identified, the image data is processed using a visual analysis model to identify the image scene category, and the target device information or the service credential is extracted from the image based on the image scene category. If text data is identified, semantic analysis is performed on the text data to extract the target device information and / or the dialogue intent.

[0008] Furthermore, the step of performing semantic analysis on the text data to extract the dialogue intent includes: Map the text data and historical work order data to a high-dimensional space; Within the high-dimensional space, the dialogue intent is determined by comparing the distance between the text data and the dialogue intent in the historical work order data.

[0009] Furthermore, based on the service credentials, the step of determining whether the current session meets the automatic service handover conditions includes: Perform integrity and validity checks on the service credentials; If all verifications pass, the current session is determined to meet the automatic service handover conditions. Otherwise, the current session is determined not to meet the automatic service handover conditions.

[0010] Furthermore, the step of retrieving matching knowledge units from a pre-built knowledge base based on the target device information and the dialogue intent, and using these units as a response to the input information, includes: Based on the device model in the target device information, locate the corresponding target knowledge partition from the knowledge base; Within the target knowledge partition, a hybrid retrieval is performed based on the dialogue intent, and the retrieval results are reordered and filtered for relevance in order to obtain matching knowledge units; The matched knowledge units are combined with the dialogue intent to generate a response to the input information.

[0011] Furthermore, the steps of performing a hybrid retrieval based on the stated dialogue intent, and reordering and filtering the retrieval results according to relevance to obtain matching knowledge units include: Within the target knowledge partition, semantic similarity-based matching retrieval and keyword-based matching retrieval are performed simultaneously, and the retrieval results of the two are fused to obtain a preliminary retrieval result set; For each knowledge unit in the preliminary search result set, calculate its relevance score to the dialogue intent; Based on a preset relevance threshold and the relevance score, the knowledge units in the preliminary search result set are filtered to obtain matching knowledge units.

[0012] A second aspect of the present invention discloses an automatic question-answering system for device support, comprising: The parsing module is used to respond to received input information, start the current session, perform multimodal parsing on the input information, and extract session parameters, which include target device information, dialogue intent, and service credentials. The judgment module is used to determine, based on the service credentials, whether the current session meets the automatic service handover conditions; The generation module is used to generate a remote request work order to start the manual service process based on the session parameters when the judgment result of the judgment module is satisfied. The retrieval module, when the judgment result of the judgment module is not satisfied, retrieves a matching knowledge unit from the pre-built knowledge base based on the target device information and the dialogue intent, as a response to the input information.

[0013] A third aspect of the present invention discloses an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of any of the automatic question-answering methods for device support disclosed in the first aspect of the present invention.

[0014] The fourth aspect of the present invention discloses a storage medium storing a computer program that, when executed by a processor, implements the steps of any of the device-supported automatic question-answering methods disclosed in the first aspect of the present invention.

[0015] This invention introduces a session parameter extraction mechanism based on multimodal parsing, which transforms user input into a structured semantic representation containing target device information and dialogue intent. This enables the system to perform knowledge retrieval and matching based on accurate device context and real user intent, thereby solving the problems of intent comprehension bias and inaccurate answers caused by relying on keyword surface matching in the prior art. This significantly improves the accuracy of identifying and solving complex practical problems. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating an automatic question-and-answer method for device support disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an automatic question-and-answer system for device support disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device disclosed in the embodiments of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, or product comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, apparatus, or products.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] Please see Figure 1 As shown, Figure 1 This is a flowchart illustrating an automatic question-and-answer method for device support disclosed in an embodiment of the present invention. Figure 1 As shown, the automated question-answering method for device support may include the following operations: S101. In response to the received input information, start the current session, perform multimodal parsing on the input information, and extract session parameters, the session parameters including target device information, dialogue intent and service credentials; In an optional embodiment, the step of performing multimodal parsing on the input information to extract session parameters includes: Identify the data type of the input information; If an image type is identified, the image data is processed using a visual analysis model to identify the image scene category, and the target device information or the service credential is extracted from the image based on the image scene category. If text data is identified, semantic analysis is performed on the text data to extract the target device information and / or the dialogue intent.

[0022] In this invention, the parsing of input information and the construction of responses are both based on the Big Prophet model.

[0023] In this optional embodiment, the data type includes image types and / or text data.

