Tourist route planning system based on AI large model

Abstract

The invention discloses a tourist route planning system based on an AI large model, and relates to the field of computer technology application. In order to solve the defects of lack of accuracy, reasonability and reference of recommendation results caused by tourist route recommendation in the prior art, the technical scheme provided by the invention is as follows: the tourist route planning system based on the AI large model comprises a scenic spot library construction module, a tourist route planning module and a tourist route planning module, the scenic spot database module is used for generating scenic spot information of each region through an AI large model and establishing a scenic spot database containing unique identifiers; the user input module is used for receiving travel demand information input by a user; the semantic analysis module is used for analyzing travel demands input by natural languages and generating structured data; the route generation module is used for generating a tour route from a scenic spot library according to the structured data; and the visual display module is used for displaying the text information and the map information of the tour route. The method is suitable for intelligent travel planning work of automatically generating a high-quality and visual travel route according to individual requirements of the user.

Description

TECHNICAL FIELD

[0001] The application relates to the field of computer technology, and particularly relates to a tourism route planning system based on an AI large model. BACKGROUND

[0002] With the continuous improvement of people's living standards and the increasing convenience of transportation, tourism has become an indispensable part of people's lives. Tourism not only carries multiple functions such as leisure and relaxation, cultural exploration, but also promotes the development of the economy and society of the destination. In order to improve travel efficiency and tourism experience, users usually rely on multiple information channels to obtain destination recommendations and tourism route planning suggestions when planning a trip.

[0003] At present, there are mainly the following technical means: Tourism application program based on manual editing Tourism Apps on the market, such as Ma Bumble, Where to Go, and Ctrip, usually provide guide content edited by local guides or professional tour guides. Although this type of application has a certain reference value, the recommended content has strong commercial orientation, and is often aimed at selling tourism products and guiding consumption as the main target, lacking neutrality and individualization, and it is difficult to generate a unique itinerary according to the interests and preferences of different users.

[0004] Information retrieval method based on search engine Users also often use search engines such as Baidu and Google to input keywords to find tourism guides and route information by themselves. Although this method has rich information, the search results come from various sources and have uneven quality, and users need to spend a lot of time screening useful content, and for users who are not familiar with the tourist destination, the information integration is difficult.

[0005] Recommendation method based on general AI question and answer robot With the development of artificial intelligence technology, AI question and answer robots based on large models (such as ChatGPT, Wenxin Yiyang, etc.) are gradually used in tourism route planning scenarios. Users can propose requirements through dialogue, and the AI returns recommended routes. However, it is found in practice that such recommendations often have problems such as "unreasonable routes, rigid content, lack of customization", and because of the "hallucination" problem of AI large models, the generated scenic spot information may not be accurate or does not exist in reality, affecting the user's trust in using it.

[0006] Lack of structured user preference input and systematic route generation mechanism Most of the existing systems are driven by "destination + days" to generate routes, which are difficult to combine user interest preferences (such as liking history, culture, natural scenery, etc.), existing travel plans (multiple cities), or personalized needs (traveling with children, avoiding crowds, focusing on niche attractions, etc.) for deep customization. At the same time, the lack of structured scenic spot database support also limits the accuracy and scalability of recommendations.

[0007] In summary, the existing technology has the defects of lack of structured user demand analysis, personalized preference adaptation and high-quality scenic spot database support in tourism route recommendation, resulting in lack of accuracy, rationality and reference of recommended results. SUMMARY

[0008] In order to solve the defects of lack of structured user demand analysis, personalized preference adaptation and high-quality scenic spot database support in tourism route recommendation in the prior art, resulting in lack of accuracy, rationality and reference of recommended results, the technical scheme provided by the present application is: An AI large model-based tourism route planning system, comprising: A scenic spot library construction module for generating scenic spot information in each region by an AI large model and establishing a scenic spot library containing unique identifiers; A user input module for receiving user input tourism demand information; A semantic analysis module for analyzing natural language input tourism demand and generating structured data; A route generation module for generating a tourism route from the scenic spot library according to the structured data; A visual display module for displaying text information and map information of the tourism route.

