Travel plan generation method and system, server and readable storage medium

Abstract

The invention discloses a travel plan generation method and system, a server and a readable storage medium, and belongs to the technical field of computers. The method comprises the following steps: receiving a travel demand and a first constraint condition input by a user; the travel demand can adopt one or more of the following modes: voice, text, image and video; the first constraint condition is used for limiting the personalized travel of the user; constructing a first user portrait based on the travel demand and the first constraint condition; inputting the travel demand, the first constraint condition and the second user portrait into a large language model, and outputting a first travel plan; the second user portrait is obtained by performing semantic retrieval processing and evidence enhancement processing on the first user portrait; the big language model is used for generating a personalized travel plan; the first trip plan is a trip plan having a user personalized tag. According to the method, the interaction naturalness can be improved by introducing a large language model and a multi-modal input mechanism, and the obtained travel plan is more accurate.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, and specifically relates to a method, system, server, and readable storage medium for generating travel plans. Background Technology

[0002] The tourism industry encompasses a wide range of human activities centered around tourism, including sightseeing, tours, entertainment, and leisure. Characterized by its comprehensiveness, high interconnectedness, and long industrial chain, tourism covers multiple sectors such as catering, accommodation, transportation, and shopping. Driven by globalization and internet technology, the tourism industry has become a vital pillar of the global economy, bringing new economic momentum and development opportunities to various countries. Simultaneously, the tourism industry can promote economic growth, meet people's fundamental needs for a better life, and enhance cultural soft power.

[0003] With the rapid development of artificial intelligence and blockchain technologies, the tourism industry is accelerating its move towards intelligence and personalization. Currently, tourism platforms use artificial intelligence systems to analyze large-scale data such as users' search history, location information, and interests to achieve real-time recommendations and personalized itinerary construction for travelers.

[0004] However, existing travel platforms struggle to accurately construct personalized user profiles, and cannot provide exclusive, customized itineraries based on these profiles. Summary of the Invention

[0005] This application provides a travel planning generation method, system, server, and readable storage medium, which can support users to express their travel needs in natural language and images, greatly improving the naturalness of interaction. Furthermore, the travel plans output by the large language model are more accurate and better match the user's real travel needs.

[0006] Firstly, this application provides a method for generating travel plans, the method comprising:

[0007] The system receives user input regarding travel needs and a first constraint. Travel needs can take one or more modalities: voice, text, image, and video. The first constraint defines the user's personalized itinerary. Based on the travel needs and the first constraint, a first user profile is constructed. The travel needs, the first constraint, and the second user profile are input into a large language model, which outputs a first travel plan. The second user profile is obtained by performing semantic retrieval and evidence enhancement processing on the first user profile. The large language model generates a personalized travel plan. The first travel plan is a travel plan with personalized user tags.

[0008] This application, by receiving user input of travel needs and first constraints, can obtain multimodal travel needs and further constraints on the user's travel itinerary. Furthermore, based on the obtained travel needs and first constraints, a first user profile is constructed, which can integrate multimodal travel needs and personalized itinerary constraints to understand and identify the user's true travel intentions, thereby accurately constructing the first user profile. Even further, by performing semantic retrieval processing and evidence enhancement processing on the first user profile, a second user profile is obtained. Based on the above description, by inputting the travel needs, first constraints, and second user profile into a large language model, a first travel plan is output, which can generate a more comprehensive, accurate, and consistent first travel plan with the user's actual itinerary.

[0009] Therefore, by introducing a large language model and a multimodal input mechanism, users can express their travel needs using natural language and images, greatly improving the naturalness of the interaction. Furthermore, the large language model can identify and parse users' actual travel needs, and based on these needs, can construct high-precision user profiles (e.g., the first user profile). This results in more accurate and comprehensive travel planning based on these high-precision user profiles (e.g., the first travel plan), enabling more personalized user itineraries to be flexibly generated while dynamically interacting with users.

[0010] In some possible implementations, before inputting the second user profile into the large language model, the method further includes: performing semantic retrieval processing and evidence enhancement processing on the first user profile using a retrieval enhancement generation method to obtain the second user profile.

[0011] In some possible implementations, after generating the corresponding first travel plan, the method further includes: updating the first travel plan based on the user's input travel demand to obtain a second travel plan; and updating the first user profile based on the second travel plan to obtain a third user profile.

[0012] In the above implementation, users can change their travel needs at any time. In response to the user's change of travel needs, the first travel plan can be updated, and then the first user profile can be updated based on the updated second travel plan. This makes the interaction stronger and ensures that the subsequent travel plans are generated based on the updated user profile, making the travel planning more accurate.

[0013] In some possible implementations, First Travel Planning includes at least the user's personalized itinerary, rankings of multiple scenic spots, candidate guides matched for each scenic spot, ticket prices for each scenic spot, visitor flow for each scenic spot, weather data, and road condition data leading to each scenic spot.

[0014] In some possible implementations, travel demand includes at least one or more of the following: travel destination, travel preferences, travel dates, number of travelers, mode of transportation, travel budget, and travel type; travel preferences include at least one or more of the following: academic travel, business travel, medical travel, educational travel, and tourism travel; travel type includes tourism, shopping, conferences, educational study tours, or team building activities.

[0015] Secondly, this application provides a travel planning generation system, including:

[0016] The receiving module is used to receive the user's travel request and first constraint; the travel request can take one or more of the following modalities: voice, text, image and video; the first constraint is used to limit the user's personalized itinerary;

[0017] The processing module is used to construct the first user profile based on travel demand and the first constraint.

[0018] The processing module is also used to input travel demand, first constraints, and second user profile into the large language model and output the first travel plan; the second user profile is obtained by semantic retrieval processing and evidence enhancement processing of the first user profile; the large language model is used to generate personalized travel plans; the first travel plan is a travel plan with personalized user tags.

[0019] In some possible implementations, the processing module is also used to perform semantic retrieval processing and evidence enhancement processing on the first user profile using retrieval enhancement generation methods to obtain a second user profile.

[0020] In some possible implementations, the system also includes an update module;

[0021] The update module is used to update the first travel plan based on the user's input travel needs to obtain the second travel plan;

[0022] The update module is also used to update the first user profile based on the second travel plan to obtain the third user profile.

[0023] In some possible implementations, after generating the corresponding first travel plan, the receiving module is also used to receive the first travel plan.

[0024] Thirdly, this application provides a server including a processor; when the processor executes computer code or instructions in memory, the processor causes the processor to perform the travel planning generation method in the first aspect and any possible design of the first aspect.