[0024] For image types, a visual analysis model is used to analyze the image content. If features containing specifications or product nameplates are detected, an optical character recognition task is further performed to extract key identifier strings. The non-standard strings identified are then standardized into preset standard key values ​​through regular expression matching, thereby completing the extraction of the target device information. If features containing remote control software interfaces are detected, the visual analysis model is configured to output structured data containing specific remote connection identifiers and remote authorization credentials to complete the extraction of the service credentials.

[0025] As can be seen, this optional embodiment achieves adaptive parsing of multimodal information by identifying and processing the data types of input information separately. For image data, a visual analysis model is used to identify the scene and extract key information, enabling the system to understand and utilize non-textual information such as physical images of equipment, screenshots of faults, or photos of vouchers, thus expanding the channels for information input. For text data, semantic analysis is performed to extract core elements. This multimodal parsing mechanism ensures that the extraction of session parameters is not limited to textual descriptions but can also comprehensively utilize visual information for supplementation and verification. Especially when users find it difficult to accurately describe the equipment model or fault phenomenon in words, they can directly obtain target equipment information or service vouchers through images, thereby significantly enhancing the system's adaptability in complex real-world scenarios and the comprehensiveness and accuracy of parameter extraction.

[0026] In an optional embodiment, the step of performing semantic analysis on the text data to extract the dialogue intent includes: Map the text data and historical work order data to a high-dimensional space; Within the high-dimensional space, the dialogue intent is determined by comparing the distance between the text data and the dialogue intent in the historical work order data.

[0027] In this optional embodiment, embedding both text and historical work order data into the same high-dimensional semantic vector space allows semantically similar text vectors to be positioned close to each other within that space. Specific steps include: First, using a pre-trained language model, converting the text descriptions in the text data to be analyzed and the historical work order data into high-dimensional vector representations, completing the mapping from text to a high-dimensional space. Then, within this high-dimensional space, calculating the semantic distance between the vector of the current text data and the vectors of each historical work order data point. Finally, by comparing these distances, identifying one or more historical work order data samples that are closest to the current text data vector, and determining the dialogue intent corresponding to these samples as the intent of the current text data, or using it as a basis for intent classification. This method effectively captures the deep semantic features of text, improving the accuracy of intent recognition and robustness to diverse expressions. In this optional embodiment, the language model is such as BERT, the dialogue intent in the historical work order data is already labeled, and the semantic distance in the high-dimensional space can be calculated using cosine similarity.

[0028] In this optional embodiment, if the digital areas of an image are obscured, smeared, or displayed as masked characters, information is not extracted from the image.

[0029] As can be seen, this optional embodiment maps current text data and historical work order data together into a high-dimensional space, and determines the dialogue intent by calculating distance within this space, thus realizing an intent recognition method based on distributed semantics. This optional embodiment does not perform simple keyword matching, but rather utilizes vector positions in the high-dimensional space to represent the deep semantics of the text. By comparing the distance with the intents already labeled in historical work orders, the system can more accurately capture the semantic similarity between the user's current query and similar historical cases, thereby inferring potentially unstated deep needs or intent categories. This enhances the understanding of the diversity of user expressions, improves the accuracy of extracting dialogue intent from text, and enhances the ability to learn from and utilize historical experience data.

[0030] 102. Based on the service credential, determine whether the current session meets the automatic service handover conditions; In an optional embodiment, the step of determining whether the current session meets the automatic service handover conditions based on the service credentials includes: Perform integrity and validity checks on the service credentials; If all verifications pass, the current session is determined to meet the automatic service handover conditions. Otherwise, the current session is determined not to meet the automatic service handover conditions.

[0031] As can be seen, this optional embodiment establishes a clear and reliable decision-making logic for determining whether to transfer services to human intervention by performing two progressive steps: integrity verification and validity verification of service credentials. Integrity verification ensures that all credential information required for the judgment is available, avoiding misjudgments due to missing information; validity verification further verifies the authenticity and timeliness of the credentials. This step-by-step verification mechanism ensures that the system only determines that the transfer conditions are met when the credentials are complete and valid. This reliably identifies scenarios requiring human intervention, improving the accuracy and reliability of automated service transfer decisions, while also ensuring the effective operation of the automated question-and-answer service within its authorized scope.

[0032] S103. If satisfied, a remote request work order is generated based on the session parameters to initiate the manual service process. S104. If not satisfied, then based on the target device information and the dialogue intent, retrieve matching knowledge units from the pre-built knowledge base as a response to the input information.

