Guided knowledge question answering method and device, electronic equipment, product and storage medium

Abstract

The invention provides a guided knowledge question answering method and device, electronic equipment, a product and a storage medium, and relates to the technical field of knowledge question answering. The method comprises the steps that based on an initial question text input by a user, question elements corresponding to the initial question text are determined in a preset civil aviation question element set; determining question missing elements based on a matching result between the keyword information of the initial question text and question elements corresponding to the initial question text; based on the question missing elements, performing guiding type reverse questioning on the user to determine a complete question text; and based on the complete question text, determining an answer corresponding to the complete question text in a preset civil aviation question and answer data set. According to the invention, when a user encounters a difficult problem, the user can obtain a high-credibility solution by simply asking questions.

Description

Technical Field

[0001] The present invention relates to the field of knowledge question answering technology, and in particular to a guided knowledge question answering method, device, electronic device, product and storage medium. Background Art

[0002] In the field of intelligent customer service and information retrieval, users often enter questions or inquire about information in order to receive answers from intelligent customer service. In the highly specialized field of civil aviation, in particular, the numerous technical terms and complex concepts lead to a diverse range of question types, and user questions may involve multiple scenarios and specialized knowledge. Therefore, intelligent customer service must accurately identify the user's intent based on the question they enter in order to provide accurate and efficient answers. However, user-entered text often contains incomplete information and unclear expressions, making it difficult for intelligent customer service to understand user needs. This can lead to answers that deviate from the user's expectations and fail to effectively resolve the user's problem. Therefore, a guided knowledge Q&A method is urgently needed to enable users to obtain highly reliable solutions by simply asking simple questions when faced with difficult problems. Summary of the Invention

[0003] The present invention provides a guided knowledge question-and-answer method, device, electronic device, product and storage medium to address the defects in the prior art that intelligent customer service may encounter difficulties in understanding user needs, resulting in the answers it provides deviating from user expectations and failing to effectively solve user problems.

[0004] The present invention provides a guided knowledge question-answering method, comprising: Based on the initial question text input by the user, determining question elements corresponding to the initial question text from a preset civil aviation question element set; Determining missing elements of the question based on a matching result between the keyword information of the initial question text and the question elements corresponding to the initial question text; Conducting guided counter-questions to the user based on the missing elements of the question to determine the complete question text; Based on the complete question text, an answer corresponding to the complete question text is determined in a preset civil aviation question and answer dataset.

[0005] According to a guided knowledge question-answering method provided by the present invention, the method of determining, based on an initial question text input by a user, question elements corresponding to the initial question text from a preset civil aviation question element set includes: Determining keyword information of the initial question text input by the user, and determining a civil aviation service scenario corresponding to the keyword information in a preset civil aviation service scenario set; Based on the civil aviation service scenario, determining a problem element corresponding to the civil aviation service scenario from a preset civil aviation problem element set; Among them, each civil aviation service scenario in the preset civil aviation service scenario set corresponds one-to-one to each problem element in the preset civil aviation problem element set.

[0006] According to a guided knowledge question-answering method provided by the present invention, the question elements include a plurality of keyword slot information, and the determining of missing question elements based on a matching result between the keyword information of the initial question text and the question elements corresponding to the initial question text includes: Matching the keyword information of the initial question text with the keyword slot information in the question elements, and determining the keyword slot information that has not been successfully matched; The unsuccessfully matched keyword slot information is determined as a problem missing element.

[0007] According to a guided knowledge question-answering method provided by the present invention, the step of conducting guided counter-questions to the user based on missing elements of the question to determine a complete question text includes: Conducting guided counter-questions to the user based on the missing elements of the question to obtain a supplementary question text input by the user; If the keyword information of the supplementary question text successfully matches the missing elements of the question, a complete question text is determined based on the supplementary question text and the initial question text.