[0025] Fourthly, this application provides a chip system applied to an electronic device including a memory, a display screen, and sensors; the chip system includes: one or more interface circuits and one or more processors; the interface circuits and processors are interconnected via lines; the interface circuits are used to receive signals from the memory and send signals to the processors, the signals including computer code or instructions stored in the memory; the processors invoke the computer code or instructions to cause the server to execute the travel planning generation method in the first aspect and any possible design of the first aspect.

[0026] The chip system may include one chip or multiple chips; when the chip system includes multiple chips, this application does not limit the type and number of chips.

[0027] Fifthly, this application provides a readable storage medium storing code or instructions, which a processor invokes to cause a server to execute the travel planning generation method in the first aspect and any possible design of the first aspect.

[0028] Sixthly, this application provides a computer program product that, when run on a computer, causes the computer to execute the travel planning generation method in the first aspect and any possible design of the first aspect.

[0029] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A flowchart illustrating a travel planning generation method provided in an embodiment of the present invention;

[0032] Figure 2 This is a schematic diagram of the server interface in a travel planning generation method provided by an embodiment of the present invention;

[0033] Figure 3 A structural block diagram of a travel planning generation system provided in an embodiment of the present invention;

[0034] Figure 4 This is another structural block diagram of a travel planning generation system provided in an embodiment of the present invention. Detailed Implementation

[0035] The travel planning generation method of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. This embodiment is a preferred implementation, but should not be construed as limiting the scope of protection of the present invention. Those skilled in the art should understand that adjustments to the order of steps, parameter selection, and module division without departing from the essence of the invention are all within the scope of protection of the present invention.

[0036] The travel planning generation method provided in this application can be applied to various travel scenarios, such as tourism, academic exchange, business, educational research, medical and health care, international exhibitions and conferences, cross-border e-commerce, outbound study abroad, and smart transportation. Based on accurate user profiles, the travel planning generation method provides personalized travel plans. Furthermore, the travel plans generated by this application encompass comprehensive data from multiple aspects, including user data, tour guide services, scenic spots, and weather, making them more comprehensive. Through the synergistic effect of various data sources, a better travel experience can be provided to users.

[0037] Optionally, the travel planning generation method provided in this application supports semantic input, personalized recommendations, and non-fungible token (NFT) ownership verification and encryption decomposition functions, enabling it to provide a one-stop, immersive intelligent travel service. The travel planning generation method also possesses semantic awareness and personalized knowledge retrieval capabilities, serving online travel agency (OTA) customer service robots, scenic area audio guide systems, and personal tour guide assistants. Specifically, the travel planning generation method provided in this application can provide real-time question-and-answer, itinerary recommendations, and language translation services, thereby further improving service quality and increasing user satisfaction.

[0038] Optionally, the travel planning generation method provided in this application embodiment can achieve data docking with local governments or cultural and tourism authorities. Through data docking, it can realize tourism data aggregation, hot scenic spot prediction, festival event early warning, and user satisfaction tracking, and further provide comprehensive data support for digital governance of cultural and tourism, smart scenic spot construction, and intelligent upgrading of tourism public services.

[0039] Optionally, the travel planning generation method provided in this application integrates Virtual Reality (VR) technology, enabling VR navigation for users' travel plans. The method also integrates Artificial Intelligence (AI) technology, allowing AI-powered explanations for users' travel plans. This allows for the generation of stories through VR navigation and AI explanations, providing a data foundation for developing digital twin experience services for global cultural heritage sites, museums, and theme parks. While meeting the virtual tourism needs for education, science popularization, and cultural dissemination, it also improves the digital conversion rate of traditional cultural and tourism venues.

[0040] Optionally, the travel planning generation method provided in this application embodiment can also integrate methods such as NFT credentials, Proof of Attendance Protocol (POAP) and community reputation mechanisms to support users in turning their travel footprints, travel videos and immersive experiences into digital assets and circulating them for the second time, thereby promoting the synergy of interests between users and servers and the co-construction of the community.

[0041] Understandably, the visit commemorative badge is a form of digital attendance proof based on blockchain. A community reputation mechanism is an institutional design that builds trust, regulates behavior, and improves the overall health of the community by recording, evaluating, and providing feedback on user behavior.

[0042] The travel planning generation method provided in this application can be applied to a server. For example, the server can be a local server with display hardware and corresponding software support; or it can be a virtual server; or it can be a cloud server; or it can be a server cluster. This application does not impose any restrictions on the specific type or number of servers.

[0043] Please see Figure 1 , Figure 1 A flowchart of a travel planning generation method provided in an embodiment of this application is shown.

[0044] like Figure 1 As shown, the travel planning generation method in this embodiment includes the following steps:

[0045] S101 receives the user's travel needs and the first constraint.

[0046] Optionally, in some embodiments, travel requests can take one or more of the following modalities: voice, text, image, and video. Travel requests can be stored in a multimodal input cache library on the server. Specifically, voice is stored using a transcribed text index, text is stored using a text index, images are stored using image tags, and video is stored separately using image tags for each frame.

[0047] For example, the image could be an image of a tourist attraction at the travel destination downloaded by the user from the internet; or, the image could be an image of a ticket the user previously used at a tourist attraction at the travel destination. The video could be a promotional video of the travel destination downloaded by the user from the internet; or, the video could be a video taken by the user previously at a tourist attraction at the travel destination.

[0048] Optionally, in some embodiments, travel demand includes at least one or more of the following: travel destination, travel preference, travel date, number of travelers, mode of transportation, travel budget, and travel type.

[0049] For example, the travel destination can be a city entered by the user; or, the travel destination can be the name of an attraction entered by the user; or, the travel destination can be the corresponding geographical location entered by the user.

[0050] Optionally, in some embodiments, travel preferences include at least one or more of the following: academic travel, business travel, medical travel, educational travel, and tourism travel.

[0051] Optionally, in some embodiments, the travel type includes tourism, shopping, conferences, educational study tours, or team building activities.

[0052] For example, the travel destination can be a city; the travel preference can be academic travel; the number of travelers is 3; the mode of transportation is airplane, high-speed rail or long-distance bus; the travel budget is 8,000 yuan; and the travel type is tourism.

[0053] It should be noted that the aforementioned travel destination, number of travelers, and mode of transportation can all be one or more, and this application embodiment does not limit this. Furthermore, in addition to the travel preferences listed above, the travel preferences in this application embodiment also include other common user preferences known to those skilled in the art (e.g., cross-border e-commerce and smart transportation), and this application embodiment does not limit these. In addition to the travel types listed above, the travel types in this application embodiment also include other common user types known to those skilled in the art (e.g., studying abroad), and this application embodiment does not limit these.

