A multi-modal travel knowledge retrieval enhancement method based on a large language model and an intelligent agent itinerary planning method

Abstract

This invention discloses a multimodal cultural tourism knowledge retrieval enhancement method and an intelligent agent itinerary planning method based on a large language model. The retrieval enhancement method includes step S201, constructing a multimodal cultural tourism knowledge base, which includes travel product data, scenic spot real-time information, scenic area activity information, traffic and weather dynamic data, and tourist evaluation feedback information; step S202, constructing a retrieval enhancement framework based on a RESTful API interface and the MCP protocol; and step S203, using the retrieval enhancement framework to retrieve real-time cultural tourism data from the multimodal cultural tourism knowledge base. This invention, through a multimodal knowledge base and a dual-protocol retrieval framework, achieves real-time integration and related retrieval of scattered data, solving the problem of fragmented cultural tourism information and ensuring information timeliness.

Description

Technical Field

[0001] This invention relates to information processing technology in the fields of artificial intelligence and cultural tourism, and particularly to a multimodal cultural tourism knowledge retrieval enhancement method and an intelligent agent itinerary planning method based on a large language model. Background Technology

[0002] With the accelerated digital transformation of the cultural tourism industry, tourists' demands for tourism information services have upgraded to real-time, personalized, diversified, and high-quality content display, making traditional service models increasingly inadequate. Currently, cultural tourism information services face multiple core pain points: On the one hand, data sources are highly fragmented and heterogeneous, encompassing travel agency ERP product information, scenic spot announcements, weather and traffic data, social media content, tourist reviews, and multimodal audio and video, etc. Data updates frequently and are in complex formats, making it difficult for traditional static knowledge bases to achieve efficient integration and dynamic management. On the other hand, general-purpose large language models lack professional knowledge of the cultural tourism industry and do not possess real-time update capabilities, easily resulting in lag when faced with dynamic information such as ticket price adjustments and event changes, and the reliability of output results cannot meet the needs of real business scenarios.

[0003] Meanwhile, existing cultural tourism recommendation systems mostly output results in simple text form, lacking visualization, multimedia display and multi-round dialogue optimization capabilities, and have insufficient explainability and interactivity, making it difficult to match users' in-depth decision-making needs; while human customer service is limited by human resources, making it difficult to handle massive inquiries during peak periods, and the service quality is easily affected by individual differences, making it difficult to guarantee consistency and comprehensive knowledge. Summary of the Invention

[0004] This invention aims to address the problems of outdated knowledge, limited scalability, lack of multimodal capabilities, and weak planning capabilities in existing cultural and tourism information services. It aims to achieve real-time retrieval and fusion of multimodal knowledge and automation of intelligent agent tasks for itinerary planning and product recommendation. This invention provides a multimodal cultural and tourism knowledge retrieval enhancement method and an intelligent agent itinerary planning method based on a large language model.

[0005] The technical solution of the present invention is as follows: A multimodal cultural and tourism knowledge retrieval enhancement method based on a large language model includes the following steps: Step S201: Construct a multimodal cultural tourism knowledge base, which includes at least travel product data, scenic spot real-world information, scenic area activity information, traffic and weather dynamic data, and tourist evaluation feedback information. Step S202: Build a retrieval enhancement framework based on the RESTful API interface and the MCP protocol; Step S203: Use the retrieval enhancement framework to retrieve real-time cultural and tourism data from the multimodal cultural and tourism knowledge base.

[0006] Further, in step S201, the travel product data is imported through the travel agency platform's ERP business system. The travel product data includes at least itinerary routes, attraction tickets, hotel accommodations, transportation arrangements, and tour packages; and / or The real-time information about the scenic spots is obtained through social media platforms, travel information websites, and the scenic spots' own platforms. This real-time information includes at least real-time photos and videos of the scenic spots; and / or The scenic area activity information is obtained through the scenic area's website or news platform, and includes at least announcements of cultural activities, festival activities, and tourism promotions; and / or The traffic and weather dynamic data is obtained through a map-based meteorological service platform, and includes at least distance, travel time, road condition forecast, temperature, precipitation probability, and air quality; and / or The tourist evaluation feedback information is obtained through travel agency platforms and social media platforms, and includes at least the travel experience, travel suggestions, and product feedback information.