[0008] According to a guided knowledge question-answering method provided by the present invention, determining an answer corresponding to the complete question text in a preset civil aviation question-answering dataset based on the complete question text includes: Vectorizing the complete question text and the preset civil aviation question and answer dataset to determine a question text vector for the complete question text and a civil aviation question and answer data vector set of the preset civil aviation question and answer dataset; Determine a similarity calculation result between the question text vector and each civil aviation question and answer data vector in the civil aviation question and answer data vector set; Based on the similarity calculation result with the greatest similarity among the similarity calculation results, determining target civil aviation question and answer data similar to the complete question text from the preset civil aviation question and answer dataset; splicing the complete question text with the target civil aviation question and answer data to obtain a target input text; The target input text is input into a civil aviation knowledge question-answering model to obtain an answer corresponding to the complete question text output by the civil aviation knowledge question-answering model; the civil aviation knowledge question-answering model is constructed based on a large language model.

[0009] According to a guided knowledge question-answering method provided by the present invention, after determining an answer corresponding to the complete question text in a preset civil aviation question-answering dataset based on the complete question text, the method further includes: Acquiring conversation data between the user and the civil aviation knowledge question-answering model; Identifying sensitive information in the conversation data; The sensitive information is desensitized to determine the desensitized conversation data.

[0010] The present invention also provides a guided knowledge question-answering device, comprising: A question element determination module is used to determine, based on the initial question text input by the user, question elements corresponding to the initial question text from a preset civil aviation question element set; a question missing element determination module, configured to determine the question missing element based on a matching result between the keyword information of the initial question text and the question elements corresponding to the initial question text; A guided questioning module, configured to conduct guided questioning of the user based on missing elements of the question to determine a complete question text; An answer determination module is used to determine an answer corresponding to the complete question text in a preset civil aviation question and answer dataset based on the complete question text.

[0011] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the guided knowledge question-answering method as described above is implemented.

[0012] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the guided knowledge question-answering methods described above.

[0013] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-described guided knowledge question-answering methods.

[0014] The guided knowledge question-answering method, device, electronic device, product, and storage medium provided by the present invention determine, based on an initial question text input by a user, question elements corresponding to the initial question text from a preset civil aviation question element set; determine missing question elements based on a match between keyword information in the initial question text and the question elements corresponding to the initial question text; and conduct guided counter-questions to the user based on the missing question elements to determine a complete question text. Based on the complete question text, the answer corresponding to the complete question text is determined in a preset civil aviation question and answer dataset. The present invention matches the keyword information in the initial question text input by the user with the corresponding question elements at the semantic level. If the keyword information is found to be not completely matched, the specific missing part is clearly located, that is, the missing element of the question is determined. Based on the missing element of the question, targeted guiding questions are formed to prompt the user to supplement key information. This guided question design enhances the interactivity between the system and the user and improves the integrity of information collection. Through multiple rounds of interaction, the user's complete intention is gradually collected to determine the complete question text. The complete question text is used as input, and the most relevant answer is searched based on the complete question text in the preset civil aviation question and answer dataset. This gradual improvement method based on question elements avoids misjudgment caused by incomplete information at a single time, thereby improving the accuracy of the answer. When users encounter difficult problems, they can obtain highly reliable solutions by simply asking simple questions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the 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.

[0016] Figure 1 It is a flow chart of the guided knowledge question-answering method provided by the present invention.

[0017] Figure 2 It is a flow chart of determining target civil aviation question and answer data in the guided knowledge question and answer method provided by the present invention.

[0018] Figure 3 It is a structural diagram of the guided knowledge question-answering device provided by the present invention.

[0019] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0021] After research and analysis, the applicants of this patent have discovered that in civil aviation customer service systems, questions raised by users often involve multiple scenarios and complex operational contexts, such as reporting an issue, booking a ticket, and refunding a ticket. However, due to the randomness and incompleteness of user input, customer service systems struggle to accurately determine user intent, resulting in inaccurate or irrelevant answers. In particular, when questions lack critical information, existing question-and-answer systems struggle to accurately supplement and understand users' true needs, often providing only vague or inappropriate answers based on partial information.

