Civil aviation large language model-based reply information acquisition method

Abstract

The application provides a civil aviation large language model-based reply information acquisition method, which comprises the following steps: acquiring user-inputted to-be-replied information; if the target intention is a civil aviation knowledge intention, determining the target knowledge intention from a preset knowledge intention tree; if the target knowledge intention has a target identifier, calling a real-time information acquisition port corresponding to the current target knowledge intention according to user account information to acquire target real-time information corresponding to the user; and inputting the to-be-replied information, the target real-time information and target civil aviation knowledge information into a civil aviation large language model after splicing to obtain reply information output by the civil aviation large language model. The civil aviation large language model-based reply information acquisition method provided by the application can query target real-time information according to a real-time information acquisition port when the determined target knowledge intention has a target identifier, so that the civil aviation large language model can further accurately determine the real intention of the user according to the target real-time information.

Description

Technical Field

[0001] This application relates to the field of intelligent question answering, and in particular to a method for obtaining response information based on a civil aviation large language model. Background Technology

[0002] Because different users have different language expression habits, many users encounter problems such as "missing subject" and "ambiguity" when asking questions. These issues are generally minor in typical knowledge-based question-and-answer scenarios. However, after researching historical data and conducting necessary user surveys, staff discovered that these issues can become significant in many scenarios within the civil aviation field, causing the civil aviation language model's answers to not meet user expectations. For example, when a user enters "How to change a flight ticket," their intent might be related to civil aviation knowledge—they want to understand the ticket change process and related regulations. However, a considerable number of users actually have a civil aviation business intent—they want to change a ticket they have already purchased. Therefore, when user intent recognition is flawed, the content of the answer will also be significantly off-target. Summary of the Invention

[0003] In view of this, this application provides a method for obtaining response information based on a civil aviation large language model, which at least partially solves the problems existing in the prior art.

[0004] In one aspect of this application, a method for obtaining response information based on a civil aviation large language model is provided, comprising the following steps:

[0005] Obtain the pending response information input by the user;

[0006] If the target intent corresponding to the information to be answered is a civil aviation knowledge-related intent, then the target knowledge-related intent is determined from a preset knowledge-related intent tree; the target knowledge-related intent is any leaf node in the knowledge-related intent tree; at least some leaf nodes in the knowledge-related intent tree are set with target identifiers; leaf nodes with target identifiers have corresponding real-time information acquisition ports; if the leaf node is set with a target identifier, it indicates that the intent corresponding to the information to be answered has a probability of being a civil aviation business-related intent;

[0007] If the target knowledge intent has a target identifier, then the real-time information acquisition port corresponding to the current target knowledge intent is called according to the user's account information to obtain the user's target real-time information; and the target knowledge intent and the information to be answered are input into the preset civil aviation knowledge database query port to obtain the target civil aviation knowledge information returned by the civil aviation knowledge database query port.

[0008] The information to be answered, the target real-time information, and the target civil aviation knowledge information are concatenated and input into the civil aviation big language model to obtain the response information output by the civil aviation big language model.

[0009] Beneficial effects:

[0010] This application provides a method for obtaining response information based on a civil aviation big language model. By setting target identifiers on specific leaf nodes in the civil aviation knowledge class intent tree, when a target knowledge class intent is identified with a target identifier, the response is not directly based on the civil aviation knowledge class intent. Instead, the response queries the user's relevant real-time information based on the real-time information acquisition port corresponding to the target knowledge class intent. This allows the civil aviation big language model to further accurately determine the user's true intent based on the target real-time information. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a method for obtaining response information based on a civil aviation large language model, provided in an embodiment of this application. Detailed Implementation

[0013] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0014] It should be noted that, in the absence of conflict, the following embodiments and features can be combined with each other; and, based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0015] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0016] Definitions:

[0017] Root node: The top node of the tree. It has no parent node and is the starting node of the tree.

[0018] Leaf node: A node that has no child nodes is called a leaf node, also known as a terminal node.

[0019] Parent node: The direct parent node of a node is called its parent node.

[0020] Child node: The direct subordinate node of a node is called its child node.

[0021] Please refer to Figure 1 As shown in the embodiment of this application, a method for obtaining response information based on a civil aviation big language model specifically includes the following steps:

[0022] S100: Obtain the user-inputted information to be replied to. Specifically, the user can input the information to be replied to through a preset interactive interface. The information to be replied to can include at least one of the following information formats: text, image, video, and audio.