[0009] Further, an preferred embodiment is provided, wherein the scenic spot library construction module classifies domestic regions by province, city and county at three levels, and assigns unique IDs to each level and each scenic spot.

[0010] Further, an preferred embodiment is provided, wherein the user input module includes a form mode and an intelligent mode, the form mode is used for inputting structured information, and the intelligent mode is used for inputting natural language questions.

[0011] Further, an preferred embodiment is provided, wherein the semantic analysis module calls an AI large model through a preset prompt template to extract the destination, the number of travel days and the preferred scenic spot type.

[0012] Further, an preferred embodiment is provided, wherein the route generation module combines the user's favorite scenic spots, popular scenic spots and niche scenic spots according to a ratio of 3 scenic spots per day to form a trip, and sorts the path according to the latitude and longitude.

[0013] Further, a preferred embodiment is provided, wherein the route generation module classifies the scenic spots into at least user-favorite scenic spots, user-unfavorite scenic spots, popular scenic spots and less-popular scenic spots, and excludes the user-unfavorite scenic spots before generating the travel itinerary.

[0014] Also provided is an AI large model-based travel route planning method, based on the system implementation, comprising: Step 1, constructing a scenic spot database, generating scenic spot information of each region through an AI large model, and storing as a data structure containing a unique identifier; Step 2, receiving user input travel demand information, including destination, travel days and keywords; Step 3, analyzing user input content, and if it is natural language, calling an AI large model to extract structured data; Step 4, filtering scenic spots from the scenic spot database according to the structured data, and composing a travel route with three scenic spots per day; Step 5, sorting the generated travel route by latitude and longitude to form a complete itinerary; Step 6, displaying the itinerary in text and map, and providing 3D view and animation playback display.

[0015] Also provided is a computer storage medium for storing a computer program, wherein when the computer program is read by a computer, the computer executes the method.

[0016] Also provided is a computer comprising a processor and a storage medium, wherein when the processor reads the computer program stored in the storage medium, the computer executes the method.

[0017] Also provided is a computer program product as a computer program, which, when executed, implements the method.

[0018] Compared with the prior art, the technical solution provided by the present application has the following advantages: By constructing a structured scenic spot database, the basic information of each scenic spot, such as administrative division, geographical location, heat value and type, is clearly recorded as a knowledge benchmark database for large model recommendation. This method significantly improves the accuracy and controllability of travel route generation, avoids the false scenic spot information generated by traditional large model "illusion", ensures that the recommended results are real and traceable, the information structure is uniform, and has higher reliability and stability compared with AI question and answer tools relying on scattered network data.

[0019] By introducing the "intelligent mode + form mode" dual-channel input mechanism, users can express their personalized travel needs through natural language questioning or structured form filling. This design balances flexibility and accuracy, addressing the limitations of traditional tourism Apps that only support standardized options and lack the ability to express complex needs. It also overcomes the shortcoming of existing AI question-and-answer tools, which often lose information when understanding complex questions.

[0020] Using AI large models for semantic understanding and specially designed question templates to format information such as destination city, travel days, and keywords, the system effectively improves the accuracy of user demand analysis. This approach enhances the system's understanding of natural language and maintains high accuracy in structured processing even when faced with ambiguous, vague, or mixed information. Compared to existing intelligent question-and-answer recommendation systems, it has a clear advantage in complex scenarios.

[0021] The "3n" structure is used to plan daily itineraries, with each day allocated to three attractions. The "user preference + popular attractions + niche attractions" ratio allocation mechanism ensures that each route is reasonable, diverse, and personalized. This generation strategy balances user preferences and popular trends, significantly improving upon traditional App recommendations, which suffer from severe homogeneity and single itineraries. It also outperforms large model recommendations, which are highly random and lack control.