[0022] Further research shows that existing knowledge question-answering technologies often rely on simple keyword matching or static question-answering models based on single-round interactions when handling incomplete user questions. While this approach can provide basic answers to certain clear questions, it still has some shortcomings when faced with complex scenarios or multi-factor questions. First, existing systems lack effective strategies to dynamically detect and guide users to supplement missing key information. When a user's question lacks core elements, existing systems are unable to intelligently generate counter-questions to obtain complete information, which directly affects the accuracy of the answer. Second, existing systems are mostly designed for single-interaction modes. When information is incomplete or user feedback is unclear, the system lacks a multi-round dialogue mechanism to continuously collect key information from the user, resulting in a loss of contextual continuity across multiple rounds of interaction. Furthermore, technologies that rely solely on keyword matching ignore the semantic context and scenario characteristics of the question. This can easily lead to the system misjudging the user's intent, especially in situations involving multiple steps or multiple error codes.

[0023] In view of the above problems, the present invention proposes the following embodiments.

[0024] Figure 1 It is a flow chart of the guided knowledge question-answering method provided by the present invention. Figure 1 As shown, the method includes the following: Step 110 : Based on the initial question text input by the user, determine question elements corresponding to the initial question text in a preset civil aviation question element set.

[0025] Here, the initial question text refers to the original question initially entered by the user, without guidance or supplementation. The pre-set civil aviation question element set is a pre-set knowledge set that includes all key elements required for different civil aviation customer service scenarios. Each scenario has corresponding question elements. For example, the question elements required for a fault reporting scenario include the platform system, operation type, and error code. These elements provide the foundational information required to resolve the user's question and determine the complete intent.

[0026] Regarding the specific steps of how to determine the question elements corresponding to the initial question text in the preset civil aviation question element set, reference may be made to the following embodiments, which will not be elaborated herein.

[0027] Step 120 : Determine missing elements of the question based on a matching result between the keyword information of the initial question text and the question elements corresponding to the initial question text.

[0028] Here, keyword information refers to keywords or key information extracted from the initial question text input by the user through natural language processing technology.

[0029] Here, question elements refer to the key information required to accurately solve the problems raised by users; Missing elements in a question refer to key information that is not provided by the user in the initial question text but is necessary to solve the problem.

[0030] For example, in a fault reporting scenario, the problem elements include "platform system", "operation steps", and "error code"; but the initial problem text does not provide all the key information, such as the lack of "platform system" or "error code", and these unprovided elements are called missing problem elements.

[0031] Step 130 : Conduct guided counter-questions to the user based on the missing elements of the question to determine a complete question text.

[0032] Here, the complete question text refers to the final question statement after the user's initial question text is supplemented with all necessary information through guided questioning.

[0033] It should be noted that the design of guided rhetorical questions is based on dialogue logic and preset rules, which can simply and clearly prompt users to fill in missing information; after the user answers the guided rhetorical questions, the system will combine the supplementary content with the original initial question text to obtain a complete question text; this complete text contains all the key information needed to solve the problem.

[0034] Step 140: Based on the complete question text, determine an answer corresponding to the complete question text in a preset civil aviation question and answer dataset.

[0035] Here, the pre-built civil aviation Q&A dataset refers to a pre-built civil aviation Q&A knowledge base containing a large number of frequently asked questions and their answers. This dataset can be built based on historical Q&A records, document content, and FAQs (Frequently Asked Questions).

[0036] The guided knowledge question-answering method provided by the embodiment of the present invention matches the keyword information in the initial question text input by the user with the corresponding question elements at the semantic level. If it is found that the keyword information cannot be completely matched, the specific missing part is clearly located, that is, the missing elements of the question are determined; based on the missing elements of the question, targeted guiding questions are formed to prompt the user to supplement key information. This guided question-answering design enhances the interactivity between the system and the user and improves the integrity of information collection; through multiple rounds of interaction, the user's complete intention is gradually collected, and then the complete question text is determined; the complete question text is used as input, and the most relevant answer is found based on the complete question text in the preset civil aviation question-answering data set; this gradual improvement method based on question elements avoids misjudgment caused by incomplete information at a single time, thereby improving the accuracy of the answer. When users encounter difficult problems, they can obtain highly reliable solutions by simply asking questions.

