Scenic area intelligent space-time navigation method based on digital twinning

Abstract

The invention relates to the crossing field of application of a natural language model to tour guide service of the tourist industry, and provides relatively fixed tour guide service for specific exhibits and scenic spots depending on simple audio tour guide and radio frequency equipment when a traditional tourist attraction tour guide technology is applied. Personalized and accurate answers cannot be provided in real time in combination with tourist demands and space-time positions; the invention provides a tourist attraction intelligent space-time guide method based on digital twinning, which comprises the following steps of: constructing a virtual three-dimensional space model of a tourist attraction by using a digital twinning technology, and constructing a space-time intelligent question and answer data set containing space-time position information, user questions and answers through a retrieval enhancement generation technology; the pre-trained natural language model is finely adjusted by using the space-time intelligent question and answer data set, it is ensured that the natural language model accurately responds to specific needs of the scenic spot, the method is applied to a scenic spot server, personalized services are achieved on a browser terminal according to the position and viewing time of a tourist in the scenic spot, and the guide immersion is improved.

Description

Technical Field

[0001] The present invention relates to the cross - field of applying natural language models to cultural and tourism industry guiding services. More specifically, it relates to an intelligent spatio - temporal guiding method for tourist attractions based on digital twin. Background Art

[0002] With the continuous development of information technology, the tourism industry has gradually entered the digital and intelligent era. The most common scenic area guiding methods are map guiding and manual explanation. Although they can provide basic guiding services for tourists, tourists are prone to getting lost or missing important scenic spots. Manual explanation has high requirements for the personal ability of tour guides, and the training results of tour guides vary. There may be problems such as omissions, patternized explanations, etc. Some tour guides tend to explain unofficial history to attract attention and convey wrong historical knowledge. Moreover, during the peak tourist season, there is often a shortage of guiding services. For wireless interpreters and QR code scanning, this method usually relies on external devices. The devices are bulky to wear, have insufficient sound quality, consume a large amount of power, and the explanation content is fixed and unidirectionally transmitted to tourists, with poor interactivity. The explanation process is inevitably rigid and inhumane, and it is difficult to provide personalized services according to the real - time needs of tourists, making it difficult to stimulate the interest and sense of participation of tourists. The development of intelligent guiding, virtual reality (VR), and augmented reality (AR) technologies has brought a brand - new experience to tourists. By using AR / VR technologies to provide virtual tour guide services, the immersive experience makes tourists feel on - the - spot. However, long - term use will cause discomfort reactions such as motion sickness in some users, and the cost of VR / AR devices is high, resulting in a limited number of users willing to purchase the devices. In recent years, a large number of AI tour guide services have been constructed based on natural language models. Most of these AI tour guides are directly fine - tuned based on corpora, lacking an understanding of the spatial structure of scenic areas, and having certain limitations: the scene relevance is not strong, and they fail to provide accurate question - answering functions by combining the location of tourists and the time of questions, and cannot achieve dynamic path planning and real - time guiding recommendations based on spatial coordinates. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide an intelligent spatio - temporal guiding method for tourist attractions based on digital twin. By using digital twin technology to reconstruct the tourist attraction in the digital space, within the digital twin model of the scenic area, the intelligent agent links the natural language model. Based on the general Q&A pairs, personalized Q&A pairs, and text materials of the scenic area, through retrieval - enhanced generation technology, a triple Q&A pair containing spatio - temporal location information, user questions, and answers is constructed, and the natural language model is fine - tuned with the triple Q&A pair as the basic corpus. Finally, a natural language model with spatial intelligence is constructed, which can provide personalized answers and guiding recommendations to the questions according to the location and time of tourists.

[0004] To achieve the above - mentioned purpose, the present invention provides the following technical solutions:

[0005] A scenic area intelligent spatio-temporal navigation method based on digital twin constructs a virtual three-dimensional space model of the scenic area using digital twin technology, constructs a spatio-temporal intelligent Q&A dataset containing spatio-temporal position information, user questions and answers through retrieval-augmented generation technology, fine-tunes a pre-trained natural language model using the spatio-temporal intelligent Q&A dataset to ensure that the natural language model responds accurately to the specific needs of the scenic area, and the trained natural language model is applied to the general natural language model of the cultural and tourism industry and applied to the scenic area server, and realizes voice navigation and text Q&A through the browser terminal user interface, specifically including the following steps:

[0006] Step 1. Use digital twin technology to construct a virtual three-dimensional space model of the scenic area, and obtain the geographical information, infrastructure locations and spatial coordinate data of each scenic spot in the scenic area;

[0007] Step 2. Based on the general Q&A pairs of the scenic area, personalized Q&A pairs and text content about the scenic area crawled from the Internet, an intelligent agent that can move within the digital twin model of the scenic area is used to link the natural language model to generate Q&A pairs containing different spatio-temporal information, and finally integrated into a spatio-temporal intelligent Q&A dataset;

[0008] Step 3. Use the spatio-temporal intelligent Q&A dataset to fine-tune the pre-trained natural language model to obtain a general natural language model applied to the cultural and tourism industry, apply the general natural language model applied to the cultural and tourism industry to the spatio-temporal intelligent navigation module of the scenic area, and realize the update through the update module of the spatio-temporal intelligent navigation of the scenic area.