[0033] In an optional embodiment, the step of retrieving matching knowledge units from a pre-built knowledge base based on the target device information and the dialogue intent, and using these units as a response to the input information, includes: Based on the device model in the target device information, locate the corresponding target knowledge partition from the knowledge base; Within the target knowledge partition, a hybrid retrieval is performed based on the dialogue intent, and the retrieval results are reordered and filtered for relevance in order to obtain matching knowledge units; The matched knowledge units are combined with the dialogue intent to generate a response to the input information.

[0034] In this optional embodiment, if the device model in the extracted target device information is an incomplete prefix, for example, the user only entered the series code, and there are multiple sub-models with that code as a prefix in the knowledge base, then all possible complete models are listed for the user to select, and the user is waited for input until the complete device model is obtained.

[0035] As can be seen, this optional embodiment locates knowledge partitions based on device model, rapidly converging the search scope from the entire knowledge base to the subset most relevant to the current device, significantly improving search efficiency and reducing interference from irrelevant information; the hybrid search within the target partition can find matching content from different dimensions; subsequently, the search results are reordered and filtered for relevance, ensuring that the returned knowledge units are not only relevant but also of high quality; the selected matching knowledge units are combined with the dialogue intent to generate a response, so that the final response not only contains factual knowledge but is also organized and expressed in accordance with the user's intent, significantly improving the targeting of knowledge retrieval and the matching degree and practicality of the response content.

[0036] In an optional embodiment, the steps of performing a hybrid retrieval based on the dialogue intent and reordering and filtering the retrieval results according to relevance to obtain matching knowledge units include: Within the target knowledge partition, semantic similarity-based matching retrieval and keyword-based matching retrieval are performed simultaneously, and the retrieval results of the two are fused to obtain a preliminary retrieval result set; For each knowledge unit in the preliminary search result set, calculate its relevance score to the dialogue intent; Based on a preset relevance threshold and the relevance score, the knowledge units in the preliminary search result set are filtered to obtain matching knowledge units.

[0037] In this optional embodiment, semantic similarity matching is achieved by calculating the cosine similarity between the semantic vector representation of the user's dialogue intent and the vector representations of various knowledge units in the knowledge base; keyword matching utilizes an inverted index to precisely match the core terms in the intent description. The results of these two retrieval methods are then merged, for example, by weighting and combining them according to predefined weights, to form a preliminary retrieval result set.

[0038] Subsequently, based on the Rerank mechanism, a cross-encoder is used to recalculate the relevance score of each knowledge unit in the initial result set to the dialogue intent. This score, compared to the similarity score in the initial retrieval, can more accurately measure the value of the knowledge unit in answering the user's actual question.

[0039] Finally, based on a confidence threshold preset by analyzing historical question-and-answer data or business needs, the re-ranked results are filtered, retaining only highly relevant knowledge units with relevance scores above the threshold. This ensures that the knowledge fragments ultimately recalled and used to generate responses are highly accurate and targeted.

[0040] As can be seen, this optional embodiment forms a complementary hybrid retrieval strategy by simultaneously executing semantic similarity-based matching retrieval and keyword-based matching retrieval. Semantic similarity retrieval can understand the deeper meaning behind the user's intent, matching knowledge with different expressions but similar semantics; keyword retrieval can accurately capture specific terms or model codes explicitly mentioned by the user. Fusing the results of both yields a more comprehensive preliminary result set. Subsequently, by calculating the relevance score of each knowledge unit to the dialogue intent and applying threshold filtering, a refined selection of the preliminary results is achieved, effectively excluding knowledge units that, although retrieved, have low actual relevance, significantly improving the comprehensiveness of the knowledge retrieval process and the relevance and purity of the final result set.

[0041] In an optional embodiment, the knowledge base is pre-built according to the following steps: Obtain equipment manuals and historical work order data files as raw knowledge data, and retain their original format; The organization structure of the original knowledge data is analyzed to obtain multiple knowledge fragments, and the logical relationships between the knowledge fragments are extracted and identified. Based on the logical relationship, the knowledge fragments are structured to form knowledge units with related relationships; The knowledge units are semantically vectorized to generate their vectorized representations; The knowledge units and their vectorized representations are associated according to the logical relationship, and stored in different knowledge partitions according to the device model to which the knowledge units belong, thus completing the construction of the knowledge base.