[0054] It should also be noted that the above-described travel needs are merely illustrative examples. The travel needs in this application embodiment may also include other travel needs of the user, and this application embodiment does not limit these needs. For example, travel needs may also include hotel preferences, which can be sorted and categorized according to the hotel's star rating. Furthermore, travel needs may also include the traveler's group (adults, seniors, children, or people with disabilities, etc.). Additionally, travel needs may include activity preferences, which may include shopping, theatrical performances, concerts, etc.

[0055] Optionally, in some embodiments, the first constraint is used to limit the user's personalized itinerary.

[0056] Optionally, in some embodiments, the first constraint can be further limited to the user's itinerary through different tools pre-set in the server. For example, if the user's travel preference is academic travel, the first constraint can be a pre-set academic conference and personal planning tool in the server. Or, if the user's travel preference is medical travel, the first constraint can be a pre-set medical travel and personal planning tool in the server.

[0057] It should be noted that the first constraint listed above is only illustrative and is not limited in this application. For example, the first constraint may also be an automatic prompt generation tool, an automatic flight booking tool, or an automatic hotel booking tool.

[0058] In response to the user's input on the server's first interface, the server can receive the user's travel request and the first constraint.

[0059] For example, such as Figure 2 As shown, in the first interface 10, the user can select to input via text, voice, image, or video by clicking control 101. The user can select a travel destination by clicking control 102. The user can select travel preferences by clicking control 103. The user can select the first constraint by clicking control 104.

[0060] Optionally, in some embodiments, controls 101, 102, 103, and 104 can also be configured as controls comprising multiple sub-controls. When controls 101, 102, 103, and 104 are configured to include multiple sub-controls, the user can complete the corresponding selection by clicking on the sub-controls of different controls. For example, ... Figure 2 As shown, control 103 includes sub-controls for academic travel 1031, business travel 1032, medical travel 1033, educational travel 1034, and leisure travel 1035. When the user's travel preference is academic travel, the user can click... Figure 2 The first interface shown contains the academic travel sub-control 1031. In response to the user's click, the server can receive the user's travel preference as academic travel.

[0061] Optionally, in another embodiment, controls 101, 102, 103, and 104 can be configured as controls with drop-down lists. When controls 101, 102, 103, and 104 are all controls with drop-down lists, users can click on the drop-down lists of different controls to display the corresponding drop-down lists, and then select travel destinations, travel preferences, etc., that match their actual travel needs from the drop-down lists.

[0062] It should be noted that the above-mentioned settings of controls and sub-controls are all exemplary. Those skilled in the art can also use other layouts to set controls and their corresponding sub-controls. This application does not limit this.

[0063] It should also be noted that, Figure 2 The first interface shown is for illustrative purposes only. In practical applications, the first interface may include more than [other interfaces]. Figure 2 The number of controls, whether more or fewer, is not limited in this embodiment.

[0064] For example, if a user's travel preference is academic travel, and the first constraint is a pre-set academic conference and personal planning tool on the server, the server can receive the first constraint in response to the user's relevant operations, thereby obtaining a more realistic travel plan for the user. Specifically, in response to the user's search operation in the academic and personal planning tool, the server can obtain basic information about multiple pre-stored conferences. In response to the user's operation of selecting a conference from multiple conferences (i.e., the target conference), the server can determine the basic information of the target conference. The basic information of the target conference includes the conference time, conference location, participants, and conference host, etc. This application, through the user's further selection of the target conference, enables the server to receive the target conference selected by the user, thereby enabling the server to integrate the user's selected target conference with the travel needs, preparing for the subsequent output of a more suitable first travel plan.

[0065] It should be noted that after a user selects a target meeting, in response to the user adding the target meeting to the meeting log, the server obtains a meeting log containing the target meeting. The meeting log is used to store historical meeting data and is a pre-built storage module on the server.

[0066] Based on the above description, the server can receive the user's travel needs and the first constraint, thereby preparing for the generation of a travel plan. This embodiment of the application, by setting a multimodal input mechanism, supports users in expressing their travel needs through natural language, images, and videos, greatly improving the naturalness of the interaction and providing a data foundation for subsequent travel planning, thereby improving the accuracy of travel recommendations.

[0067] S102, based on travel demand and the first constraint, constructs the first user profile.

[0068] Optionally, in some embodiments, the first user profile is obtained by Large Language Model (LLM) after performing intent recognition and slot parsing on travel demand and first constraints.

[0069] It should be noted that the first user profile can also be constructed by combining user behaviors such as liking, commenting, collecting, and creating NFT credentials, but this application embodiment does not limit this.

[0070] Optionally, in some embodiments, the first user profile may be stored in a user profile feature vector library on a local server. Alternatively, the first user profile may also be stored in a user profile feature vector library on a cloud server. The user profile feature vector library includes pre-stored user preference embeddings, historical question and answer data, internal system behavior, semantic slot structure, and user itinerary drafts.

[0071] After receiving the travel request and the first constraint, the server constructs the first user profile based on the travel request and the first constraint.

[0072] The first user profile constructed in this application is not only based on shallow tags or historical click records, but also capable of identifying and analyzing the user's true travel intentions, behavioral patterns, and contextual depth. This results in a more accurate first user profile that can dynamically evolve. By constructing an accurate first user profile, this application can support the subsequent generation of travel plans highly adapted to different users.

[0073] S103 inputs the travel demand, the first constraint, and the second user profile into the large language model and outputs the first travel plan.

[0074] Optionally, in some embodiments, the second user profile is obtained by performing semantic retrieval processing and evidence enhancement processing on the first user profile.

[0075] Optionally, in some embodiments, the large language model is used to generate personalized travel plans. The large language model is a pre-trained model based on deep learning techniques and employing a Transformer architecture, capable of understanding and generating natural language. The large language model can be stored on a local server or a cloud server.

[0076] Optionally, in some embodiments, the large language model of this application uses users' historical travel preferences (e.g., academic, conference, business, medical, and educational study tours), flight and travel inventory structure, landmark and scenic spot knowledge, and training data covering visas, access rules, etc., for continuous training and instruction fine-tuning. The large language model of this application can be geared towards high-end tourism and complex travel scenarios, and by training the large language model with more comprehensive and professional data, it can recognize the real travel semantics under multiple users, multiple destinations, and multiple constraints.

[0077] It should be noted that the training process of the above-mentioned large language model can refer to the training process of the corresponding model in the existing technology, and will not be elaborated here.

[0078] It should also be noted that the historical travel preferences of the aforementioned users (such as academic, conference, business, medical and educational study tours, etc.), flight and travel inventory structure, landmark and scenic spot knowledge, and training data covering visas, passage rules, etc. are data pre-stored in the server.