[0022] A complete attraction filtering process is designed, including user "like / dislike" tagging mechanisms, heat range judgments, and attraction type matching. This multi-dimensional filtering mechanism improves recommendation quality by effectively excluding attractions that users are not interested in and avoiding duplicate arrangements. In practical applications, it significantly improves user satisfaction and reference value, overcoming the shortcomings of existing AI routes that lack memory and filtering capabilities for user preferences.

[0023] Tourist routes are visualized on the system page, with interactive methods such as map markers, 3D remote sensing animations, and flying animation videos enhancing user experience. This approach not only improves the intuitiveness and immersion of information display but also enhances user engagement and understanding of the route plan, outperforming most Apps that only provide text or static illustrations. It also breaks through the limitations of AI text output lacking visual support.

[0024] In summary, through data structuring, multi-modal input of needs, accurate semantic analysis, personalized itinerary allocation, multi-dimensional attraction evaluation mechanisms, and high-interactive visual display, the system effectively addresses the core issues of existing tourism route planning methods, such as lack of personalization, susceptibility to errors, low user engagement, and poor reference value.

[0025] This system is suitable for intelligent travel planning that automatically generates high-quality, visualized tourist routes based on user individual needs. Attached Figure Description

[0026] Figure 1 This is a schematic diagram illustrating the format of the scenic spot location information content of the present invention. Figure 2 This is a schematic diagram of the problem analysis process based on the AI ​​large model of the present invention; Figure 3 This is a schematic diagram of the tourism route planning process for a single city according to the present invention. Figure 4 : This is a schematic diagram of the scenic spot database of the present invention; Figure 5 : This is a schematic diagram illustrating the formatting of the user question parsing results of the present invention. Detailed Implementation

[0027] To make the advantages and benefits of the technical solution provided by the present invention clearer, the technical solution provided by the present invention will now be described in further detail with reference to the accompanying drawings, specifically: Implementation Method 1: This implementation method provides a travel route planning system based on an AI large-scale model, including: The attraction database construction module is used to generate attraction information for various regions through a large AI model and establish an attraction database containing unique identifiers. The user input module is used to receive travel request information input by the user. The semantic parsing module is used to parse natural language input of travel needs and generate structured data; The route generation module is used to generate tourist routes from the attraction database based on the structured data; The visualization module is used to display the text and map information of the tourist route.

[0028] The scenic spot database construction module classifies domestic regions into three levels: province, city, and county, and assigns a unique ID to each level and each scenic spot.

[0029] The user input module includes a form mode and an intelligent mode. The form mode is used to input structured information, and the intelligent mode is used to input natural language questions.

[0030] The semantic parsing module uses a preset prompt template to call the AI ​​big model to extract the destination, number of days of the trip, and preferred attraction types.

[0031] The route generation module combines three attractions per day from users' favorite attractions, popular attractions, and lesser-known attractions to form an itinerary, and sorts the routes according to latitude and longitude.

[0032] The route generation module classifies the scenic spots, including at least user-liked scenic spots, user-unliked scenic spots, popular scenic spots and unpopular scenic spots, and excludes the user-unliked scenic spots before generating the travel itinerary.

[0033] Embodiment two, this embodiment is a further detailed description of the technical solution provided in embodiment one, specifically: Step 1: Construct a scenic spot database. In this embodiment, the domestic regions are classified according to provinces (autonomous regions, municipalities directly under the Central Government), cities (prefectures) and counties (districts) in three levels, and each level of classification is given a unique identification (ID). The system uses a cyclic call AI large model interface to collect and generate scenic spots in each region in batches. Specifically, the generation interface of the AI large model is combined with the scenic spot location content format template to obtain scenic spot information from the provincial level to the municipal level and then to the county level. The scenic spot location content format is shown in the following table: Figure 1 The table includes fields such as scenic spot unique identification, scenic spot name, belonging province, belonging city, belonging county, longitude, latitude, and scenic spot introduction. Through this way, the system generates a record for each scenic spot and gives it a unique scenic spot ID, and stores the record in the scenic spot database. The scenic spot database can be initialized and created before the system is first used, or it can be dynamically updated and enriched during subsequent use. Figure 4 The table shows the completed scenic spot database according to Figure 1 The format.