[0007] Furthermore, the map-based meteorological service platform includes a map service platform and a meteorological service platform; The distance, travel time, and road condition predictions in the traffic and weather dynamic data are obtained in real time through the WebService API of the map service platform and used for one or more route planning methods, including driving, public transportation, and walking. The temperature, precipitation probability, and air quality in the traffic and weather dynamic data are obtained through the meteorological service platform to provide multi-day weather forecasts for the destination.

[0008] Further, in step S202, the RESTful API interface is a standardized interface used to connect to the ERP business system of the travel agency platform. The RESTful API interface includes at least a full-text search interface, a product details query interface, and a route search interface for querying cultural and tourism service targets; wherein, The full-text search interface is used to perform full-text search based on the key content of the user's initial needs, and to filter out alternative products with high matching degree from travel product data; The product details query interface is used to return complete information about the user's selected alternative products; The route retrieval interface is used to return a list of itineraries based on the user's travel itinerary list and / or daily itinerary plan.

[0009] Furthermore, in step S202, the MCP protocol is used to call data information from external platforms, which include at least social media platforms and map and weather service platforms.

[0010] Furthermore, the method includes: After obtaining the user's query command, the large language model automatically parses the query command based on the MCP protocol and transforms it into a structured result or natural language expression that the user can understand; whereby... When a user requests a text command to search for travel guides for a destination, the system calls data from social media platforms via the MCP protocol and returns a results list. The results list includes at least the post ID, author, text summary, multimedia links, and interaction data from the social media platforms. The results list is sorted by popularity and relevance, and the Top-K results are selected as supplementary knowledge.

[0011] This invention also provides an intelligent agent task planning method, which employs the above-mentioned multimodal cultural tourism knowledge retrieval enhancement method and includes the following steps: Step S1: Construct a cultural tourism intelligent agent based on the ReAct mode. The intelligent agent extracts user departure point, destination, travel time, number of days, budget, number of companions, preferences and special needs obtained from the user terminal through intent recognition and slot filling. For ambiguous needs, guide confirmation through questioning to form a structured user profile. Step S2: Based on the user profile, the intelligent agent connects to a multimodal cultural tourism knowledge base, including at least the ERP business system of a travel agency platform, social media platform, and map and weather service platform, through a retrieval enhancement framework, to achieve cross-modal information query and retrieval. Step S3 involves sorting and filtering the retrieved cultural and tourism products using a multi-objective optimization algorithm, and then integrating them to generate a cultural and tourism product recommendation scheme or itinerary planning scheme that includes recommendation reasons, product details, and itinerary arrangements, supporting multiple rounds of interactive optimization.

[0012] Furthermore, in step S2, the information query function of the travel agency platform's ERP business system in the intelligent agent includes: Product index library construction: Extract key fields including at least product name, route number, price, departure date, availability information, hotel class, and attractions to form a searchable index library; Semantic retrieval interface: Transforms user intent into product index query statements through natural language query mapping, realizing the conversion from fuzzy question answering to exact matching; Results parsing and formatting: The returned alternative product data is parsed in a structured manner to generate table or card-style results with accompanying jump links.

[0013] Furthermore, in step S2, the information retrieval function of the social media platform in the intelligent agent includes: Content recall and popularity analysis function: Retrieve text, images and short video content through keywords and geographic tags obtained from the user terminal, calculate interaction popularity and sentiment, and identify current popular attractions and activities; Multimodal knowledge extraction function: Using text summarization technology, extract scenic spot features and tourist evaluations to construct a "scenic spot impression vector"; Assisted recommendation generation function: Associates the popularity of social media content with the travel nodes of alternative products to optimize the sorting of alternative products or generate dynamic recommendation descriptions.

[0014] Furthermore, in step S2, the information retrieval function of the map weather service platform in the intelligent agent includes: Route optimization and spatiotemporal calculation functions: Call the map service platform to calculate the optimal route to multiple points, taking into account transportation modes and real-time traffic conditions, to achieve the optimization of minimizing travel time or maximizing user experience; Attraction spatial matching function: Automatically matches spatial coordinates based on attractions included in the alternative products or popular landmarks recommended by social media platforms to form a geographic visualization itinerary map; Intelligent weather and season linkage function: Invoke the meteorological service platform to query the future weather of the destination and generate itinerary tips adapted to the weather.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention achieves real-time integration and related retrieval of scattered data through a multimodal cultural tourism knowledge base and a dual-protocol retrieval framework, solving the problem of fragmented cultural tourism information and ensuring information timeliness; the intelligent agent based on the ReAct mode can accurately identify user needs and select the optimal solution through a multi-objective optimization algorithm.