[0037] Based on any of the above embodiments, in the method, determining, based on the initial question text input by the user, question elements corresponding to the initial question text from a preset civil aviation question element set includes: Determining keyword information of the initial question text input by the user, and determining a civil aviation service scenario corresponding to the keyword information in a preset civil aviation service scenario set; Based on the civil aviation service scenario, determining a problem element corresponding to the civil aviation service scenario from a preset civil aviation problem element set; Among them, each civil aviation service scenario in the preset civil aviation service scenario set corresponds one-to-one to each problem element in the preset civil aviation problem element set.

[0038] Here, the preset civil aviation service scenario set is a system-predefined scenario library, which includes a variety of civil aviation-related service scenarios, such as "flight change", "luggage loss", "ticket inquiry", etc.; each scenario contains a set of specific problem elements.

[0039] It should be noted that based on the identified service scenario, the system will further identify specific problem elements corresponding to that scenario within the pre-set civil aviation problem element set. Each service scenario will correspond to a specific set of problem elements. For example, the problem elements for the "flight change" scenario may include "flight number," "passenger information," and "reason for change."

[0040] For example, in natural language processing, keyword information of the initial question text can be realized through technologies such as TF-IDF (Term Frequency-Inverse Document Frequency), part-of-speech tagging, and named entity recognition.

[0041] The guided knowledge question-and-answer method provided by the embodiment of the present invention can effectively identify the specific civil aviation service scenarios that the user is concerned about by extracting keyword information and matching service scenarios from the initial question text entered by the user. Based on the identified service scenarios, the system finds relevant question elements in the preset civil aviation question element set, such as "flight number" and "passenger information". Through this series of processing, the system can accurately lock in the user's needs and identify the required information elements, thereby improving the accuracy of question understanding and the efficiency of service response. This not only helps to improve the user experience, but also ensures that users receive faster and more accurate answers when asking questions, reducing ambiguity and unnecessary inquiries in subsequent interactions, thereby improving the overall service quality.

[0042] Based on any of the above embodiments, in this method, the question elements include a plurality of keyword slot information, and determining the missing question elements based on the matching results between the keyword information of the initial question text and the question elements corresponding to the initial question text includes: Matching the keyword information of the initial question text with the keyword slot information in the question elements, and determining the keyword slot information that has not been successfully matched; The unsuccessfully matched keyword slot information is determined as a problem missing element.

[0043] It should be noted that keyword slot information refers to the basic unit that constitutes the question elements, and one or more keyword slot information can constitute a question element.

[0044] Exemplarily, the function of the keyword slot information is to compare with the keyword information of the question text entered by the user to determine whether it is consistent. If the keyword information of the question text entered by the user is consistent with all the keyword slot information of the corresponding question elements, the question text entered by the user is considered to be a complete question text, and an accurate answer can be determined based on this complete question text; if the keyword information of the question text entered by the user does not meet all the keyword slot information of the corresponding question elements, then the unmatched keyword slot information will be regarded as a missing element of the question, and then the user will be asked guided questions based on this missing element of the question until all the keyword slot information of the question elements can be matched, and then the accurate answer will be output.

[0045] The guided knowledge question-and-answer method provided by the embodiment of the present invention can effectively identify missing elements of the question by matching the keyword information in the initial question text input by the user with the keyword slot information in the preset question elements. First, the system can accurately identify the key information that is lacking in the user's question, thereby reducing misunderstandings and improving the accuracy of question understanding; secondly, by confirming which keyword slot information has not been successfully matched, the system can guide the user to ask targeted supplementary questions to ensure that complete information is collected and optimize the user interaction experience; finally, timely identification of missing elements of the question can significantly improve the efficiency of service response and enhance user satisfaction. Therefore, this method not only improves the accuracy and efficiency of problem handling, but also provides a smoother interactive experience in the user service process.