[0034] Step 2: Open the system application page. The user accesses the front-end page of the system through the browser and enters the system initial interface. The system initial interface is a login / register page. After the user completes the registration and login in the interface, he / she can enter the travel route planning function module.

[0035] Step 3: Fill in the travel route planning requirements. In the system interface, the user can choose different input modes according to the requirements. If the user has specified the travel destination and the travel time period, he / she can choose the "form mode". In the form mode, the user fills in the destination city name, the travel start time and the end time (multiple destinations and time periods are supported) in the designated input box, and can fill in the interest keywords (such as food, culture, etc.) in the keyword input bar to assist the planning. Otherwise, the user can choose the "intelligent mode", describes the travel planning requirements in the text input box, and inputs the requirements in the form of natural language, for example, "three-day travel route planning for a family of three in Dalian City". After completing the requirement input, the user clicks the "create itinerary" button, and the system starts the route generation phase.

[0036] Step 4: Create a travel route plan. After the user clicks "Create Trip", the system frontend page displays a loading animation and prompts the generation progress. The system starts the route planning algorithm based on the demand information obtained in step 3. In this embodiment, when the user inputs the text demand in the intelligent mode, the system calls the AI large model to perform semantic analysis on the user demand. As shown in Figure 2 FIG. 1 1, the semantic analysis module based on the AI large model designs a series of step-by-step analysis tasks to convert the user's natural language question into structured data through a pre-designed prompt template. Specifically, the following subtasks and corresponding question templates are included: Subtask 1 (question classification): Determine whether the user input is a travel route planning demand. The question template is: "Is the question " [user input question] " related to travel route planning? If yes, please answer 1 directly; if not, please answer 0 directly." For example, for the question "Dalian City three-day tour travel route planning for a family of three", the model returns 1, indicating that the question belongs to the travel route planning demand.

[0037] Subtask 2 (destination resolution): Analyze the travel destination city. The question template is: "According to this question, what are the destination cities that the user wants to travel to? Please answer the city name directly, and if there are multiple cities, use ; to separate. For example: Beijing City; Shanghai City." In the example question, the model returns "Dalian City", which means the output destination city is "Dalian City".

[0038] Subtask 3 (trip duration resolution): Analyze the travel duration. The question template is: "According to this question, how many days does the user want to stay in the destination city? Please answer the number of days directly, and if there are multiple destination cities, use ; to separate. For example: 3; 2." For example, the model returns "3" for the example question, indicating that the trip duration is 3 days.