[0016] This invention addresses user needs and product recommendation solutions by presenting users with rich cultural and tourism information, including text, voice, charts, and images, based on multimodal information integration and large language model content generation technology. Furthermore, it combines multimodal information into user-friendly interactive Q&A, achieving an intuitive, convenient, and enjoyable user experience, thus improving the quality of cultural and tourism products and services. In addition, it achieves plug-and-play functionality through APIs and MCP protocols, adapting to the application needs of small and medium-sized enterprises in the cultural and tourism sector. Attached Figure Description

[0017] Figure 1 This is a flowchart of a multimodal cultural tourism knowledge retrieval enhancement method according to the present invention; Figure 2 This is a flowchart of an intelligent agent task planning method according to the present invention. Detailed Implementation

[0018] The following is in conjunction with the appendix Figures 1-2 The technical solution of the present invention will be further described in detail below with reference to the embodiments.

[0019] Combination Figure 1A multimodal cultural tourism knowledge retrieval enhancement method based on a large language model, the specific steps of which include: Step S201: Construct a multimodal cultural tourism knowledge base, which includes at least travel product data, scenic spot real-world information, scenic area activity information, traffic and weather dynamic data, and tourist evaluation feedback information. The travel product data is imported from the ERP business system of travel agency platforms, and at least covers core product information such as itineraries, attraction tickets, hotel accommodations, transportation arrangements, and tour packages. This data primarily serves as an index for travel products, supporting core functions such as itinerary planning, price calculation, and order generation. The upper layer exposes service capabilities through a RESTful API interface, supporting queries based on keywords, destination, and number of trip days.

[0020] The real-world information about tourist attractions comes from social media platforms, travel information websites, and the attractions' own platforms, including at least real-world photos and videos. Taking Xiaohongshu (Little Red Book) as an example, to compensate for the abstract nature of text descriptions, the platform integrates publicly available content, acquiring a large amount of real-world photos, short videos, travel guides, and hand-drawn maps. These self-media resources are uploaded by real tourists, reflecting the current true appearance of the attractions and popular photo spots. Xiaohongshu content is highly timely—for example, spring flower fields, winter snowscapes, and festival activities are presented immediately. Furthermore, the self-media commentary videos provide an immersive experience, helping users intuitively determine whether a visit is worthwhile.

[0021] When accessing information from external platforms, due to copyright and data compliance issues, the relevant data is not stored locally. Instead, resource links, such as image URLs and video IDs, are dynamically obtained through online search interfaces. The data is returned in semi-structured JSON format, containing metadata such as title, body text, tags, likes, and favorites. Although the total amount of data is theoretically unlimited, high-value content can be extracted through keyword filtering and sorting mechanisms.

[0022] Information about activities at the scenic area comes from the scenic area's official website and news platforms, and includes at least announcements of cultural activities, festivals, and tourism promotions.

[0023] Traffic and weather dynamic data are sourced from a map-based weather service platform and include at least distance, travel time, road condition forecast, temperature, precipitation probability, and air quality. The map-based weather service platform comprises a map service platform and a weather service platform. Distance, travel time, and road condition forecasts in the traffic and weather dynamic data are obtained in real-time through the map service platform's Web Service API and are used for one or more route planning options, including driving, public transportation, and walking. Temperature, precipitation probability, and air quality in the traffic and weather dynamic data are obtained through the weather service platform's interface, providing at least a 7-day weather forecast for the destination. Traffic and weather dynamic data are used to determine travel feasibility. For example, if a user plans to visit two attractions 80 kilometers apart within a day, the system can determine feasibility based on real-time road conditions; if heavy rain is forecast, the system will proactively remind the user to adjust their itinerary or prepare rain gear.