[0079] Optionally, in some embodiments, the large language model of this application can work in conjunction with an executable itinerary domain-specific language (Itinerary DSL), so that while generating an executable first travel plan and explanatory basis, the large language model outputs budget constraints and resource dependencies, so as to subsequently link with the constraint solver in the server for feasibility verification and automatic revision.

[0080] Among them, Executable Trip Domain-Specific Language (EPL) is a small, executable programming language designed specifically for the field of travel trip planning. The goal of EPL is to translate user travel needs and initial constraints into scripts that can be directly executed by the server, thereby automating or dynamically adjusting the generated initial travel plan.

[0081] Optionally, in some embodiments, the large language model of this application can integrate the multimodal travel needs and first constraints input by the user, thereby enabling functions such as tour guide certificate verification, point of interest recognition (POI), ticket accounting and location trajectory verification to be implemented on the server.

[0082] Optionally, in some embodiments, the large language model of this application can utilize enhanced tool-use and proxy reasoning to enable the large language model to cyclically call the retrieval, inventory, compliance and pricing engines under the scheduling of the orchestrator in the server, and to reflectively correct intermediate conclusions, so as to ensure that the first travel plan output by the large language model has higher feasibility and traceability.

[0083] Alternatively, in another embodiment, the large language model of this application embodiment can display a summary and citation of traceable evidence related to the first trip plan on the server's display screen, thereby enabling users to more intuitively determine the feasibility of the first trip plan.

[0084] Optionally, in some embodiments, the large language model of this application may adopt ChatGPT or OpenAI, and this application does not limit this. Furthermore, the large language model is typically a decoder-based architecture or an encoder-based architecture.

[0085] Optionally, in some embodiments, the first travel plan is a travel plan with user-personalized tags.

[0086] Optionally, in some embodiments, the first travel plan includes at least the user's personalized itinerary, rankings of multiple scenic spots, candidate guides matched for each scenic spot, tickets for each scenic spot, visitor flow for each scenic spot, weather data, and road condition data leading to each scenic spot.

[0087] It should be noted that the above-described first travel plan is merely an example and is not intended to limit the scope of this application. For instance, the first travel plan may also include recommended travel time, recommended attraction addresses, and various basic information about the attractions. The basic information about the attractions should at least include their location, name, images, travel routes, transportation connections, and opening hours.

[0088] It should also be noted that the first trip plan can be stored in the user trip draft database on a local server or in the user trip draft database on a cloud server. This application embodiment does not limit this.

[0089] Optionally, in some embodiments, after generating the first travel plan, the server can generate an NFT credential for the first travel plan and manage its travel records and rights credentials through a wallet embedded in the server.

[0090] It should be noted that, in addition to the initial travel plan, the user's itinerary draft database may also include other data, which is not limited in this embodiment. For example, the user's itinerary draft database may also include user asset and interaction record chains, user-generated NFT credentials, and proof-of-travel (PoT), etc.

[0091] Optionally, in some embodiments, before inputting the second user profile into the large language model, the server uses the Retrieval Augmented Generation (RAG) method to perform semantic retrieval processing and evidence enhancement processing on the first user profile to obtain the second user profile.

[0092] Optionally, in some embodiments, the retrieval in the retrieval enhancement generation method refers to encoding the user's first user profile into a first user profile feature vector, performing relevant searches in knowledge graphs, legal databases, visa databases, destination and point-of-interest identification databases, inventory and price streams, user profile and historical preference vector databases pre-stored in the server, and external databases of the server, and selecting evidence fragments. The retrieval can employ dense vector retrieval, sparse retrieval, or hybrid retrieval.

[0093] For example, the server performs semantic retrieval on the first user profile in the user profile feature vector library of the knowledge graph.

[0094] Optionally, in some embodiments, the augmentation in the retrieval augmentation generation method refers to augmenting the retrieved evidence fragments with context before further processing them in the large language model.

[0095] Optionally, in some embodiments, the generation in the retrieval enhancement generation method refers to the large language model generating the corresponding first travel plan based on the aforementioned evidence fragments and the second user profile.

[0096] Based on the above description, the server inputs the travel demand, the first constraint, the second user profile, and the retrieved evidence into the large language model for corresponding processing, and then outputs the first travel plan. The first travel plan in this embodiment of the application better reflects the user's actual travel needs, and the first travel plan covers data from various aspects related to the user's travel. Based on more comprehensive data, it can provide the user with a better travel experience.

[0097] Considering that users' travel needs may change temporarily, users can further input their changed travel needs on the server's first interface when their travel needs need to be changed.

[0098] Optionally, in some embodiments, after generating the corresponding first travel plan and receiving the user's changed travel request, the server can update the first travel plan based on the user's changed travel request to obtain a second travel plan. This results in a second travel plan that is more closely matched to the user's actual travel needs, further enhancing the user's travel experience and making the interaction more flexible.

[0099] It should be noted that the second trip plan can be stored in a database on a local server or in a database on a cloud server; this application embodiment does not limit this.

[0100] Optionally, in some embodiments, to construct a more accurate user profile, after obtaining the second travel plan, the server can update the first user profile based on the second travel plan to obtain a third user profile. This results in a more accurate user profile, which in turn provides more accurate data for subsequently generating corresponding travel plans for the user.

[0101] It should be noted that the third user profile can be stored in the user profile feature vector library on a local server or the user profile feature vector library on a cloud server, and this application embodiment does not limit this.

[0102] Optionally, in some embodiments, after generating a travel plan (e.g., a first travel plan or a second travel plan), in response to a user's posting of a community task, the server may generate a reputation score and token incentives.

[0103] The travel planning method in this embodiment receives user-input travel needs and first constraints to obtain multimodal travel needs and further constraints on the user's travel itinerary. Furthermore, based on the obtained travel needs and first constraints, a first user profile is constructed, which integrates multimodal travel needs and personalized itinerary constraints to understand and identify the user's true travel intentions, thereby accurately constructing the first user profile. Even further, by performing semantic retrieval processing and evidence enhancement processing on the first user profile, a second user profile is obtained. Based on the above description, inputting the travel needs, first constraints, and second user profile into a large language model and outputting a first travel plan can generate a more comprehensive, accurate, and consistent first travel plan that reflects the user's actual itinerary.

[0104] Therefore, by introducing a large language model and a multimodal input mechanism, users can express their travel needs using natural language and images, greatly improving the naturalness of the interaction. Furthermore, the large language model can identify and parse users' complex real travel needs, and based on these needs, can construct high-precision user profiles (e.g., the first user profile). This results in more accurate and comprehensive travel planning based on these high-precision user profiles (e.g., the first travel plan), enabling more personalized user itineraries to be flexibly generated while dynamically interacting with users.