[0042] In this optional embodiment, the format of the original knowledge data may include text files, structured documents, scanned documents, images, or combinations thereof. To handle different formats and preserve the original content and structure, this optional embodiment parses and transforms the original knowledge data. For example, it uses the Minuer parsing tool to convert the original knowledge data into an intermediate format, such as Markdown, that preserves the structure of images, tables, and paragraphs to the greatest extent possible. Based on an understanding of the document's organizational structure, the transformed content is extracted and segmented according to preset rules. This process employs a hierarchical approach, such as identifying and establishing logical relationships between chapter-level or function-level "parent" topic blocks and "child" content blocks such as specific operation steps and parameter descriptions, thereby parsing the unstructured original data into multiple knowledge fragments with clear logical connections. By structurally reorganizing and encapsulating these knowledge fragments, the related knowledge units are finally formed.

[0043] In this optional embodiment, non-image-based knowledge units can be stored using the Dify platform, while image-based knowledge units can be stored using the MinIO object storage service built into the Dify platform.

[0044] As can be seen, this optional embodiment effectively avoids information loss or structural damage caused by format conversion during the data preprocessing stage by acquiring and retaining the original format of equipment manuals and historical work order data as raw knowledge data, thus preserving the complete context for subsequent processing. By parsing the organizational structure and extracting the logical relationships between knowledge fragments, the originally unstructured document content is transformed into a knowledge network with clear relationships, enhancing the structure and internal coherence of the knowledge. Further semantic vectorization and associated storage of knowledge units not only lays the foundation for subsequent semantic retrieval, but also greatly improves the efficiency and accuracy of locating target knowledge during retrieval by adopting a knowledge partitioning storage strategy based on equipment model, thereby systematically improving the completeness, usability, and retrieval accuracy of the knowledge base.

[0045] Please see Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of an automatic question-answering system for device support disclosed in an embodiment of the present invention, including: The parsing module 201 is used to respond to the received input information, start the current session, perform multimodal parsing on the input information, and extract session parameters, the session parameters including target device information, dialogue intent and service credentials; The judgment module 202 is used to determine whether the current session meets the automatic service handover conditions based on the service credential; The generation module 203 is used to generate a remote request work order to start the manual service process based on the session parameters when the judgment result of the judgment module is satisfied. The retrieval module 204, when the judgment result of the judgment module is not satisfied, retrieves a matching knowledge unit from the pre-built knowledge base based on the target device information and the dialogue intent, as a response to the input information.

[0046] Specific limitations regarding the automated question-and-answer system for device support can be found in the limitations of the automated question-and-answer method for device support described above, and will not be repeated here. Each module in the aforementioned automated question-and-answer system for device support can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware format within or independently of the processor in the electronic device, or stored in software format in the memory of the electronic device, so that the processor can call the corresponding operations of each module.

[0047] It should be noted that, in order to highlight the innovative aspects of this invention, this embodiment does not include modules that are not closely related to solving the technical problems proposed by this invention, but this does not mean that there are no other modules in this embodiment.

[0048] like Figure 3 As shown, the electronic device 1 provided by the present invention may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and executable on the processor 13, such as an automatic question-and-answer program for device support.

[0049] The memory 12 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 12 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 1. Furthermore, the memory 12 can include both internal and external storage units of the electronic device 1. The memory 12 can be used not only to store application software and various types of data installed on the electronic device 1, such as code for automatic question-and-answer support, but also to temporarily store data that has been output or will be output.

[0050] In some embodiments, the processor 13 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 13 is the control unit of the electronic device 1, connecting various components of the electronic device 1 via various interfaces and lines. It executes programs or modules stored in the memory 12 (e.g., automatic question-and-answer programs for device support) and calls data stored in the memory 12 to perform various functions of the electronic device 1 and process data.

[0051] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-described automatic question-and-answer method for device support.

[0052] For example, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into a parsing module 201, a judging module 202, a generating module 203, and a retrieving module 204.

[0053] The integrated unit implemented as a software functional module described above can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The software functional module stored in the storage medium includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute some of the functions of the automatic question-answering method for device support described in the various embodiments of this application.