[0046] Based on any of the above embodiments, in the method, the step of conducting guided counter-questions to the user based on the missing elements of the question to determine the complete question text includes: Conducting guided counter-questions to the user based on the missing elements of the question to obtain a supplementary question text input by the user; If the keyword information of the supplementary question text successfully matches the missing elements of the question, a complete question text is determined based on the supplementary question text and the initial question text.

[0047] Here, supplementary question text refers to the additional information provided by the user after the guided question. This part of the text is the user's direct response to the system's questions, and its purpose is to fill in the key information missing in the initial question text.

[0048] For example, the user may input the question by voice or text. If the user inputs the question by voice, the voice is translated into text based on ASR technology.

[0049] In one embodiment, the keyword information of the supplementary question text input by the user is matched with the missing elements of the question. If the match is successful, the complete question text is determined based on the supplementary question text and the initial question text; if the match fails, the user is asked guided questions again until a complete match with the missing elements of the question is achieved.

[0050] For example, let's take the fault reporting and consultation scenario as an example. First, the missing elements of the problem in this scenario are as follows:

[0051] The problem elements corresponding to each scenario have keyword slot information that needs to be collected. For example, in the fault reporting consultation scenario, the keyword slot information includes the platform system, operation type, and fault reporting code; based on the missing elements of the problem, reverse questions are asked to the user.

[0052] In one embodiment, the guided questions are designed based on the interrogation dialogue turn. During the first interaction into the interrogation dialogue, the system sets the flag passed to the large model to 0. This indicates that the current interrogation dialogue turn is a new interrogation dialogue turn. In subsequent interactions, the large model determines whether the user's input belongs to the current interrogation dialogue turn. If the user's input belongs to the current interrogation dialogue turn, the flag is incremented by 1, indicating that the conversation is still in the same turn. If the user's input does not belong to the current interrogation dialogue turn, the flag is reset to 0, indicating that the user has started a new conversation. When the flag is 0, the system determines that a new conversation turn has begun. At this point, the large model discards the previous conversation context information and does not incorporate the previous content into the current processing logic. When the flag is not 0, it indicates that the conversation is still in the same turn. In this case, the model retains the previous information and combines it with the current input to better understand the user's needs and provide an appropriate response. This design effectively ensures that the large model can correctly associate context information during the interrogation process, thereby improving the accuracy of the interaction and user experience.

[0053] The guided knowledge question-and-answer method provided by the embodiments of the present invention effectively improves interaction quality and problem-solving efficiency by using guided counter-questions based on missing elements in the question. Specifically, the missing elements of the question are first identified, and targeted counter-questions are conducted accordingly. This process not only guides the user to provide necessary supplementary information but also ensures that the collected information is accurate and relevant, thereby quickly constructing a complete question text. This ensures accurate understanding of the user's intent, reduces repeated communication caused by insufficient information, and thus improves service efficiency.

[0054] Based on any of the above embodiments, in the method, determining an answer corresponding to the complete question text in a preset civil aviation question and answer dataset based on the complete question text includes: Vectorizing the complete question text and the preset civil aviation question and answer dataset to determine a question text vector for the complete question text and a civil aviation question and answer data vector set of the preset civil aviation question and answer dataset; Determine a similarity calculation result between the question text vector and each civil aviation question and answer data vector in the civil aviation question and answer data vector set; Based on the similarity calculation result with the greatest similarity among the similarity calculation results, determining target civil aviation question and answer data similar to the complete question text from the preset civil aviation question and answer dataset; splicing the complete question text with the target civil aviation question and answer data to obtain a target input text; The target input text is input into a civil aviation knowledge question-answering model to obtain an answer corresponding to the complete question text output by the civil aviation knowledge question-answering model; the civil aviation knowledge question-answering model is constructed based on a large language model.