[0039] Subtask 4 (attraction type recommendation): Identify suitable attraction types. The question template is: "According to this question, what are the most suitable attraction types for the questioner? Please use 0 for natural scenery, 1 for museum, 2 for theme park, 3 for ecological landscape, and 4 for cultural landscape. Use numerical codes instead of text in the answer. Recommend at least 1 type and at most 3 types, separated by ;. For example: 0; 2; 4." For example, for the example question, the model may return "2; 0", indicating that theme parks and natural scenery are recommended. By sequentially executing the above analysis tasks, structured demand information can be obtained. For example, the analysis result for the example question is as follows: Figure 5As shown: the destination is "Dalian City", the trip duration is 3 days, and the attraction types are "theme parks" and "natural scenery". The above parsed results, after formatting, will be used as input for subsequent path planning algorithms. In this embodiment, the route planning module follows... Figure 3 The system demonstrates a single-city tour route planning process. First, based on the destination and attraction type, it retrieves a set of attractions from the attraction database that meet the criteria, further categorizing them into several groups: list_like (user-marked favorites), list_notlike (user-marked dislikes), list_hot (popular attractions), and list_unique (unique attractions). During retrieval and combination, the system automatically excludes attractions from the list_notlike. For example, based on the categories "theme park" and "natural scenery," popular attractions (list_hot) in Dalian might include "Tiger Beach Ocean Park, Dalian Forest Zoo, Dalian Haichang Discovery Kingdom Theme Park," while unique attractions (list_unique) might include "Dalian Luneng Ocean Hot Spring, Dalian Art Museum," etc. Next, the final allocation of attractions is determined based on the number of days (n) and the length of the list_like list. Specifically, all elements in the list_like list are prioritized, with the remaining attractions provided by the list_hot and list_unique lists. Assuming the trip duration n=3 days, when the length of list_like is a=2, the number of attractions selected from list_hot can be calculated as b=3n-a-1 (assuming a maximum of 3n attractions can be visited per destination, deducting 1 from the starting point and 'a' from existing preferred attractions). Therefore, b=3×3-2-1=6. The system then merges two attractions from list_like, six attractions randomly selected from list_hot, and one attraction selected from list_unique to form the complete itinerary list list_journey. The merged list_journey contains a total of 9 attractions as the planning result. Subsequently, the system sorts list_journey counter-clockwise based on the latitude and longitude information of each attraction to determine the sequential path of the itinerary.

[0040] Step 5: Display the results of the travel route planning. After generating the final journey list list_journey and sorting, the system formats the journey data and renders it on the front end. The system displays both a text list and a map visualization of the route planning results on the application page. The text list lists the attractions for each day in order, while the map interface displays the path and attraction markers at corresponding geographic locations. Users can view the planned route point by point using the "forward" and "backward" buttons, and view the planned route by day using the "switch day" button. Users can also click on the attraction links in the text list or the markers on the map to view detailed information about that attraction. When the user clicks the "3D" button, they can view a three-dimensional remote sensing image animation of the corresponding attraction location; clicking the "animation video" button will play a flight animation video automatically generated based on this itinerary.

[0041] Embodiment three, in combination Figures 1-5 To illustrate the present embodiment, the above-mentioned technical solutions are further described in detail through specific embodiments, specifically: In order to overcome the above-mentioned problems existing in the prior art, a tourism route planning system based on AI large model is provided.

[0042] The solution adopted to solve the above technical problems is: a tourism route planning system based on AI large model, comprising the following steps: Step 1: Build an attraction library. Divide the domestic area into provinces (autonomous regions), cities (municipalities), and counties (districts) and add a unique ID to each level of classification. Use batch loop to call the large model, generate the attractions in a certain area according to the attraction location information content format template, and add a unique ID to the attractions. The attraction location information content format is shown in Figure 1 The attraction library can be updated and enriched gradually before or during the use of the system, preparing for the subsequent steps.

[0043] Step 2: Open the system application page. Open the system interface using a browser, and after the user registers and logs in, they can start using the system to generate a travel route.

[0044] Step 3: Fill in the travel route planning requirements. According to the travel requirements, select the appropriate mode and input the corresponding questions or fill in the form.

[0045] Step 4: Create a travel route plan. Click "Create Journey" to start generating a travel route plan, and the page will display a loading animation and a progress prompt.

[0046] Step 5: View the travel route plan. After the travel route plan is generated, it is displayed visually on the system application page.

[0047] The tourism route planning system based on the AI large model, the specific process of the step 3 inputting the question or filling in the form comprises: Step 3.1: if the user is clear about the destination and the travel time, the "form mode" is selected, the destination and the travel time period are inputted, and multiple destinations are supported; and the keywords of the trip such as food and culture can be filled in. Otherwise, this step is omitted.