[0009] Furthermore, in Step 1, the information of the scenic area and the spatial coordinate data of each scenic spot are obtained according to the following method:

[0010] Step 1.1 Use satellite remote sensing technology to collect images of the scenic area, extract the geographical information and features of the scenic area through image matching technology, generate a three-dimensional space model with surface geometry for the processed scenic area images through structured light method, dense reconstruction technology and RealityCapture software, and convert the features of the terrain, buildings and roads in the scenic area into polygon meshes;

[0011] Step 1.2 Use geographic information system technology to complete the spatial information annotation of the three-dimensional model obtained in Step 1.1 to ensure that tourists can find the corresponding location in the real world according to the virtual tour guide.

[0012] Furthermore, in Step 2, the spatio-temporal intelligent Q&A dataset is obtained as follows:

[0013] Step 2.1 Create a knowledge base of the scenic area as the basic data source, and the dynamic information in the basic data source is updated in a timely manner by the staff of the scenic area;

[0014] Step 2.2 Use natural language technology to construct a knowledge graph of all scenic spots in the scenic area. In the knowledge graph, the basic data sources of the scenic spots are associated through nodes, edges, and attributes;

[0015] Step 2.3 Convert all the information in the scenic area knowledge graph constructed in Step 2.2 into natural language corpus pairs to obtain personalized Q&A pairs for the scenic area. Then, combine the general Q&A pairs of the scenic area and the text content of the scenic area crawled from the Internet to construct a spatio-temporal intelligent Q&A database for the scenic area.

[0016] Furthermore, in Step 3, the general natural language model fine-tuning method applied to the cultural and tourism industry is as follows:

[0017] Step 3.1 Construct a general natural language model evaluation benchmark for the cultural and tourism industry as a reference standard for the fine-tuning results;

[0018] Step 3.2 Select a pre-trained natural language model for fine-tuning. Fine-tune the pre-trained natural language model according to the spatio-temporal intelligent Q&A database of the scenic area obtained in Step 2. When the fine-tuning result of the natural language model meets the general natural language model evaluation benchmark for the cultural and tourism industry, the fine-tuning of the general natural language model for the cultural and tourism industry is completed.

[0019] In summary, the present invention has the following beneficial effects:

[0020] In order to improve the ability of the scenic area AI guide to provide real-time immersive services, the present invention proposes a new Q&A pair construction mode that combines spatio-temporal coordinates and questions. Using the scenic area digital twin and retrieval enhancement technology, a Q&A pair construction method based on basic questions and characteristic questions is realized, and the automated construction of a massive Q&A database with spatio-temporal information is achieved. The pre-trained natural language model is fine-tuned using the database to obtain a general natural language model applied to the cultural and tourism industry.

[0021] The present invention significantly reduces the need for manual intervention by scenic area staff, enabling the scenic area to operate more efficiently. The general natural language model of the present invention applied to the cultural and tourism industry overcomes the limitations of existing natural language models in dealing with scenic area tasks of traditional culture. By introducing a Q&A database with spatio-temporal information specific to the cultural and tourism industry for fine-tuning, it can better understand and respond to queries related to scenic area culture, providing more appropriate services for tourists. The general natural language model of the present invention applied to the cultural and tourism industry can also be accurate to the spatio-temporal location and specific needs of tourists, providing accurate path recommendations and guiding services, and providing personalized services according to the location and viewing time of tourists in the scenic area, greatly improving the satisfaction and experience of tourists and enhancing the immersion of the guide. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic diagram of the method of the present invention;

[0023] Figure 2 It is the flow chart of the intelligent guided tour Q&A system. Specific implementation manners

[0024] The present invention will be further described in detail below with reference to the accompanying drawings.

[0025] As Figure 1 shown in the flow chart, the present invention proposes a scenic area intelligent spatio-temporal guided tour method based on digital twin, which uses digital twin technology and spatial intelligent interaction system to provide a more convenient, intelligent and personalized scenic area guided tour experience for tourists. First, use digital twin technology to construct a virtual three-dimensional space model of the scenic area, and obtain the geographical information of the scenic area, the location of infrastructure, and the spatial coordinate data of each scenic spot. Construct a scenic area spatio-temporal intelligent Q&A data set containing spatio-temporal position information, user questions and answers through retrieval-augmented generation technology, and use the spatio-temporal intelligent Q&A data set to fine-tune a pre-trained natural language model to obtain a general natural language model applied to the cultural and tourism industry, ensuring that the general natural language model applied to the cultural and tourism industry can accurately respond to the specific needs of the scenic area, especially when dealing with personalized questions with spatio-temporal information. Then, configure the natural language model, that is, the general natural language model applied to the cultural and tourism industry, as a service on the server, and provide a user interface through the H5 page of the browser end. The user page supports multi-modal interaction methods of voice and text input, and has functions of voice navigation and text Q&A, improving the guided tour experience of tourists.