[0055] For example, vectorization typically uses natural language processing techniques, such as word embedding, to map text into a high-dimensional space. Common similarity calculation methods include cosine similarity and Euclidean distance.

[0056] like Figure 2 As shown, Figure 2 The process of obtaining target civil aviation question and answer data is demonstrated.

[0057] For example, by comparing the similarity between the complete question text entered by the user and each civil aviation question and answer data in a preset civil aviation question and answer dataset, target civil aviation question and answer data is extracted from the preset civil aviation question and answer dataset. Furthermore, the complete question text and the target civil aviation question and answer data are concatenated to form a target input text. This target input text is then fed into a pre-trained civil aviation knowledge question and answer model to obtain an answer output by the civil aviation knowledge question and answer model. This answer is the answer generated for the complete question text entered by the user.

[0058] In one embodiment, the base model is incrementally pre-trained based on civil aviation data and a question element table to obtain an initial civil aviation knowledge question-answering model; further, based on the civil aviation question-answering data, the initial civil aviation knowledge question-answering model is fine-tuned using a thinking chain method to determine the civil aviation knowledge question-answering model.

[0059] It should be noted that the civil aviation data and problem element tables used for training are in plain text format.

[0060] For example, civil aviation data includes documents in various formats, including Word, Excel, and PDF, as well as conversation logs from various professional systems. To ensure the effectiveness of subsequent model training, preprocessing and cleaning this civil aviation data is crucial. The main preprocessing steps include manual collection, review and annotation, knowledge generalization, and data vectorization. At the same time, when creating a question element table for civil aviation scenarios, the key information elements that user questions must include are recorded. This information will serve as the basis for subsequent training models to guide question-and-answer interactions, ensuring that the model can understand the user's intent and provide accurate responses. The training goal is to enable the model to predict the next token given a piece of text (token sequence). The incremental pre-training phase focuses on allowing the model to learn a large amount of civil aviation domain knowledge and basic text grammar, allowing the model to learn to follow human instructions and output.

[0061] For example, the pre-set civil aviation Q&A dataset can be a FAQ (historical Q&A) dataset. Based on this FAQ dataset, the initial civil aviation Q&A model is fine-tuned using a chaining approach. The chaining approach is as follows: First, the knowledge characteristics of civil aviation customer service are analyzed to identify the essential elements in the user's question. In the initial round of user questions, the large-scale model is responsible for extracting key information related to the question. Using the essential elements identified in the first step, the large-scale model determines whether the user's question contains all the necessary key information. If any information is missing, the model outputs guiding counter-questions to encourage the user to provide the missing elements. Once the large-scale model has collected all the necessary key information, it outputs this information to the user for confirmation to ensure the accuracy of the collected information. After the user confirms the collected key information, the large-scale model generates and outputs the user's complete intent based on this information to help resolve the user's question. This chaining approach trains the initial aviation Q&A model.

[0062] The guided knowledge question-answering method provided by the present invention uses vectorization technology to convert the complete question text and the civil aviation question-answering dataset into vector form, making similarity calculation more efficient and accurate. The system then identifies the answer that is most similar to the user's question, thereby improving the relevance and accuracy of the answer, further enhancing the response quality of the civil aviation knowledge question-answering model. This approach not only ensures the accuracy of the answer but also enhances the model's flexibility and intelligence when handling complex questions, ultimately optimizing the user experience and reducing the possibility of information omissions and misunderstandings.

[0063] Based on any of the above embodiments, in the method, after determining an answer corresponding to the complete question text in a preset civil aviation question and answer dataset based on the complete question text, the method further includes: Acquiring conversation data between the user and the civil aviation knowledge question-answering model; Identifying sensitive information in the conversation data; The sensitive information is desensitized to determine the desensitized conversation data.

[0064] Specifically, special characters can be used to desensitize sensitive information.