[0054] In summary, the automatic question-answering method, system, device, and medium disclosed in this invention, by introducing a session parameter extraction mechanism based on multimodal parsing, transforms user input into a structured semantic representation containing target device information and dialogue intent. This enables the system to perform knowledge retrieval and matching based on accurate device context and genuine user intent, thereby solving the problems of intent comprehension bias and inaccurate answers caused by relying on keyword surface matching in existing technologies. This significantly improves the accuracy of identifying and solving complex practical problems. Therefore, this invention effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0055] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. An automatic question-and-answer method for device support, characterized in that, The method includes: In response to the received input information, the current session is started, the input information is parsed in a multimodal manner, and session parameters are extracted, including target device information, dialogue intent and service credentials; Based on the service credentials, determine whether the current session meets the automatic service handover conditions; If satisfied, a remote request ticket is generated based on the session parameters to initiate the manual service process; If the conditions are not met, then based on the target device information and the dialogue intent, a matching knowledge unit is retrieved from the pre-built knowledge base as a response to the input information.

2. The automatic question-and-answer method for device support according to claim 1, characterized in that, The knowledge base is pre-built according to the following steps: Obtain equipment manuals and historical work order data files as raw knowledge data, and retain their original format; The organization structure of the original knowledge data is analyzed to obtain multiple knowledge fragments, and the logical relationships between the knowledge fragments are extracted and identified. Based on the logical relationship, the knowledge fragments are structured to form knowledge units with related relationships; The knowledge units are semantically vectorized to generate their vectorized representations; The knowledge units and their vectorized representations are associated according to the logical relationship, and stored in different knowledge partitions according to the device model to which the knowledge units belong, thus completing the construction of the knowledge base.

3. The automatic question-and-answer method for device support according to claim 1, characterized in that, The steps of performing multimodal parsing on the input information and extracting session parameters include: Identify the data type of the input information; If an image type is identified, the image data is processed using a visual analysis model to identify the image scene category, and the target device information or the service credential is extracted from the image based on the image scene category. If text data is identified, semantic analysis is performed on the text data to extract the target device information and / or the dialogue intent.

4. The automatic question-and-answer method for device support according to claim 3, characterized in that, The steps of performing semantic analysis on the text data to extract the dialogue intent include: Map the text data and historical work order data to a high-dimensional space; Within the high-dimensional space, the dialogue intent is determined by comparing the distance between the text data and the dialogue intent in the historical work order data.

5. The automatic question-and-answer method for device support according to claim 1, characterized in that, Based on the service credentials, the steps for determining whether the current session meets the automatic service handover conditions include: Perform integrity and validity checks on the service credentials; If all verifications pass, the current session is determined to meet the automatic service handover conditions. Otherwise, the current session is determined not to meet the automatic service handover conditions.

6. The automatic question-and-answer method for device support according to claim 1, characterized in that, Based on the target device information and the dialogue intent, the step of retrieving matching knowledge units from a pre-built knowledge base as a response to the input information includes: Based on the device model in the target device information, locate the corresponding target knowledge partition from the knowledge base; Within the target knowledge partition, a hybrid retrieval is performed based on the dialogue intent, and the retrieval results are reordered and filtered for relevance in order to obtain matching knowledge units; The matched knowledge units are combined with the dialogue intent to generate a response to the input information.

7. The automatic question-and-answer method for device support according to claim 6, characterized in that, The steps of performing a hybrid retrieval based on the stated dialogue intent, and reordering and filtering the retrieval results according to relevance to obtain matching knowledge units include: Within the target knowledge partition, semantic similarity-based matching retrieval and keyword-based matching retrieval are performed simultaneously, and the retrieval results of the two are fused to obtain a preliminary retrieval result set; For each knowledge unit in the preliminary search result set, calculate its relevance score to the dialogue intent; Based on a preset relevance threshold and the relevance score, the knowledge units in the preliminary search result set are filtered to obtain matching knowledge units.

8. An automatic question-and-answer system for device support, characterized in that, include: The parsing module is used to respond to received input information, start the current session, perform multimodal parsing on the input information, and extract session parameters, which include target device information, dialogue intent, and service credentials. The judgment module is used to determine, based on the service credentials, whether the current session meets the automatic service handover conditions; The generation module is used to generate a remote request work order to start the manual service process based on the session parameters when the judgment result of the judgment module is satisfied. The retrieval module, when the judgment result of the judgment module is not satisfied, retrieves a matching knowledge unit from the pre-built knowledge base based on the target device information and the dialogue intent, as a response to the input information.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the automatic question-answering method for device support as described in any one of claims 1 to 7.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the automatic question-answering method for device support as described in any one of claims 1 to 7.

Images