[0023] S200, if the target intent corresponding to the information to be answered is a civil aviation knowledge-related intent, then the target knowledge-related intent is determined from a preset knowledge-related intent tree. The target knowledge-related intent is any leaf node in the knowledge-related intent tree, and at least some leaf nodes in the knowledge-related intent tree are set with a target identifier; the leaf nodes with target identifiers have corresponding real-time information acquisition ports. If a leaf node is set with a target identifier, it indicates that the intent corresponding to the information to be answered has a probability of being a civil aviation business-related intent.

[0024] It is worth noting that, in this embodiment, the child node mentioned here can be understood as any node that is not the root node in the knowledge intent tree.

[0025] S300, if the target knowledge class intent carries a target identifier, then based on the user's account information, the real-time information acquisition port corresponding to the current target knowledge class intent is invoked to obtain the user's corresponding target real-time information; and the target knowledge class intent and the information to be answered are input into a preset civil aviation knowledge database query port to obtain the target civil aviation knowledge information returned by the civil aviation knowledge database query port. Specifically, the civil aviation knowledge database query port corresponds to a preset civil aviation knowledge database. Specifically, the civil aviation knowledge database can be a graph database, table database, vector database, etc., or it can be composed of multiple databases of the same type or at least some different types.

[0026] S400, the information to be answered, the target real-time information, and the target civil aviation knowledge information are concatenated and input into the civil aviation big language model to obtain the response information output by the civil aviation big language model.

[0027] In some exemplary embodiments of this application, the response information includes knowledge-based text and / or page redirection hyperlinks; the page redirection hyperlinks are used to redirect to the corresponding civil aviation business processing page.

[0028] In this embodiment, by setting a target identifier on a specific leaf node in the civil aviation knowledge intent tree, when a target knowledge intent is identified with a target identifier, the response is not directly based on the civil aviation knowledge intent. Instead, the response queries the user's relevant real-time target information based on the real-time information acquisition port corresponding to the target knowledge intent. This allows the civil aviation big language model to further accurately determine the user's true intent based on the real-time target information.

[0029] Taking the above scenario as an example, when a user inputs "how to reschedule a flight ticket," the knowledge intent located in the civil aviation knowledge intent tree might be "rescheduling procedures." In this embodiment, "rescheduling procedures" is set with a target identifier, and the real-time information acquisition port set for "rescheduling procedures" is the user's itinerary acquisition port. At this time, the user's itinerary acquisition port will be called based on the user's account information to obtain the user's itinerary information returned by the port. Simultaneously, the knowledge related to the rescheduling process (i.e., the target civil aviation knowledge information) obtained through the civil aviation knowledge database query port will also be retrieved.

[0030] After inputting the above information into the civil aviation big data language model, if the user's itinerary information indicates that the user has unexecuted trips, it will be processed according to the civil aviation business intent. In this case, relevant operation instructions can be output and / or a hyperlink can be redirected to the rebooking interface for the unexecuted trips. If the user's itinerary information indicates that the user has no unexecuted trips, it will be processed according to the civil aviation knowledge intent, and a response will be made based on the acquired knowledge related to the rebooking process.

[0031] Therefore, the solution provided in this embodiment can proactively acquire background knowledge (such as user itinerary, flight status, and user profile) from the real-time information acquisition port corresponding to the knowledge intent with target identifier in the knowledge intent tree. This assists the civil aviation big language model in processing, enabling it to output response information that better meets the user's needs.

[0032] In some exemplary embodiments of this application, the civil aviation big language model is configured to prioritize the output of the response information as a civil aviation business processing task when the input information simultaneously contains the target real-time information and the target civil aviation knowledge information.

[0033] In practice, if a user has unexecuted itineraries, the questions they raise are most likely related to those itineraries. Therefore, when recognizing intent, if the user's input intent contains both civil aviation business-related intents and civil aviation knowledge-related intents, they probably want to perform business processing on those itineraries. Therefore, in this embodiment, the civil aviation big language model is configured to prioritize outputting the response information based on civil aviation business processing tasks when the input information simultaneously contains the target real-time information and the target civil aviation knowledge information.