[0065] Sensitive information here refers to user privacy data or other information requiring special protection, such as contact information, flight details, and bank card numbers. Desensitization is a data protection technology that replaces sensitive information with specific characters to conceal its true content, thereby ensuring data security. Conversational data refers to all textual content of user interactions with the civil aviation knowledge question-and-answer model, including user questions, the model's responses, and user supplementary information.

[0066] In one embodiment, when users interact with the civil aviation knowledge question-and-answer model, conversation data is generated. This data may contain sensitive information, such as the user's personal profile, contact information, and flight details. This conversation data is stored for subsequent analysis or optimization. Furthermore, a specific sensitive information identification algorithm is used to automatically detect and mark all possible sensitive information from the conversation data. This identification method can be based on regular expressions, natural language processing (NLP) technology, etc., to automatically detect sensitive information such as phone numbers, bank account numbers, and flight numbers. Furthermore, the identified sensitive information is replaced with specific special characters. This process is designed to prevent this information from being displayed in plain text during storage or subsequent processing, and it is also displayed encrypted in system logs.

[0067] The guided knowledge question-and-answer method provided by an embodiment of the present invention obtains conversation data between a user and a civil aviation knowledge question-and-answer model and identifies sensitive information therein. The method can automatically desensitize the information and replace sensitive content such as personal data such as ID number, telephone number, flight details, etc. with specific special characters. In this way, the processed conversation data can effectively prevent the risk of information leakage during storage, analysis or subsequent use; in addition, this process ensures the compliance of data processing and storage; through the identification and desensitization of sensitive information, not only can user privacy be protected, but also the security and efficiency of data analysis and model training can be improved, so as to achieve the goal of reducing the possibility of data leakage while making full use of conversation data.

[0068] The guided knowledge question and answer device provided by the present invention is described below. The guided knowledge question and answer device described below and the guided knowledge question and answer method described above can be referenced to each other.

[0069] Figure 3 This is a schematic diagram of the structure of the guided knowledge question-answering device provided by the present invention, as shown in FIG. Figure 3 As shown, the guided knowledge question-answering device includes: A question element determination module 310 is configured to determine, based on an initial question text input by a user, question elements corresponding to the initial question text from a preset civil aviation question element set; A missing question element determination module 320 is configured to determine missing question elements based on a matching result between keyword information of the initial question text and question elements corresponding to the initial question text; A guided questioning module 330 is configured to conduct guided questioning of the user based on the missing elements of the question to determine a complete question text; The answer determination module 340 is configured to determine, based on the complete question text, an answer corresponding to the complete question text in a preset civil aviation question and answer dataset.

[0070] Figure 4 An example of a physical structure diagram of an electronic device is shown below. Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communications bus 440. The processor 410 may call logic instructions in the memory 430 to execute a guided knowledge question-answering method, which includes: based on an initial question text input by a user, determining question elements corresponding to the initial question text in a preset civil aviation question element set; determining missing question elements based on a match between keyword information in the initial question text and question elements corresponding to the initial question text; conducting guided counter-questions to the user based on the missing question elements to determine a complete question text; and based on the complete question text, determining an answer corresponding to the complete question text in a preset civil aviation question-answering dataset.

[0071] Furthermore, the logic instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0072] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the guided knowledge question and answer method provided by the above methods, which includes: based on the initial question text input by the user, determining the question element corresponding to the initial question text in a preset civil aviation question element set; based on the matching result between the keyword information of the initial question text and the question element corresponding to the initial question text, determining the missing element of the question; based on the missing element of the question, conducting guided counter-questions to the user to determine the complete question text; based on the complete question text, determining the answer corresponding to the complete question text in a preset civil aviation question and answer data set.

[0073] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the guided knowledge question-answering method provided by the above-mentioned methods, the method comprising: based on the initial question text input by the user, determining the question element corresponding to the initial question text in a preset civil aviation question element set; determining the missing elements of the question based on the matching results between the keyword information of the initial question text and the question elements corresponding to the initial question text; conducting guided counter-questions to the user based on the missing elements of the question to determine the complete question text; and based on the complete question text, determining the answer corresponding to the complete question text in a preset civil aviation question-answering data set.