[0048] Step 3.2: if the user is not clear about the destination and the travel time, or the user wants to describe the tourism route planning requirement as a question in the form of text to inquire the answer, the "intelligent mode" is selected, and the text question is inputted. Otherwise, this step is omitted.

[0049] The tourism route planning system based on the AI large model, the specific process of the step 4 generating the tourism route planning comprises: Step 4.1: the destination, the travel days and the keywords of the trip are obtained according to the input of the user in the system page. If the user selects the "intelligent mode" in the step 3, the destination city, the travel time and the keywords of the trip in the question of the user are extracted based on the AI large model, and are processed in a format, and the question analysis process based on the AI large model is as shown in Figure 2 If the user selects the "form mode" in the step 3, the destination city, the travel days and the keywords of the trip are directly obtained through the form.

[0050] Step 4.2: the tourism route planning of a single city is generated according to the destination, the travel days and the keywords of the trip, and the tourism route planning process of the single city is as shown in Figure 3

[0051] Step 4.3: all the destination cities in the step 4.1 are traversed, and the step 4.2 is executed in a loop until the corresponding trip arrangement of all the destination cities is generated, and the complete trip is obtained by sorting according to the date.

[0052] The tourism route planning system based on the AI large model, the specific process of the step 4.1 based on the AI large model comprises: Step 4.1.1: whether the question inputted by the user belongs to the tourism route planning type is judged through the AI large model, if the question inputted by the user belongs to the tourism route planning type, the following step is continued to be executed; if the question inputted by the user does not belong to the tourism route planning type, the following step is terminated to be executed, and an error prompt is given.

[0053] Step 4.1.2: the destination city and the travel days are analyzed through the AI large model according to the question inputted by the user. If the destination city and the travel days cannot be successfully analyzed, the following step is terminated to be executed, and an error prompt is given.​

[0054] Step 4.1.3: According to the user input question, the AI large model determines the most suitable type of scenic spot for the user's trip.

[0055] Step 4.1.4: Format the parsed destination city, trip duration, and other information. The above-mentioned AI large model-based tourism route planning system, in step 4.2, if the user's trip duration in a certain destination city is n days, the system will recommend 3n scenic spots to form a complete tourism route planning, i.e. 3 scenic spots will be arranged in one day. The scenic spots in the tourism route planning are composed of a*like + b* famous + 1* unique, where a can be 0, and a maximum of (3n-1) / 2, at which point a=b.

[0056] The above-mentioned AI large model-based tourism route planning system, the specific process of single city tourism route planning in step 4.2 includes: Step 4.2.1: Query all scenic spots in the city in the scenic spot database, and sort the scenic spots by scenic spot ID in ascending order.

[0057] Step 4.2.2: Determine if the current scenic spot is liked by the user. If the current scenic spot has been marked as liked by the user, add this scenic spot to the list_like list (this list is a list of all liked scenic spots in this city by the user).

[0058] Step 4.2.2: Determine if the current scenic spot is not liked by the user. If the current scenic spot has been marked as not liked by the user, add this scenic spot to the list_notlike list (this list is a list of all disliked scenic spots in this city by the user).

[0059] Step 4.2.3: Determine if the current scenic spot is a famous scenic spot. If the current scenic spot's popularity is greater than 8 and the scenic spot type meets the user's target scenic spot type, add this scenic spot to the list_hot list (this list is a list of all famous scenic spots in this city by the user).

[0060] Step 4.2.4: Determine if the current scenic spot is a unique scenic spot. If the current scenic spot's popularity is less than 6 and the scenic spot type meets the user's target scenic spot type, add this scenic spot to the list_unique list (this list is a list of all unique scenic spots in this city by the user).

[0061] Step 4.2.5: Traverse all scenic spots in the city, repeat steps 4.2.2 to 4.2.4.