[0105] Furthermore, compared to existing technologies, this application is more intelligent, more personalized, and produces more feasible travel plans. Moreover, this application can cover scenarios encompassing various travel preferences (e.g., business or academic travel), and can generate highly relevant travel plans (e.g., initial travel plans) for each scenario.

[0106] It should be noted that the travel planning generation method in this application embodiment can also be integrated into an application, which can be installed on any electronic device (such as a mobile phone, smartwatch, smart glasses and other wearable devices), so that users can complete travel planning through the corresponding electronic device.

[0107] Furthermore, in this application embodiment, the travel planning generation method is integrated into an application. When this application is installed on the target electronic device, the aforementioned large language model requiring computational processing, the retrieval enhancement generation method, and various data requiring storage (e.g., travel preferences, first constraints, first user profile, and first travel plan) can all be stored in a cloud server communicatively connected to the target electronic device. The target electronic device can be invoked through modules within the target electronic device (e.g., the semantic interaction and trip construction module) using an Application Programming Interface (API). For example, the target electronic device can invoke the large language model to complete the corresponding processing through the semantic interaction and trip construction module using the large language model interface and model context protocol in the API. This allows the target electronic device to be used only for intent capture and privacy preprocessing, without undertaking core inference and generation, thus reducing the energy consumption of the target electronic device and improving its processing performance.

[0108] The above combination Figure 1 and Figure 2 This paper details the specific implementation process of a travel planning generation method provided in the embodiments of this application. The following section, in conjunction with... Figure 3 This application provides a detailed description of a travel planning generation system based on embodiments.

[0109] Please see Figure 3 , Figure 3 A structural block diagram of a travel planning generation system provided in an embodiment of this application is shown.

[0110] like Figure 3 As shown, a travel planning generation system 200 in this embodiment includes a receiving module 201 and a processing module 202. Furthermore, the receiving module 201 and the processing module 202 are communicatively connected.

[0111] The receiving module 201 is used to receive the user's input travel requirements and first constraint conditions.

[0112] The travel demand can take one or more of the following modalities: voice, text, image, and video; the first constraint is used to limit the user's personalized itinerary.

[0113] It should be noted that the definitions of the terms used here can be found in the relevant definitions in S101 above, and will not be repeated here.

[0114] Processing module 202 is used to construct a first user profile based on travel demand and the first constraint.

[0115] It should be noted that the definitions of the terms used here can be found in the relevant definitions in section S102 above, and will not be repeated here.

[0116] The processing module 202 is also used to input travel demand, first constraint conditions and second user profile into the large language model and output the first travel plan.

[0117] The second user profile is obtained by semantic retrieval and evidence enhancement processing of the first user profile. A large language model is used to generate personalized travel plans. The first travel plan is a travel plan with personalized user tags.

[0118] It should be noted that the definitions of the terms used here can be found in the relevant definitions in S103 above, and will not be repeated here.

[0119] Optionally, in some embodiments, the processing module 202 is further configured to perform semantic retrieval processing and evidence enhancement processing on the first user profile using a retrieval enhancement generation method to obtain a second user profile.

[0120] Optionally, in some embodiments, after generating the corresponding first travel plan, the receiving module 201 is also used to receive the first travel plan.

[0121] Alternatively, in some embodiments, such as Figure 3 As shown, a travel planning generation system according to an embodiment of this application may further include an update module 203, which is communicatively connected to the processing module 202.

[0122] The update module 203 is used to update the first travel plan based on the user's input travel needs to obtain the second travel plan.

[0123] The update module 203 is also used to update the first user profile based on the second travel plan to obtain the third user profile.

[0124] An embodiment of this application provides a travel planning generation system that, by introducing a large language model and a multimodal input mechanism, allows users to express their travel needs using natural language and images, greatly enhancing the naturalness of the interaction. Furthermore, the large language model can identify and parse the user's actual travel needs, and based on these needs, can construct a high-precision user profile (e.g., a first user profile). This results in more accurate and comprehensive travel plans (e.g., a first travel plan) derived from the high-precision user profile, enabling the flexible generation of more personalized user itineraries while dynamically interacting with the user.

[0125] It should be noted that the travel planning system of this application can be deployed on a local server, or on a server and an electronic device; this application embodiment does not limit this. Furthermore, the server in this application embodiment can be a local server or a cloud server.

[0126] based on Figures 1 to 3 The following description, combined with Figure 4 The following example illustrates the specific implementation process of the travel planning generation system in this application.

[0127] Assume that the travel planning system 200 of this application is deployed on a server and electronic devices.

[0128] Please see Figure 4 , Figure 4 A schematic block diagram of a travel planning generation system provided in an embodiment of this application is shown. Figure 4 As shown, the travel planning generation system 200 of this application may include a server 300 and an electronic device 400. The server 300 and the electronic device 400 are communicatively connected. The electronic device 300 includes a user-facing subsystem, a tour guide service-facing subsystem, and a government management-facing subsystem. Furthermore, data interaction and auditing between the various subsystems are all implemented through the server 300.

[0129] like Figure 4 As shown, server 300 includes storage module 301 and large language model module 302.

[0130] Storage module 301 is used to store various types of data. These types of data include, but are not limited to: travel demand, primary constraints, user profiles (e.g., primary user profile), travel plans (e.g., primary travel plans), large language models, large language model interfaces, large language model context protocols, knowledge bases, meeting data, and user profile feature vector libraries, etc.

[0131] The large language model module 302 is used to build accurate user profiles and further output precise travel plans based on these accurate user profiles.

[0132] The large language model module 302 is also used to perform intent understanding, task decomposition, tool orchestration, and result authentication on instructions from electronic devices using agentic AI technology. This enables reliable matching of tour guide service resources, inventory linkage, and closed-loop fulfillment.

[0133] like Figure 4 As shown, the electronic device 400 includes a login module 401, a semantic interaction and itinerary construction module 402, a tool module 403, a map module 404, a social platform module 405, an online payment module 406, a multi-agent module 407, a ranking module 408, an intelligent tour guide module 409, a tour guide matching module 410, a public opinion monitoring module 411, a popular attraction prediction module 412, an attraction supervision module 413, and a traffic monitoring module 414.

[0134] Optionally, such as Figure 4 As shown, the login module 401, semantic interaction and itinerary construction module 402, tool module 403, map module 404, social platform module 405, online payment module 406, multi-agent module 407, and ranking module 408 are modules in the user-facing subsystem.

[0135] The login module 401 is used for user authentication and account management. Specifically, the login module 401 is used to implement registration and login, credential verification, and risk control verification. For example, user authentication can be performed using basic information such as the user's username, email address, password, gender, and travel preferences.