[0034] It is worth noting that in this embodiment, civil aviation business processing tasks are different from civil aviation business intentions. The two can be corresponding. For example, if the civil aviation big language model learns that the current user's intention is a civil aviation business intention, it can be executed as a civil aviation business processing task.

[0035] Furthermore, in this embodiment, when the input information simultaneously includes the target real-time information and the target civil aviation knowledge information, the response information is not necessarily output as a civil aviation business processing task. Instead, by fine-tuning the instructions of the civil aviation large language model, the probability of processing as a civil aviation business processing task is higher than that of processing as a civil aviation knowledge output task in this situation. This is because, in practice, not all users tend to perform business processing tasks.

[0036] Accordingly, in this embodiment, when the target knowledge class intent carries a target identifier, the target knowledge class intent will not be... Figure 1 The system inputs real-time target information and target civil aviation knowledge information into the civil aviation big language model. This aims to avoid the input intent (target knowledge-based intent) affecting the user intent judgment of the civil aviation big language model, so as to give full play to the inherent reasoning ability of the civil aviation big language model.

[0037] Of course, in some exemplary embodiments of this application, the target knowledge class intent and the possible civil aviation business class intent corresponding to the target knowledge class intent can also be simultaneously represented. Figure 1 The real-time information of the target and the target's civil aviation knowledge information are input into the civil aviation big data language model.

[0038] In some exemplary embodiments of this application, the response information includes knowledge-based text and / or business processing buttons;

[0039] The method further includes:

[0040] S500, in response to the user clicking the service processing button, a corresponding service processing instruction is generated based on the target real-time information, and the service processing instruction is sent to the service processing port corresponding to the service processing button.

[0041] In this embodiment, a corresponding message generation method can be set for each type of civil aviation business. This allows the civil aviation big data model to directly generate a corresponding processing button when it detects a user's business processing needs, and send the message directly to the relevant business processing port when the user clicks the button. Specifically, the message generation method can employ a preset template and slot approach. For example, in a ticket refund scenario, a preset message template can be set, including fixed content from the refund instruction message and setting variable content as slots. When the civil aviation big data model recognizes that a user wants to process a refund for a specific trip, it fills the corresponding slots in the message template based on user information and trip information (such as flight number) to obtain the business processing instruction, and then sends the instruction to the corresponding business processing port (API port). This reduces the complexity of business processing for users, eliminating the need to navigate to other interfaces and allowing multiple needs to be met within a single interface.

[0042] In some exemplary embodiments of this application, the method further includes:

[0043] S310, if the target knowledge class intent does not have a target identifier, then the target knowledge class intent and the information to be answered are input into the civil aviation knowledge database query port to obtain the target civil aviation knowledge information returned by the civil aviation knowledge database query port.

[0044] Specifically, the target knowledge class intent and the information to be answered can be concatenated to form a semantic vector. This semantic vector is then input into the civil aviation knowledge database query port, enabling the civil aviation knowledge database corresponding to the query port to use semantic matching to determine the target civil aviation knowledge information to be returned.

[0045] S320, the target civil aviation knowledge information, the target knowledge class intent, and the information to be answered are concatenated and input into the civil aviation big language model to obtain the response information returned by the civil aviation big language model.

[0046] The civil aviation big language model processes the concatenated target civil aviation knowledge information, target knowledge class intent, and pending response information. This allows the big language model to more accurately determine the user's purpose based on the pending response information and target knowledge class intent, and generate response information based on the acquired target civil aviation knowledge information and learned prior knowledge, achieving accurate replies. It is worth noting that in this embodiment, methods such as prompt injection can be used to restrict the civil aviation big language model from generating response information based on the target civil aviation knowledge information, thereby avoiding the occurrence of "illusion" (i.e., the big language model fabricating responses based on the probability space).

[0047] Specifically, in one exemplary embodiment of this application, at least some child nodes in the knowledge-based intent tree have corresponding necessary information. This necessary information is information used to determine whether the corresponding child node represents a target intent. Specifically, the necessary information for a child node can be stored in a preset information table for easy retrieval, or it can be directly embedded in the child node's data as information.