[0024] Tourist feedback information comes from travel agency platforms and social media platforms, and includes at least information on travel experiences, travel advice, and product feedback. Taking the social media platform Xiaohongshu as an example, UGC (User Generated Content) comments and notes containing travel experiences, travel advice, tips to avoid pitfalls, and product comparisons are scraped from Xiaohongshu. These unstructured texts contain rich subjective evaluation information, such as "child-friendly," "good for taking photos," and "long queues." Tourist feedback information serves two purposes: first, it can assist users in decision-making by providing authentic feedback from a third-party perspective; second, it can serve as training signals for intelligent agent recommendation systems to optimize recommendation strategies. For example, if a scenic spot is mentioned as having "excellent family-friendly facilities" in multiple highly-rated notes, it can be prioritized for recommendation in family travel scenarios. Tourist feedback information is also remotely accessed via RESTful API to avoid the legal risks associated with local storage. At the same time, with the help of sentiment analysis and keyword extraction technology, high-frequency evaluation dimensions can be automatically extracted to form structured rating indicators.

[0025] Step S202: Build a retrieval enhancement framework based on the RESTful API interface and the MCP protocol; Step S203: Use the retrieval enhancement framework to retrieve real-time cultural and tourism data from the multimodal cultural and tourism knowledge base.

[0026] The RESTful API interface serves as a standardized interface for connecting to the ERP business system of travel agency platforms, supporting queries for cultural and tourism services. The RESTful API interface must include at least the following: The full-text search interface is used to perform full-text searches and filter out highly relevant alternative products from travel product data; The product details query interface is used to return complete information about the user's selected alternative products; The route search interface is used to return a list of itineraries based on the user's travel itinerary list and / or daily itinerary plan.

[0027] All of the above interfaces return data in JSON format and follow a unified data model specification.

[0028] External platforms include at least social media platforms, map service platforms, and weather service platforms. For these platforms, the API capabilities are encapsulated using the MCP protocol. MCP (Model Context Protocol) is a communication protocol that allows large language models to securely call external tools. When the large language model receives a user's query command, it automatically parses and transforms the query command into a structured result or natural language expression that the user can understand, based on the MCP protocol, for use in cultural and tourism knowledge retrieval and interactive responses. Its basic principle is to enable the model to understand and request external services (such as APIs) through standardized interface descriptions, obtaining real-time data or performing operations. The system, according to predefined tool specifications, transforms the model-generated call requests into actual API calls and returns the structured results to the model, thereby expanding its capabilities, enabling dynamic interaction with external systems, and improving the accuracy and usability of responses.

[0029] When retrieving a user's text command to query destination travel guides, the system calls data from social media platforms via the MCP protocol and returns a results list. This results list includes at least the post ID, author, text summary, multimedia links, and interaction data from the social media platforms. The results list is sorted by popularity and relevance, and the Top-K results are selected as supplementary knowledge. When retrieving a user's command to query destination traffic and weather information, the large language model, based on the user's natural language needs, allows the model to automatically construct request parameters and parse the response according to semantics.

[0030] Compared to traditional plain text vector knowledge base methods, the retrieval enhancement scheme proposed in this invention has significant advantages. Firstly, it ensures high data timeliness. This invention does not perform local data synchronization; all information is obtained in real-time from the source, avoiding information lag caused by caching delays. For example, a temporary park closure notice can be reflected in the query results immediately after its release. Secondly, it provides clear semantics and explicit data relationships. Whether it's the field-based data returned by the structured RESTful API or the JSON object encapsulated by MCP, its semantics are clearly defined by the interface documentation, coupled with annotations such as the OpenAPI Schema, ensuring unambiguous data meaning. In contrast, traditional document slicing methods often lead to missing or misunderstood information due to contextual breaks. The multimodal cultural tourism knowledge retrieval enhancement method proposed in this chapter fully utilizes the existing data ecosystem to construct a low-cost, highly timely, and easily maintainable intelligent service architecture.

[0031] This invention aims to flexibly respond to users' consultation needs in different scenarios, deeply analyze the typical characteristics of user consultations in the cultural and tourism industry, and introduce intelligent agent technology to encapsulate capabilities such as web search, weather query, public opinion analysis, and navigation positioning into a standardized toolset. Based on a large language model, it performs intent recognition and task planning for user consultation questions, and realizes autonomous tool invocation based on the MCP protocol. It collaborates with users to complete tasks such as information integration, product selection, and itinerary planning, and finally generates a cultural and tourism product recommendation scheme with reasonable and effective reasons and strong interpretability throughout the process. It also supports multi-round interactive optimization, improves the accuracy of intelligent recommendation of cultural and tourism products and user experience, and provides an intelligent agent task planning method.