[0136] If a user has already registered an account, they can log in through the login module 401. After successful login, the login module 401 also retrieves user historical profiles and historical travel preferences from the storage module 301 of the server 300, and uses the obtained user historical profiles and historical travel preferences as input to provide a more comprehensive and accurate data foundation for the generation of subsequent travel plans.

[0137] The semantic interaction and trip construction module 402 is used to collect users' travel needs. Specifically, the semantic interaction and trip construction module 402 is used to accept input from multiple modalities such as voice, text, images, and video.

[0138] After a user successfully logs in and the login module 401 successfully retrieves data such as the user's historical profile and historical travel preferences, the user inputs multimodal travel needs through semantic interaction and the trip construction module 402.

[0139] The semantic interaction and trip construction module 402 is also used to transmit the received multi-modal travel requests to the storage module 301 of the server 300 for storage.

[0140] The semantic interaction and trip construction module 402 is also used to call the large language model in the large language model module 302 through the corresponding large language model interface set in the server 300 to complete the slot parsing of the user's travel needs, and build a user profile (e.g., the travel needs in S101) based on the travel intent after slot parsing.

[0141] The semantic interaction and trip construction module 402 is also used to call the retrieval enhancement generation method in the large language model module 302 to complete the retrieval of the knowledge graph in the server 300 and the external database, obtain evidence fragments, and upload the evidence fragments to the large language model module 302. This enables the large language model module 302 to generate the user's travel plan (e.g., the first travel plan output in S103) based on the user profile and evidence fragments.

[0142] After receiving the travel plan transmitted by the large language model module 302, the semantic interaction and itinerary construction module 402 is also used to select on-chain evidence and manage the user's travel plan in the long term.

[0143] The semantic interaction and itinerary building module 402 is also used for itinerary editing, versioning, and storage. Specifically, itinerary editing can include at least adding or deleting attractions, changing modes of transportation and hotels, and adjusting time windows.

[0144] In response to the user's itinerary editing and confirmation of changes in the semantic interaction and itinerary construction module 402, the semantic interaction and itinerary construction module 402 can trigger the large language model and retrieval enhancement generation method in the large language model module 302 to perform local recalculation and consistency verification in order to update the user's travel plan (e.g., second travel plan).

[0145] Tool module 403 is used for professional resource matching and data injection. Tool module 403 includes tools for academic conferences and personal planning, as well as tools for medical travel and personal planning. Tool module 403 is equivalent to the first constraint in S101.

[0146] After the user successfully inputs multimodal travel needs through semantic interaction and the trip construction module 402, the user can further limit the trip through the tool module 403.

[0147] In response to the user's relevant operations in the tool module 403, the semantic interaction and trip construction module 402, after receiving the output of the tool module 403, can use the output of the tool module 403 as new evidence or constraints to trigger the process of calling the large language model module 302 to generate a trip plan (e.g., the first trip plan).

[0148] After generating a travel plan (e.g., the first travel plan), the user can modify or re-enter their travel needs and the already generated travel plan through the semantic interaction and trip building module 402. This will trigger the semantic interaction and trip building module 402 to call the large language model in the large language model module 302 again to regenerate the travel plan and obtain the updated travel plan (e.g., the second travel plan).

[0149] It should be noted that the above process of regenerating the travel plan is equivalent to the process of updating module 203 mentioned above.

[0150] The map module 404 is used for spatial visualization and travel assistance. Specifically, the map module 404 displays the location of attractions, their opening hours, travel routes, and transportation connections. The map module 404 interacts in real time with the semantic interaction and itinerary construction module 402, and can update map elements (such as the location of attractions and travel routes) at any time based on updates to travel plans.

[0151] The social platform module 405 is used for content sharing and interaction, as well as data feedback for governance. Specifically, content sharing and interaction include sharing text and videos, liking, commenting, collecting, and reposting. Data feedback for governance provides data support for tour guide matching in the tour guide matching module 410. Data feedback for governance also provides data support for public opinion monitoring in the public opinion monitoring module 411 and visitor flow prediction in the popular tourist attraction prediction module 412.

[0152] It should be noted that the social platform module 405 allows users to browse relevant governance data without logging in.

[0153] The online payment module 406 is used for payment settlement and on-chain rights confirmation. Specifically, the types of payment settlement can include membership top-ups, service purchases, and crypto asset settlements. For example, the online payment module 406 can include a wallet or asset sub-module.

[0154] The online payment module 406 is also used to initiate on-chain rights confirmation and contract calls for itinerary plan NFT certificates and travel certificate PoTs. The generated on-chain certificates are displayed and managed in the semantic interaction and itinerary construction module 402.

[0155] The multi-agent module 407 is used for one-click planning and tour guide agent collaboration. The multi-agent module 407 is also used to trigger one-click generation of travel plans and intelligent tour guides for popular scenic spots. Furthermore, the multi-agent module 407 is used in server 300 to schedule agents for tasks such as retrieval, price comparison, compliance, and tour guide collaboration to complete complex travel planning and execution suggestions.

[0156] Ranking module 408 is used for generating and displaying popular lists. Specifically, ranking module 408 aggregates indicators such as social interaction, visitor flow, ratings, and fulfillment quality to generate popular scenic spot lists, popular food lists, and route lists. Ranking module 408 also provides anti-fraud and de-stigmatization processing for popular lists.

[0157] It should be noted that multiple modules in the aforementioned user-facing subsystems support natural language input, multi-turn dialogue optimization, NFT credential generation, and community incentives. Through the synergistic effect of these modules in the user-facing subsystems, and by employing mechanisms for perceiving multimodal travel needs and understanding language, the travel planning generation system of this embodiment can improve its ability to analyze users' real travel needs. Based on flexible dynamic interaction, it further constructs a high-precision user profile, thereby preparing for the generation of adapted travel plans.

[0158] Optionally, such as Figure 4 As shown, the intelligent tour guide module 409 and the tour guide matching module 410 are modules in a subsystem oriented towards the tour guide service side.

[0159] The intelligent tour guide module 409 is used for service execution and full-process recording. Specifically, the intelligent tour guide module 409 is used to complete on-chain registration and settlement of key events such as order acceptance, cancellation, confirmation of completion, and evaluation through a blockchain contract gateway, thereby forming an auditable performance ledger and revenue distribution record. For example, after a user completes the corresponding itinerary based on travel planning, the intelligent tour guide module 409 is also used to confirm the completion of settlement and generate digital credentials such as a travel certificate (PoT) and a visit commemorative badge (or attendance certificate) for record-keeping.

[0160] The intelligent tour guide module 409 is also used to generate a task planning tree based on the complexity of the user's itinerary, breaking down complex itineraries into several sub-tasks. Furthermore, after breaking down complex itineraries into several sub-tasks, the intelligent tour guide module 409 works in conjunction with the multi-agent module 407 in the user-facing subsystem to complete sub-tasks such as retrieval, price comparison, compliance verification, and on-site guidance.