[0048] The step of determining the target knowledge category intent from the preset knowledge category intent tree includes:

[0049] S201, based on the information to be answered, the target knowledge-type intent is matched layer by layer from several knowledge-type intents contained in the knowledge-type intent tree. Specifically, during the matching process, the knowledge-type intent tree can be searched based on existing tree search algorithms to complete the matching of the target knowledge-type intent. Furthermore, during the matching process, if the currently matched node (which can be any node in the knowledge-type intent tree) meets preset matching conditions (e.g., the matching degree reaches a preset threshold), it is determined as the currently matched knowledge-type intent.

[0050] S202, if the currently matched knowledge class intent is a leaf node, then the currently matched knowledge class intent is determined as the target knowledge class intent. Otherwise, continue matching in the child nodes corresponding to the currently matched knowledge class intent.

[0051] S203, if there is no matching sub-node among the sub-nodes corresponding to the currently matched knowledge-type intent, then generate an information retrieval prompt based on the necessary information corresponding to the sub-nodes corresponding to the currently matched knowledge-type intent.

[0052] S204, output the information acquisition prompt to obtain the supplementary information input by the user;

[0053] S205, after concatenating the information acquisition prompt, the supplementary information, and the information to be answered into new information to be answered, the target knowledge intent is determined from the preset knowledge intent tree based on the new information to be answered.

[0054] Because different users have different language expression habits, many users will have missing necessary information (i.e., information to be answered) when they first ask a question. Although the Civil Aviation Big Language Model can respond to questions with missing necessary information, the response often fails to meet the user's expectations and cannot solve the user's problem. At this point, the user can only judge which necessary information is missing in their input question based on the Civil Aviation Big Language Model's response and re-edit the question. This method is inefficient, and users sometimes find it difficult to accurately determine what necessary information they need to add, further degrading the user experience.

[0055] In this embodiment, precise intent matching is performed before inputting the information to be answered into the Civil Aviation Big Language Model. If matching at a lower level is not possible (i.e., no matching sub-node exists among the sub-nodes of the currently matched knowledge-type intent), an information retrieval prompt is generated based on the necessary information corresponding to the sub-nodes of the current node. This allows the user to directly input supplementary information based on the prompt for a more accurate response. Specifically, the information retrieval prompt can also be generated using the Civil Aviation Big Language Model based on the corresponding necessary information.

[0056] The following example illustrates this concept. If a user inputs the question "What is the maximum weight of luggage I can bring when flying?", the initial intent matching will be placed under the civil aviation knowledge category. At this point, a layer-by-layer matching process will be performed within the knowledge intent tree. If the knowledge intent corresponding to a node in the current tree is "baggage allowance regulations," its child nodes will contain intents such as "Airline A baggage allowance regulations," "Airline B baggage allowance regulations," and "Airline C baggage allowance regulations." The necessary information for "Airline A baggage allowance regulations," "Airline B baggage allowance regulations," and "Airline C baggage allowance regulations" is the airline name. Since the user's input does not include the airline name, the airline name will be identified as necessary information, and a corresponding information retrieval prompt will be generated (e.g., "Since different airlines have different baggage allowance regulations, which airline's regulations would you like guidance on?"). This allows the user to input supplementary information, enabling the civil aviation big data model to provide an accurate response.

[0057] In existing large language models, even those trained with civil aviation expertise, when a user inputs "What is the maximum baggage allowance for flying?", most can only provide a direct response. Because a response is required, to ensure accuracy, they often output all baggage allowance regulations from all airlines. This results in two problems: firstly, the output text is too long, leading to extended processing time; secondly, it places a heavy reading burden on the user, and the response lacks specificity, failing to meet the user's needs.

[0058] The solution provided in this embodiment employs a question-and-answer interactive method. When a user's question lacks necessary information, the system proactively prompts the user to provide supplementary information and then provides a targeted and precise answer based on the supplemented content. This approach reduces the length of the output text, better meets user needs, reduces the overall number of interactions, and eliminates the need for users to repeatedly think about how to write their questions.

[0059] In some exemplary embodiments of this application, the method can be applied to an intelligent system, which includes: a civil aviation big language model, a civil aviation business database query port, and several real-time information acquisition ports.