[0032] Combination Figure 2 The intelligent agent task planning method of the present invention, employing the above-mentioned multimodal cultural tourism knowledge retrieval enhancement method, includes: Step S1: Construct a cultural tourism intelligent agent based on the ReAct model. The intelligent agent first extracts the user's departure point, destination, travel time, number of days, budget, number of companions, preferences and special needs through intent recognition and slot filling. For ambiguous needs, it guides confirmation through questioning to form a structured user profile.

[0033] In step S2, the intelligent agent, based on the user profile, connects to a multimodal cultural tourism knowledge base, including at least the ERP business system of travel agency platforms, social media platforms, and map and weather service platforms, through a retrieval enhancement framework to achieve cross-modal information query and retrieval. The information query data includes relevant destination information, transportation methods, hotel resources, scenic spot tickets, featured restaurants and activities, etc.

[0034] The information query function of the travel agency platform's ERP business system includes: Product index library construction: Extract key fields including at least product name, route number, price, departure date, availability information, hotel class, and attractions to form a searchable index library; Semantic retrieval interface: Transforms user intent into product index query statements through natural language query mapping, realizing the conversion from fuzzy question answering to exact matching; Results parsing and formatting: The returned data of alternative products is parsed in a structured manner to generate table or card-style results with accompanying jump links.

[0035] The information retrieval functions of social media platforms include: Content recall and popularity analysis function: Retrieve text, images and short video content through keywords and geographic tags, calculate interaction popularity and sentiment, and identify current popular attractions and activities; Multimodal knowledge extraction function: Using text summarization technology, extract scenic spot features and tourist evaluations to construct a "scenic spot impression vector"; Assisted recommendation generation function: Associates the popularity of social media content with the travel nodes of alternative products to optimize the sorting of alternative products or generate dynamic recommendation descriptions.

[0036] The information retrieval functions of the map-based weather service platform include: Route optimization and spatiotemporal calculation functions: Call the map service platform to calculate the optimal route to multiple points, taking into account transportation modes and real-time traffic conditions, to achieve the optimization of minimizing travel time or maximizing user experience; Attraction spatial matching function: Automatically matches spatial coordinates based on attractions included in the alternative products or popular landmarks recommended by social media platforms to form a geographic visualization itinerary map; Intelligent weather and season linkage function: Call the meteorological service platform to check the future weather of the destination and generate travel tips that are adapted to the weather.

[0037] Step S3 involves sorting and filtering the retrieved cultural and tourism products using a multi-objective optimization algorithm, and integrating them to generate a recommended cultural and tourism product proposal or itinerary planning scheme that includes recommendation reasons, product details, and itinerary arrangements, supporting multi-round interactive optimization. In this step, the intelligent agent selects the most suitable combination from multiple product schemes based on user preferences and constraints, such as choosing a suitable route package, hotel package, or self-driving route.

[0038] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multimodal cultural tourism knowledge retrieval enhancement method based on a large language model, characterized in that, Includes the following steps: Step S201: Construct a multimodal cultural tourism knowledge base, which includes at least travel product data, scenic spot real-world information, scenic area activity information, traffic and weather dynamic data, and tourist evaluation feedback information. Step S202: Build a retrieval enhancement framework based on the RESTful API interface and the MCP protocol; Step S203: Use the retrieval enhancement framework to retrieve real-time cultural and tourism data from the multimodal cultural and tourism knowledge base.

2. The method according to claim 1, characterized in that, In step S201, the travel product data is imported through the ERP business system of the travel agency platform. The travel product data includes at least itinerary routes, attraction tickets, hotel accommodations, transportation arrangements, and tour packages; and / or The real-time information about the scenic spots is obtained through social media platforms, travel information websites, and the scenic spots' own platforms. This real-time information includes at least real-time photos and videos of the scenic spots; and / or The scenic area activity information is obtained through the scenic area's website or news platform, and includes at least announcements of cultural activities, festival activities, and tourism promotions; and / or The traffic and weather dynamic data is obtained through a map-based meteorological service platform, and includes at least distance, travel time, road condition forecast, temperature, precipitation probability, and air quality; and / or The tourist evaluation feedback information is obtained through travel agency platforms and social media platforms, and includes at least the travel experience, travel suggestions, and product feedback information.