[0161] The intelligent guide module 409 is also used for state synchronization and feedback calibration via a dialogue interface. Specifically, the intelligent guide module 409 is also used to synchronize the completion status of several sub-tasks via a dialogue interface.

[0162] The intelligent tour guide module 409 is also used to integrate third-party VR tour guides or augmented reality (AR) tour guide services and AI explanation services. This allows virtual explanations, 3D maps and scenes, and AI explanation assistants to be embedded into the itinerary nodes. In case of abnormal situations such as congestion, delays, or park closures, alternative solutions can be automatically triggered, such as triggering third-party VR tours, thereby improving the user experience.

[0163] The intelligent tour guide module 409 is also used to call interfaces between online travel agencies and other resource providers. This allows it to aggregate and dynamically monitor inventory such as air tickets, hotels, tickets, and transportation, enabling inventory locking and price updates. Consequently, it ensures the integrity and timeliness of tour guide services, forming a fulfillment monitoring and closed-loop feedback system.

[0164] The tour guide matching module 410 is used to construct and maintain a semantic feature index for tour guides, and to vectorize tour guide information and service tags based on BERT and Sentence-BERT. Furthermore, the tour guide matching module 410 is used to perform reverse retrieval and reordering based on factors such as attractions, theme preferences, language habits, and time arrangements in the user's travel needs, using BERT and Sentence-BERT, to generate a set of candidate tour guides (e.g., real tour guides or intelligent tour guides) and the optimal matching combination. Even further, the tour guide matching module 410 can synchronize the generated set of candidate tour guides and the optimal matching combination to the user-facing subsystem via server 300.

[0165] Understandably, BERT and Sentence-BERT are both encoder-based bidirectional Transformers. Both BERT and Sentence-BERT are used to map text into fixed-length vectors to support semantic retrieval or similarity matching.

[0166] The tour guide matching module 410 is also used to call the RAG pipeline in the large language model module 302, and jointly call the data sources in the internal knowledge graph of the server 300 and the external database to complete the evidence. Furthermore, the tour guide matching module 410 is also used to output interpretable matching reasons and constraint hit status based on the completed evidence fragments, which facilitates the tour guide's subsequent confirmation and order acceptance.

[0167] It should be noted that the aforementioned subsystems for tour guide services integrate smart contracts, RAG semantic reverse matching, multi-agent task decomposition, and immersive technologies to achieve intelligent scheduling and trust authentication, thereby improving resource distribution efficiency and further enhancing service quality. Furthermore, these subsystems for tour guide services integrate Web3 technology, decentralized identity (DID), and incentive mechanisms to achieve trusted matching of tour guide resources, service assetization, user incentives, and decentralized governance.

[0168] Optionally, such as Figure 4 As shown, the public opinion monitoring module 411, the popular tourist attraction prediction module 412, the tourist attraction supervision module 413, and the traffic monitoring module 414 are modules in the subsystem oriented towards the government management side.

[0169] The public opinion monitoring module 411 is used for the collection, cleaning, and aggregation of interactive data and content sentiment on social media platforms. Specifically, the public opinion monitoring module 411 is used to identify public opinion topics, sentiment polarity, and abnormal upward trends. The public opinion monitoring module 411 is also used to generate early warnings and handling suggestions based on the identified public opinion topics, sentiment polarity, and abnormal upward trends. The public opinion monitoring module 411 is also used to call the large language model in the large language model module 302 to complete multi-source text disambiguation, key point extraction, emergency script drafting, and automatic generation of daily and weekly reports. The public opinion monitoring module 411 is also used to push risk tags and early warning events through the server 300 to the map module 404 in the user-facing subsystem for visualization.

[0170] The popular tourist attraction prediction module 412 is used for time-series modeling and hotspot identification of tourist attractions and activities, and combines factors such as holidays, weather, marketing campaigns, and historical visitor flow to form predictions. The module 412 also calls upon the large language model in the large language model module 302 to output peak and trough visitor flow and the probability of becoming a hotspot, and provides strategy suggestions for time-slot reservations, zoned traffic diversion, and activity promotion. The module 412 further transmits these strategy suggestions to the map module 404 via server 300 for display. Finally, the module 412 transmits these strategy suggestions to the tourist attraction monitoring module 413 for monitoring.

[0171] The scenic spot supervision module 413 is used for scenic spot-related governance and compliance rules. These rules may include visitor flow threshold management, ticket verification and reconciliation, blacklists and whitelists, complaint arbitration, and emergency response. The module also implements data access control and sharing strategies, employing decentralized identities and verifiable credentials to define the scope, granularity, and timeliness of data sharing based on Decentralized Autonomous Organization (DAO) governance rules, and ensuring controllable rights through access control and audit logs. Furthermore, the module 413 tags and measures data contributors according to policies, incentivizes contracts and settlement services to generate verifiable incentive records, and transmits these records back to the map module 404 via server 300 for visualization. Finally, the module generates flow-limited zones based on scenic spot carrying capacity thresholds and compliance rules, and allows for the issuance of governance commands such as increased security and tour guide deployment.

[0172] The traffic monitoring module 414 is used to predict road accessibility and congestion. Specifically, by accessing traffic flow, road condition events, and public transportation data stored in an external database, the traffic monitoring module 414 can output time-segmented scheduling suggestions and temporary traffic organization plans. The traffic monitoring module 414 also calls the large language model in the large language model module 302 to generate early warnings and works in conjunction with the scenic spot monitoring module 413 to output suggestions for on-site traffic management and shuttle resource allocation. The traffic monitoring module 414 also transmits the output early warnings, on-site traffic management, and shuttle resource allocation suggestions back to the map module 404 via the server 300 for visualization.

[0173] It should be noted that the aforementioned government-management-oriented subsystem, by utilizing various data from the user-oriented subsystem and the tour guide service subsystem stored on server 300, as well as public data sources stored in external databases, can provide real-time and visualized governance assistance and strategy output for scenic area scheduling and government management. The government-management-oriented subsystem can provide key indicators such as tourist distribution, service satisfaction, tour guide workload, holiday popularity, and congestion levels. Several key indicators can be displayed in a split-screen format with map routes in map module 404. Furthermore, the government-management-oriented subsystem supports one-click issuance of instructions for time-slot reservations, additional tour guides, and vehicle traffic management, and can also retain strategy playback for effect evaluation.