[0062] Step 4.2.6: Determine the length of the list_like list. If the length of the list_like list is 0, then a = 0; if the length of the list_like list is less than (3n-1) / 2, then a = length(list_like); otherwise, a = (3n-1) / 2. Randomly select a attractions from the list_like list and add them to the list_journey list of candidate attractions.

[0063] Step 4.2.7: Remove elements from the list_hot list that overlap with the list_notlike list. Let b = 3n-a-1. Randomly select b attractions from the list_hot list and add them to the list_journey list of candidate attractions.

[0064] Step 4.2.8: Remove elements from the list_unique list that overlap with the list_notlike list. Randomly select 1 attraction from the list_unique list and add it to the list_journey list of candidate attractions.

[0065] Step 4.2.9: Arrange the elements in the list_journey list of candidate attractions in a clockwise order based on their latitude and longitude information, and format them. This completes the planning of the n-day travel itinerary for the destination city.

[0066] The beneficial effects of the present embodiment are: Travel planning in travel apps is usually designed to guide users to make purchases, with a strong commercial purpose. The quality of search engine results is uneven, and users cannot avoid spending a lot of time filtering and selecting. The cost of obtaining objective and personalized travel route recommendation results is high. Based on the above two points, the present application proposes a method of quickly and accurately generating a travel route plan by constructing a point database and AI large model analysis, reducing user filtering time, improving search efficiency, and meeting the personalized needs of users.

[0067] The travel routes directly output by traditional generative AI large models have the problems of rigidity and serious irrationality. In actual tests, the reference degree is low. The illusion problem of AI large models also leads to incorrect attraction results and incorrect information, which may mislead users to some extent. Based on the above two points, the travel route generation algorithm proposed in the present application uses the pre-constructed point database as the baseline knowledge base and fully utilizes the capabilities of the AI large model. It fully considers user needs, personal preferences, attraction popularity, and other multi-dimensional information to quickly complete route recommendations, enhance the degree of personalization, and ensure information accuracy.

[0068] In specific embodiments: Establish a domestic scenic spot database, the database field is according to the scenic spot location information content format shown in Figure 1 , and the database is built as shown in Figure 5 .

[0069] Taking the requirement of "three-day tour route planning for a family of three in Dalian City" as an example, the user opens the system interface using the browser and completes the login. Select the intelligent mode, enter "three-day tour route planning for a family of three in Dalian City" in the question input box, and click Create Itinerary.

[0070] Through the question "three-day tour route planning for a family of three in Dalian City", the question is analyzed according to the AI large model question analysis process shown in Figure 3 , and the AI large model input question template of each link is designed: a. Ask the question template: "

Fill in the user's input question

[0071] b. Analyze the destination question template: "

Fill in the user's input question

[0072] c. Analyze the trip duration question template: "

Fill in the user's input question

[0073] d. Determine the suitable scenic spot type question template: "

Fill in the user's input question

[0074] Using the above designed question template, fill in the user's input question "three-day tour route planning for a family of three in Dalian City", and execute it in turn, as shown in Figure 5 .

[0075] From this, we can get the destination of the user's trip: Dalian City; trip duration: 3, i.e. n = 3; suitable scenic spot types: theme park, natural scenery. Format the result and continue to execute the following steps.

[0076] If the user has marked the favorite and the dislike scenic spots in advance, the favorite scenic spots in Dalian City are: Star Sea Square and Dalian Shell Museum, and the dislike scenic spots are: Dalian Old Wharf Scenic Area. According to the single-city tourism route planning process shown in the schematic diagram, the tourism route of Dalian City can be planned as follows: Figure 3 list_like = [Star Sea Square, Dalian Shell Museum]; list_notlike = [Dalian Old Wharf Scenic Area]; list_hot = [Tiger Beach Ocean Park, Dalian Forest Zoo, Dalian Haichang Discovery Kingdom Theme Park, …]; list_unique = [Dalian Luenen Ocean Hot Spring, Dalian Art Museum, …]; The length of the list_like list is 2, which is less than (3n-1) / 2=4, so a=2, b=3n-a-1=6, that is, all the scenic spots in the list_like list are added to the list_journey list, 6 scenic spots in the list_hot list are selected and added to the list_journey list, and 1 scenic spot in the list_unique list is selected and added to the list_journey list.