[0060] Specifically, in this embodiment, the civil aviation large language model can be a complete large language model or a large language model group formed by combining multiple models. Furthermore, if the civil aviation large language model is a large language model group formed by combining multiple models, it can also include other artificial intelligence models (such as machine models, neural network models, etc.) that are not large language models. It is only necessary to ensure that the large language model group contains at least one large language model that has undergone targeted training using civil aviation knowledge (including pre-training, instruction fine-tuning, reinforcement learning, etc.). In this embodiment, the large language model generally refers to a large language model based on the Transformer architecture. Specifically, this large language model can include both an encoder and a decoder, or only a decoder, or only an encoder.

[0061] The civil aviation business database query port has a corresponding civil aviation data storage database, which can be a graph database, table database, vector database, etc. In this embodiment, the civil aviation data storage database is a table database, specifically, it can be an SQL database. It can include flight data for all historical, currently executing, or yet-to-be-executed flights, user travel data, user order data, or travel data for other modes of transportation, etc.

[0062] Several real-time information acquisition ports can be pre-configured API ports, such as flight status acquisition ports, user itinerary acquisition ports, and user profile acquisition ports. Specifically, these real-time information acquisition ports can directly reuse API ports from other itinerary management software, systems, and platforms.

[0063] In some exemplary embodiments of this application, between step S100 and step S200, the method further includes:

[0064] S110, based on the pending response information, determine the target intent corresponding to the pending response information from the civil aviation intent list. The civil aviation intent list includes civil aviation knowledge-related intents, several first-category civil aviation business intents, and several second-category civil aviation business intents, wherein the task complexity corresponding to the first-category civil aviation business intents is higher than the task complexity of the second-category civil aviation business intents.

[0065] Specifically, the first type of civil aviation business intent does not have a corresponding real-time information acquisition port, while each second type of civil aviation business intent has at least one real-time information acquisition port. Specifically, in this embodiment, the civil aviation business intent... Figure 1 Generally, this refers to users' intentions regarding civil aviation matters, such as ticket refunds or rebookings. It may also involve querying or analyzing real-time, near-real-time, or frequently changing data, such as the current location or status of a specific flight. Civil aviation business-related intentions are primarily distinguished from civil aviation knowledge-related intentions. Figure 1 This generally refers to users looking for fixed civil aviation information that won't change in the short term, such as the customer service number for XX airport, the official website of XX airport, or the baggage requirements for a certain flight.

[0066] Step S200 further includes:

[0067] S210, if the target intent is any of several first-class civil aviation business intents, then the civil aviation big language model is used to obtain a database query statement based on the information to be answered. Specifically, in this embodiment, the database query statement can be an SQL statement. The civil aviation big language model can be trained to generate a database query statement based on the information to be answered by specifically training its text2sql capability.

[0068] S211, the database query statement is sent to the civil aviation business database query port to obtain the target dataset returned by the civil aviation business database query port.

[0069] S212, the target dataset and the information to be answered are concatenated and then input into the civil aviation big language model to obtain the response information returned by the civil aviation big language model.

[0070] S220, if the target intent is any of several second-category civil aviation business intents, then obtain a list of port information corresponding to the target intent. The list of port information includes port information for at least a portion of the several real-time information acquisition ports; the port information includes a description of the corresponding real-time information acquisition port's functions. The port function description may include information on which fields can be obtained from the real-time information acquisition port, and how to invoke the port, etc.

[0071] S221, the port information list and the information to be replied to are concatenated and input into the civil aviation big language model, so that the civil aviation big language model calls at least one of the real-time information acquisition ports corresponding to the port information list to obtain the target real-time information, and outputs reply information according to the target real-time information and the information to be replied to.

[0072] That is, in this embodiment, the port information list is input as a prompt command into the civil aviation big language model, so that the civil aviation big language model can select the real-time information acquisition port that can be used this time from several real-time information acquisition ports corresponding to the port information list according to the thinking chain technology.

[0073] Specifically, in this embodiment, the list of port information corresponding to each second type of civil aviation business intent can be set by staff based on their work experience. Alternatively, it can be determined based on real-time information acquisition ports that are statistically calculated from historical data and whose correlation with the current second type of civil aviation business intent is greater than a preset threshold.