[0074] The device embodiments described above are merely illustrative. 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, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0075] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A guided knowledge question-answering method, characterized in that: include: Based on the initial question text input by the user, determining the question element corresponding to the initial question text in the preset civil aviation question element set; Determine missing elements of the question based on a matching result between the keyword information of the initial question text and the question elements corresponding to the initial question text; Conducting guided questioning to the user based on the missing elements of the question to determine the complete question text; Based on the complete question text, an answer corresponding to the complete question text is determined in a preset civil aviation question and answer dataset.

2. The guided knowledge question-answering method according to claim 1, characterized in that: The step of determining, based on the initial question text input by the user, a question element corresponding to the initial question text from a preset civil aviation question element set includes: Determine keyword information of the initial question text input by the user, and determine the civil aviation service scenario corresponding to the keyword information in the preset civil aviation service scenario set; Based on the civil aviation service scenario, determining a problem element corresponding to the civil aviation service scenario from a preset civil aviation problem element set; Among them, each civil aviation service scenario in the preset civil aviation service scenario set corresponds one-to-one to each problem element in the preset civil aviation problem element set.

3. The guided knowledge question-answering method according to claim 1, characterized in that: The question elements include a plurality of keyword slot information, and the determining of the missing question elements based on the matching result between the keyword information of the initial question text and the question elements corresponding to the initial question text includes: Matching the keyword information of the initial question text with the keyword slot information in the question elements, and determining the keyword slot information that has not been successfully matched; The keyword slot information that has not been successfully matched is determined as a problem missing element.

4. The guided knowledge question-answering method according to claim 1, characterized in that: The step of conducting guided questioning of the user based on the missing elements of the question to determine the complete question text includes: Conducting guided questioning to the user based on the missing elements of the question to obtain a supplementary question text input by the user; If the keyword information of the supplementary question text successfully matches the missing elements of the question, a complete question text is determined based on the supplementary question text and the initial question text.

5. The guided knowledge question-answering method according to claim 1, characterized in that: The step of determining an answer corresponding to the complete question text in a preset civil aviation question and answer dataset based on the complete question text includes: Vectorizing the complete question text and the preset civil aviation question and answer data set to determine a question text vector of the complete question text and a civil aviation question and answer data vector set of the preset civil aviation question and answer data set; Determine a similarity calculation result between the question text vector and each civil aviation question and answer data vector in the civil aviation question and answer data vector set; Based on the similarity calculation result with the greatest similarity among the similarity calculation results, determining target civil aviation question and answer data similar to the complete question text from the preset civil aviation question and answer data set; Concatenating the complete question text with the target civil aviation question and answer data to obtain a target input text; The target input text is input into a civil aviation knowledge question-answering model to obtain an answer corresponding to the complete question text output by the civil aviation knowledge question-answering model; the civil aviation knowledge question-answering model is constructed based on a large language model.

6. The guided knowledge question-answering method according to claim 5, characterized in that: After determining an answer corresponding to the complete question text in a preset civil aviation question and answer dataset based on the complete question text, the method further includes: Acquiring dialogue data between the user and the civil aviation knowledge question-answering model; identifying sensitive information in the conversation data; The sensitive information is desensitized to determine the desensitized conversation data.

7. A guided knowledge question-answering device, characterized in that: include: A question element determination module, which is used to determine, based on the initial question text input by the user, question elements corresponding to the initial question text in a preset civil aviation question element set; A question missing element determination module, which is used to determine the question missing element based on the matching result between the keyword information of the initial question text and the question element corresponding to the initial question text; A guided questioning module, which is used to conduct guided questioning to the user based on the missing elements of the question to determine the complete question text; An answer determination module is used to determine an answer corresponding to the complete question text in a preset civil aviation question and answer data set based on the complete question text.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the guided knowledge question and answer method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the guided knowledge question and answer method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the guided knowledge question and answer method according to any one of claims 1 to 6 is implemented.

Images