[0174] It should be noted that the aforementioned modules such as the large language model module 302, semantic interaction and itinerary construction module 402, multi-agent module 407, intelligent tour guide module 409, tour guide matching module 410, public opinion monitoring module 411, and popular attraction prediction module 412 are equivalent to... Figure 3The processing module 202 shown above. The login module 401, tool module 403, map module 404, social platform module 405, online payment module 406, ranking module 408, scenic spot monitoring module 413, and traffic monitoring module 414 are equivalent to... Figure 3 The receiving module 201 is shown. The semantic interaction and process construction module 402, the login module 401, and the tool module 403 can also form an update module 203. Specifically, the above modules can be divided into a processing module 202 or a receiving module 201 according to their functions, which will not be elaborated here.

[0175] Based on the above description, the travel planning generation system of this application embodiment is a comprehensive system oriented towards the user side, tour guide service side, and government management side, which can comprehensively improve the intelligence level and governance capabilities of the tourism industry. The user-side subsystem, tour guide service side subsystem, and government management side subsystem work collaboratively around a closed-loop chain of "semantic input - personalized recommendation - resource matching - service execution - data governance." It can use the large language model in the large language model module 302 as its core, calling the large language model through the semantic interaction and itinerary construction module 402 and the multi-agent module 407, and combining the retrieval enhancement generation method with the internal knowledge graph and external database of the server 300 to achieve semantic understanding, RAG enhanced recommendation, and AI itinerary generation, thereby obtaining travel plans that conform to the user's personalized itinerary. Secondly, the tour guide service side subsystem is supported by Web3 intelligent agent technology and contract gateway technology, and is triggered through the intelligent tour guide module 409, tour guide matching module 410, and online payment module 406 to realize service rights confirmation, tour guide matching, and user incentives. In addition, the government management subsystem is supported by urban data. Through the synergistic effect of the public opinion monitoring module 411, the popular tourist attraction prediction module 412, the tourist attraction supervision module 413, and the traffic monitoring module 414, and combined with the governance data feedback provided by the social platform module 405 in the user-side subsystem, it can jointly achieve governance support and visualized decision-making.

[0176] In addition, the travel planning in the travel planning generation system of this application embodiment includes the user's itinerary and tour guide services, which can ensure that tour guide service resources are accurately distributed to target users.

[0177] For example, this application provides a readable storage medium storing a computer program and processor calling instructions, which cause a server to implement the method in the preceding embodiments when executed.

[0178] For example, this application provides a chip system applied to an electronic device including a memory, a display screen, and sensors; the chip system includes: one or more interface circuits and one or more processors; the interface circuits and processors are interconnected via lines; the interface circuits are used to receive signals from the memory and send signals to the processors, the signals including computer code or instructions stored in the memory; the processors invoke instructions to cause the server to execute the methods in the preceding embodiments.

[0179] For example, this application provides a computer program product that, when run on a computer, causes the computer to implement the methods described in the preceding embodiments.

[0180] In the above embodiments, all or part of the functionality can be implemented by software, hardware, or a combination of software and hardware. When implemented using software, it can be implemented wholly or partially in the form of a computer program product. A computer program product includes one or more computer codes or instructions. When the computer program code or instructions are loaded and executed on a computer, all or part of the flow or functionality according to this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer code or instructions can be stored in a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0181] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. The above embodiments are only used to illustrate the technical solutions of the present invention, facilitating understanding and implementation by those skilled in the art. The scope of protection of the present invention is not limited to the above embodiments; any equivalent substitutions or improvements made under the guidance of the present invention should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for generating travel plans, characterized in that, The method includes: The system receives user input of travel requests and a first constraint; the travel requests may take one or more of the following modalities: voice, text, image, and video; the first constraint is used to limit the user's personalized itinerary. Based on the travel demand and the first constraint, a first user profile is constructed; The travel demand, the first constraint, and the second user profile are input into a large language model to output a first travel plan; the second user profile is obtained by performing semantic retrieval processing and evidence enhancement processing on the first user profile; the large language model is used to generate personalized travel plans; the first travel plan is a travel plan with personalized user tags.

2. The method according to claim 1, characterized in that, Before inputting the second user profile into the large language model, the method further includes: The first user profile is processed using semantic retrieval and evidence enhancement methods to obtain the second user profile.

3. The method according to claim 1, characterized in that, After generating the corresponding first travel plan, the method further includes: Based on the user's input of travel needs, the first travel plan is updated to obtain the second travel plan; Based on the second travel plan, the first user profile is updated to obtain the third user profile.

4. The method according to claim 1 or 3, characterized in that, The first travel plan includes at least the user's personalized itinerary, rankings of multiple scenic spots, candidate tour guides matched for each scenic spot, tickets for each scenic spot, visitor flow for each scenic spot, weather data, and road condition data leading to each scenic spot.

5. The method according to any one of claims 1 to 4, characterized in that, The travel demand includes at least one or more of the following: travel destination, travel preference, travel date, number of travelers, mode of transportation, travel budget, and travel type; the travel preference includes at least one or more of the following: academic travel, business travel, medical travel, educational travel, and tourism travel; the travel type includes tourism, shopping, conferences, educational study tours, or team building activities.

6. A travel planning generation system, characterized in that, The system includes: The receiving module is used to receive the user's input travel request and first constraint; the travel request can take one or more of the following modalities: voice, text, image and video; the first constraint is used to limit the user's personalized itinerary; The processing module is used to construct a first user profile based on the travel demand and the first constraint. The processing module is further configured to input the travel demand, the first constraint, and the second user profile into a large language model and output a first travel plan; the second user profile is obtained by performing semantic retrieval processing and evidence enhancement processing on the first user profile; the large language model is used to generate personalized travel plans; the first travel plan is a travel plan with personalized user tags.

7. The system according to claim 6, characterized in that, The processing module is also used to perform semantic retrieval processing and evidence enhancement processing on the first user profile using a retrieval enhancement generation method to obtain the second user profile.

8. The system according to claim 6, characterized in that, The system also includes an update module; The update module is used to update the first travel plan based on the travel needs input by the user, so as to obtain a second travel plan. The update module is further configured to update the first user profile based on the second travel plan to obtain a third user profile.

9. The system according to claim 6, characterized in that, After generating the corresponding first travel plan, the receiving module is also used to receive the first travel plan; The first travel plan includes at least the user's personalized itinerary, rankings of multiple scenic spots, candidate guides matched for each scenic spot, tickets for each scenic spot, visitor flow for each scenic spot, weather data, and road condition data leading to each scenic spot. The travel demand includes at least one or more of the following: travel destination, travel preference, travel date, number of travelers, mode of transportation, travel budget, and travel type; the travel preference includes at least one or more of the following: academic travel, business travel, medical travel, educational travel, and tourism travel; the travel type includes tourism, shopping, conferences, educational study tours, or team building activities.

10. A server comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-5.

Images