[0074] This application provides an intelligent system based on a civil aviation big data language model, which sets several first-type civil aviation business intentions and several second-type civil aviation business intentions in the civil aviation intention list. The task complexity corresponding to the first-type civil aviation business intentions is higher than that of the second-type (which can also be understood as the first-type civil aviation business not having a corresponding API port). Furthermore, upon obtaining the information to be answered, the target intention is directly determined. Therefore, when the civil aviation big data language model processes the information to be answered, it does not need to perform user intention recognition itself, and can quickly decide which strategy to use based on whether the obtained target intention is a first-type or second-type civil aviation business intention. This reduces the training difficulty of the civil aviation big data language model.

[0075] Furthermore, in this embodiment, a distinction is made between the first type of civil aviation business intent and the second type of civil aviation business intent. This eliminates the need to design specific API ports (i.e., real-time information acquisition ports) for complex processing tasks. Instead, the civil aviation big data language model automatically obtains the database query statement based on the information to be answered, thereby calling the civil aviation business database query port to obtain the target dataset returned by the civil aviation business database query port, and using this as the basis for generating the response information. This avoids the need to set up a real-time information acquisition port specifically for each user's needs, and allows for the use of existing business query APIs as real-time information acquisition ports, or a small number of specific real-time information acquisition ports. On the other hand, even if a user generates a new need that staff could not anticipate, the civil aviation big data language model can still retrieve the corresponding information from the civil aviation business database query port to generate a valid response.

[0076] This embodiment illustrates the concept with an example. If a user's input is "Please help me check when my flight can be checked in," this can be categorized as a second type of civil aviation business intent. The intelligent system can obtain flight information for all flights the user hasn't checked in for yet through a preset user itinerary retrieval port, allowing the large language model to provide a targeted response based on this information. This type of intent can be addressed by setting up a simple API port. However, if the user's input is "Please help me find out how many flights I have with Airline A and Airline B in my history," or "How many of my history flights depart at night, and how many of those flights were delayed," these types of intents, which are more random and complex, are difficult to quickly obtain effective information through API ports. In this embodiment, the input of "Please help me find out how many flights I have with Airline A and Airline B in my history," or "How many of my history flights depart at night, and how many of those flights were delayed," will be classified as a first type of civil aviation business intent, allowing the civil aviation large language model to provide an effective response based on data obtained from the civil aviation business database query port.

[0077] In some exemplary embodiments of this application, the civil aviation big language model includes a civil aviation-specific model and a general knowledge model.

[0078] The step of using the civil aviation big language model to obtain a database query statement based on the information to be answered includes:

[0079] The civil aviation-specific model is used to obtain a database query statement based on the information to be answered.

[0080] The step of concatenating the target dataset with the information to be answered and inputting it into the civil aviation big language model to obtain the response information returned by the civil aviation big language model includes:

[0081] The target dataset and the information to be answered are concatenated and then input into the general knowledge model to obtain the response information returned by the general knowledge model.

[0082] The step of concatenating the port information list and the information to be replied to, and then inputting them into the civil aviation big language model, so that the civil aviation big language model can call at least one of the real-time information acquisition ports corresponding to the port information list to obtain the target real-time information, and output reply information based on the target real-time information and the information to be replied to, includes:

[0083] The process involves concatenating the port information list and the information to be replied to, and then inputting them into the civil aviation dedicated model. This enables the civil aviation dedicated model to call at least one of the real-time information acquisition ports corresponding to the port information list to obtain the target real-time information, and output reply information based on the target real-time information and the information to be replied to.

[0084] The list of civil aviation intents also includes general knowledge intents;

[0085] The intelligent system is also used to perform the following steps:

[0086] If the target intent is the general knowledge intent, then the information to be answered is input into the civil aviation big language model to obtain the response information output by the civil aviation big language model.

[0087] The step of inputting the information to be answered into the civil aviation big language model to obtain the response information output by the civil aviation big language model includes:

[0088] The information to be answered is input into the general knowledge model to obtain the response information output by the general knowledge model.

[0089] The intelligent system also includes a first intent recognition model;

[0090] The step of determining the target intent corresponding to the pending response information from the civil aviation intent list includes:

[0091] The information to be answered is input into the first intent recognition model so that the first intent recognition model can determine the target intent corresponding to the information to be answered from the civil aviation intent list.

[0092] The intelligent system also includes a second intent recognition model;

[0093] The step of determining the target knowledge category intent from the preset knowledge category intent tree includes:

[0094] The information to be answered is input into the second intent recognition model so that the second intent recognition model can determine the target knowledge-type intent from the preset knowledge-type intent tree.