3. The method according to claim 2, characterized in that, The map-based meteorological service platform includes a map service platform and a meteorological service platform. The distance, travel time, and road condition predictions in the traffic and weather dynamic data are obtained in real time through the WebService API of the map service platform and used for one or more route planning methods, including driving, public transportation, and walking. The temperature, precipitation probability, and air quality in the traffic and weather dynamic data are obtained through the meteorological service platform to provide multi-day weather forecasts for the destination.

4. The method according to claim 1, characterized in that, In step S202, the RESTful API interface is a standardized interface used to connect to the ERP business system of the travel agency platform. The RESTful API interface includes at least a full-text search interface, a product details query interface, and a route search interface for querying cultural and tourism service targets; wherein, The full-text search interface is used to perform full-text search based on the key content of the user's initial needs, and to filter out alternative products with high matching degree from travel product data; The product details query interface is used to return complete information about the user's selected alternative products; The route retrieval interface is used to return a list of itineraries based on the user's travel itinerary list and / or daily itinerary plan.

5. The method according to claim 2, characterized in that, In step S202, the MCP protocol is used to call data information from external platforms, which include at least social media platforms and map and weather service platforms.

6. The method according to claim 5, characterized in that, The method includes: After obtaining the user's query command, the large language model automatically parses the query command based on the MCP protocol and transforms it into a structured result or natural language expression that the user can understand; whereby... When a user requests a text command to search for travel guides for a destination, the system calls data from social media platforms via the MCP protocol and returns a results list. The results list includes at least the post ID, author, text summary, multimedia links, and interaction data from the social media platforms. The results list is sorted by popularity and relevance, and the Top-K results are selected as supplementary knowledge.

7. A task planning method for an intelligent agent, characterized in that, The multimodal cultural and tourism knowledge retrieval enhancement method according to any one of claims 1-6 includes the following steps: Step S1: Construct a cultural tourism intelligent agent based on the ReAct mode. The intelligent agent extracts user departure point, destination, travel time, number of days, budget, number of companions, preferences and special needs obtained from the user terminal through intent recognition and slot filling. For ambiguous needs, guide confirmation through questioning to form a structured user profile. Step S2: Based on the user profile, the intelligent agent connects to a multimodal cultural tourism knowledge base, including at least the ERP business system of a travel agency platform, social media platform, and map and weather service platform, through a retrieval enhancement framework, to achieve cross-modal information query and retrieval. Step S3 involves sorting and filtering the retrieved cultural and tourism products using a multi-objective optimization algorithm, and then integrating them to generate a cultural and tourism product recommendation scheme or itinerary planning scheme that includes recommendation reasons, product details, and itinerary arrangements, supporting multiple rounds of interactive optimization.

8. The method according to claim 7, characterized in that, In step S2, the information query function of the travel agency platform's ERP business system in the intelligent agent includes: Product index library construction: Extract key fields including at least product name, route number, price, departure date, availability information, hotel class, and attractions to form a searchable index library; Semantic retrieval interface: Transforms user intent into product index query statements through natural language query mapping, realizing the conversion from fuzzy question answering to exact matching; Results parsing and formatting: The returned alternative product data is parsed in a structured manner to generate table or card-style results with accompanying jump links.

9. The method according to claim 7, characterized in that, In step S2, the information retrieval function of the social media platform in the intelligent agent includes: Content recall and popularity analysis function: Retrieve text, images and short video content through keywords and geographic tags obtained from the user terminal, calculate interaction popularity and sentiment, and identify current popular attractions and activities; Multimodal knowledge extraction function: Utilize text summarization technology to extract scenic spot features and tourist evaluations to construct a "scenic spot impression vector"; Assisted recommendation generation function: Associates the popularity of social media content with the travel nodes of alternative products to optimize the sorting of alternative products or generate dynamic recommendation descriptions.

10. The method according to claim 7, characterized in that, In step S2, the information retrieval function of the map weather service platform in the intelligent agent includes: Route optimization and spatiotemporal calculation functions: Call the map service platform to calculate the optimal route to multiple points, taking into account transportation modes and real-time traffic conditions, to achieve the optimization of minimizing travel time or maximizing user experience; Attraction spatial matching function: Automatically matches spatial coordinates based on attractions included in the alternative products or popular landmarks recommended by social media platforms to form a geographic visualization itinerary map; Intelligent weather and season linkage function: Invoke the meteorological service platform to query the future weather of the destination and generate itinerary tips adapted to the weather.

Images