[0077] After the list_journey list is combined and arranged in a counterclockwise direction according to the latitude and longitude position information, the page rendering can be performed after the format processing.

[0078] The text list display and map visualization display are performed in the tourism route planning page. The user can view the planned travel route point by point through the “forward” and “backward” buttons, view the planned travel route day by day through the “switch day” operation, view the scenic spot information through the “text link” and “map point marker” operations, view the 3D remote sensing image animation of the point through the “3D” button, and view the automatically generated flying animation video of the travel through the “animation video” button.

[0079] ​The specific embodiments described above are merely illustrative of the present application and should not be considered limiting thereof. Those skilled in the art can make equivalent substitutions or modifications to the technical solutions of the present application to achieve the same technical effects, and these substitutions or modifications shall also be included in the protection scope of the present application. The technical solutions of the present application are further described in detail through several specific embodiments above, in order to highlight the advantages and benefits of the technical solutions provided by the present application. However, the above several specific embodiments are not used as a limitation of the present application, and any reasonable modifications and improvements, combinations and equivalent substitutions of the embodiments, etc. based on the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A tourism route planning system based on an AI large-scale model, characterized in that, include: The attraction database construction module is used to generate attraction information for various regions through a large AI model and establish an attraction database containing unique identifiers. The user input module is used to receive travel request information input by the user. The semantic parsing module is used to parse natural language input of travel needs and generate structured data; The route generation module is used to generate tourist routes from the attraction database based on the structured data; The visualization module is used to display the text and map information of the tourist route.

2. The travel route planning system based on an AI large-scale model according to claim 1, characterized in that, The scenic spot database construction module classifies domestic regions into three levels: province, city, and county, and assigns a unique ID to each level and each scenic spot.

3. The travel route planning system based on an AI large-scale model according to claim 1, characterized in that, The user input module includes a form mode and an intelligent mode. The form mode is used to input structured information, and the intelligent mode is used to input natural language questions.

4. The tourism route planning system based on an AI large-scale model according to claim 1, characterized in that, The semantic parsing module uses a preset prompt template to call the AI ​​big model to extract the destination, number of days of the trip, and preferred attraction types.

5. A tourism route planning system based on an AI large-scale model according to claim 1, characterized in that, The route generation module combines three attractions per day from users' favorite attractions, popular attractions, and lesser-known attractions to form an itinerary, and sorts the routes according to latitude and longitude.

6. The tourism route planning system based on an AI large model according to claim 1, characterized in that, The route generation module categorizes attractions, including at least those the user likes, those they dislike, popular attractions, and lesser-known attractions, and excludes attractions the user dislikes before generating the itinerary.

7. A tourism route planning method based on an AI large-scale model, characterized in that, The system implementation based on claim 1 includes: Step 1: Build a scenic spot database by generating scenic spot information for each region using a large AI model and storing it as a data structure with unique identifiers; Step 2: Receive travel request information input by the user, including destination, number of days of trip, and keywords; Step 3: Parse the user input. If it is natural language, call the AI ​​large model to extract structured data. Step 4: Select attractions from the attraction database based on structured data, and create a tour route based on the number of three attractions per day; Step 5: Sort the generated tourist routes by latitude and longitude to form a complete itinerary; Step 6: Display the itinerary in text and map format, providing 3D views and animation playback.

8. A computer storage medium for storing computer programs, characterized in that, When the computer program is read by the computer, the computer executes the method of claim 7.

9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 7.

10. A computer program product, as a computer program, is characterized by: When the computer program is executed, it implements the method of claim 7.

Images