[0095] Specifically, the first intent recognition model and / or the second intent recognition model can be included within the civil aviation large language model, or they can exist independently. The first intent recognition model and / or the second intent recognition model can also be constructed using the large language model, primarily performing classification tasks. In some embodiments, the capabilities and tasks to be processed by the first intent recognition model and / or the second intent recognition model can be simultaneously and specifically trained (e.g., instruction fine-tuning) to integrate them with the civil aviation professional model into a single model. In this embodiment, the civil aviation professional model, the general knowledge model, the first intent recognition model, and the second intent recognition model are four independent large language models. This allows each model to be trained specifically according to its functional and capability requirements, without requiring extensive training to enable a single model to handle multiple tasks simultaneously, thereby reducing the computational cost, time cost, and data volume of model training.

[0096] Existing knowledge-based question-answering systems mostly employ general language models, meaning they lack a clear focus on specific knowledge domains and can effectively answer general questions. However, due to a lack of specialized knowledge in certain fields, or because they prioritize general knowledge, they cannot effectively address specialized questions within specific professional domains during the learning process.

[0097] If a large language model that has not been initially trained is trained solely on professional knowledge, the limited corpus size of a single professional knowledge makes it impossible to train a high-quality large language model. Some approaches further train a pre-trained general knowledge model using relevant knowledge from a specific professional field. However, this approach often leads to the model forgetting the learned general knowledge, or the probability space of the large language model being influenced by the specificity of professional knowledge, resulting in errors in judging general knowledge (for example, the same noun may have significantly different meanings in the general knowledge domain and the civil aviation domain).

[0098] To address the aforementioned issues, this embodiment establishes two independent large language models for civil aviation: a civil aviation-specific model and a general knowledge model. The general knowledge model is trained using general knowledge, while the civil aviation-specific model is trained using specialized knowledge in the civil aviation field on top of a pre-trained general knowledge model (which may be the same as or different from the aforementioned general knowledge model).

[0099] In practical applications, if the target intent corresponding to the information to be answered is a general knowledge intent listed in the civil aviation intent list, the information to be answered can be directly input into the general knowledge model for response. However, if the target intent is a civil aviation domain intent (i.e., the target intent is a first-class civil aviation business intent, a second-class civil aviation business intent, or a civil aviation knowledge intent), the generation of subsequent response information mainly uses the civil aviation professional model.

[0100] In this way, it is possible to avoid the problem of failing to meet user needs due to errors in the judgment of general knowledge by civil aviation-specific models.

[0101] It is worth noting that in certain specific situations, in order to obtain response information that better meets user needs, even if the target intent is in the civil aviation field, a general knowledge model may be used to generate response information, or a civil aviation-specific model and a general knowledge model may work together to generate response information.

[0102] For example, in this embodiment of the application, if the target intent is a first type of civil aviation business intent, a civil aviation-specific model will be used to generate a database query statement, and a general knowledge model will be used to generate a response message based on the target dataset returned by the query and the information to be responded to.

[0103] In this embodiment, since the general knowledge model has not been trained on civil aviation-related knowledge or database query generation capabilities, a civil aviation-specific model is used to generate database queries to ensure their accuracy. Meanwhile, because the general knowledge model has stronger general knowledge capabilities, it responds to the information to be answered based on the target dataset returned by the query. Leveraging its superior data statistical analysis capabilities and language fluency compared to the civil aviation-specific model, the final generated response information better meets the user's needs.

[0104] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0105] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0106] In an exemplary embodiment of this disclosure, an electronic device capable of implementing the above-described method is also provided.

[0107] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0108] An electronic device according to this embodiment of the present application. The electronic device is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of this application.

[0109] Electronic devices are manifested in the form of general-purpose computing devices. Components of an electronic device may include, but are not limited to: at least one processor, at least one memory, and buses connecting different system components (including memory and processor).

[0110] The storage device stores program code that can be executed by the processor to perform the steps described in the "Exemplary Methods" section above, according to various exemplary embodiments of this application.

[0111] The storage may include readable media in the form of volatile storage, such as random access memory (RAM) and / or cache memory, and may further include read-only memory (ROM).

[0112] The storage may also include programs / utilities having a set (at least one) of program modules, including but not limited to: an operating system, one or more applications, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0113] A bus can represent one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus that uses any of the various bus architectures.

[0114] The electronic device can also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), one or more devices that enable a user to interact with the electronic device, and / or any device that enables the electronic device to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication can be performed via input / output (I / O) interfaces. Furthermore, the electronic device can communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0115] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0116] In exemplary embodiments of this disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the methods described above is stored. In some possible implementations, various aspects of this application may also be implemented as a program product comprising program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps of the various exemplary embodiments of this application described in the "Exemplary Methods" section above.

[0117] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0118] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0119] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0120] Program code for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0121] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of this application, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0122] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0123] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for obtaining response information based on a civil aviation large language model, characterized in that, Includes the following steps: Obtain the pending response information input by the user; If the target intent corresponding to the information to be answered is a civil aviation knowledge-related intent, then the target knowledge-related intent is determined from the preset knowledge-related intent tree; The target knowledge class intent is any leaf node in the knowledge class intent tree; at least some leaf nodes in the knowledge class intent tree are set with target identifiers; Leaf nodes with target identifiers have corresponding real-time information acquisition ports; if a leaf node has a target identifier, it means that the intent corresponding to the information to be answered has a probability of being a civil aviation business intent. If the target knowledge class intent has a target identifier, then the real-time information acquisition port corresponding to the current target knowledge class intent is called according to the user's account information to obtain the user's corresponding target real-time information; The target knowledge class intent and the information to be answered are input into a preset civil aviation knowledge database query port to obtain the target civil aviation knowledge information returned by the civil aviation knowledge database query port. The information to be answered, the target real-time information, and the target civil aviation knowledge information are concatenated and input into the civil aviation big language model to obtain the response information output by the civil aviation big language model.

2. The method for obtaining response information according to claim 1, characterized in that, The civil aviation big language model is configured to prioritize the output of the response information in the form of civil aviation business processing tasks when the input information simultaneously contains the target real-time information and the target civil aviation knowledge information.

3. The method for obtaining response information according to claim 2, characterized in that, The response information includes knowledge-based text and / or page redirection hyperlinks; the page redirection hyperlinks are used to redirect to the corresponding civil aviation business processing page.

4. The method for obtaining response information according to claim 2, characterized in that, The response information includes knowledge-based text and / or business processing buttons; The method further includes: In response to the user clicking the service processing button, a corresponding service processing instruction is generated based on the target real-time information, and the service processing instruction is sent to the service processing port corresponding to the service processing button.

5. The method for obtaining response information according to claim 1, characterized in that, The method further includes: If the target knowledge class intent does not have a target identifier, then the target knowledge class intent and the information to be answered are input into the civil aviation knowledge database query port to obtain the target civil aviation knowledge information returned by the civil aviation knowledge database query port; The target civil aviation knowledge information, the target knowledge class intent, and the information to be answered are concatenated and input into the civil aviation big language model to obtain the response information returned by the civil aviation big language model.

6. The method for obtaining response information according to any one of claims 1 or 5, characterized in that, At least some of the child nodes in the knowledge-type intent tree have corresponding necessary information; The step of determining the target knowledge category intent from the preset knowledge category intent tree includes: Based on the information to be answered, the target knowledge intent is matched layer by layer from several knowledge intents contained in the knowledge intent tree; If the currently matched knowledge class intent is a leaf node, then the currently matched knowledge class intent is determined as the target knowledge class intent; otherwise, the matching continues in the child nodes corresponding to the currently matched knowledge class intent. If there is no matching child node among the child nodes corresponding to the currently matched knowledge intent, then an information retrieval prompt is generated based on the necessary information corresponding to the child node corresponding to the currently matched knowledge intent. Output the aforementioned information to obtain supplementary information input by the user; After concatenating the information acquisition prompt, the supplementary information, and the information to be answered into new information to be answered, the target knowledge intent is determined from the preset knowledge intent tree based on the new information to be answered.

7. An electronic device, characterized in that, Including processor and memory; The processor executes the steps of the method as described in any one of claims 1 to 6 by invoking programs or instructions stored in the memory.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores a program or instructions that cause a computer to perform the steps of the method as described in any one of claims 1 to 6.

Images