[0026] The specific steps are as follows:

[0027] Step 1. Use digital twin technology to construct a virtual three-dimensional space model of the scenic area, and obtain the geographical information of the scenic area, the location of infrastructure, and the spatial coordinate data of each scenic spot.

[0028] Step 1.1 Use satellite remote sensing technology to collect images of the scenic area, which contain the terrain, buildings, locations, and road information of the scenic area. Through image matching technology, extract the terrain, building, location, and road features of the scenic area from these images. Then upload the scenic area images to Reality Capture software to process the scenic area images. First, perform preliminary processing to generate a sparse three-dimensional point cloud model, and these points represent the key structural information in the scenic area. Subsequently, the software will perform more refined processing to generate a dense three-dimensional point cloud model to capture more details. Based on the point cloud data, Reality Capture uses a triangular mesh reconstruction method to generate a virtual three-dimensional space model of the scenic area with surface geometry, transform the features of the terrain, buildings, and roads in the scenic area into polygon meshes, and texture the three-dimensional meshes according to the color information of the remote sensing image data of the scenic area to enhance the realism of the virtual three-dimensional space model of the scenic area.

[0029] Step 1.2 Use Geographic Information System (GIS) technology to label and integrate the spatial information of the scenic area for the 3D model obtained in Step 1.1, and label the buildings, scenic spots, and roads in the scenic area as separate entities in the virtual 3D space model of the scenic area. GIS technology can obtain elevation, longitude, latitude, and slope data of the earth's surface. Using these data and spatial registration technology, precise docking of the virtual space model of the scenic area with the buildings, scenic spots, and roads in the real scenic area is carried out to ensure that when tourists conduct virtual tours, they can find the corresponding locations in the real world. Tourists can search for the corresponding locations in the real world according to the directions of the virtual tour.

[0030] Draw the spatial location information of all the main roads, footpaths, and passageways in the scenic area in GIS, and label the attributes of each path, such as walking paths, wheelchair access, connection to main scenic spots, etc., and calculate the accessibility and passage difficulty of these paths.

[0031] The staff in the scenic area maintain and update the scenic area route information obtained by GIS according to the actual situation of the scenic area, and ensure the accuracy and reliability of the information by manually modifying the relevant data in GIS.

[0032] Step 2. Conduct spatio-temporal intelligent Q&A data integration. Based on the general Q&A pairs of the scenic area, personalized Q&A pairs, and relevant text materials of the scenic area crawled from the Internet, intelligent agents that can move within the digital twin model of the scenic area are used to link natural language models to generate Q&A pairs containing different spatio-temporal information, and finally integrate them into a spatio-temporal intelligent Q&A data set:

[0033] Step 2.1 First, it is necessary to integrate various types of information of the scenic area to create a detailed scenic area knowledge base, including name, type, spatial coordinates, historical and cultural background, opening hours, activity arrangements, and ticket information. Organize these information into a unified format, and finally create a complete and detailed scenic area knowledge base as the basic data source. The dynamic data is updated in a timely manner by the staff in the scenic area, and the dynamic data includes opening hours and ticket information. The format of the scenic area knowledge base is CSV, and each entry contains the following attributes:

[0034] Name: The name of the scenic spot, building, facility, performance, or road.

[0035] Type: Indicates whether it is a scenic spot, building, facility, or road.

[0036] Spatial coordinates: The corresponding geographical coordinates or spatial coordinates.

[0037] Background: The brief introduction or historical and cultural background of the scenic spot, building, or facility.

[0038] Opening hours: For example, the opening hours of scenic spots, roads, etc. are 8:00 - 20:30, and the performance time is 19:30 - 20:00.

[0039] Ticketing information: The ticket prices of the scenic area and ticketing-related information.

[0040] Connection relationship: The connection information with other entries, such as path relationship and time relationship.

[0041] Step 2.2 Use natural language technology to construct a knowledge graph of the scenic area, and construct a knowledge graph with all the scenic spot information and facility information in the scenic area. The knowledge graph mainly consists of three parts: nodes, edges, and attributes. The nodes in the knowledge graph include scenic spot nodes and facility nodes. Take historical and cultural background, opening hours, event arrangements, and ticketing information as the attributes of the scenic spot nodes. The edges represent the relationships between the nodes, such as location relationship, time relationship, and service relationship.

[0042] Step 2.3 Convert all the information in the scenic area knowledge graph constructed in Step 2.2 into natural language corpus pairs to obtain personalized Q&A pairs of the scenic area, and then combine the general Q&A pairs of the scenic area and the relevant text materials of the scenic area crawled from the Internet to construct a spatio-temporal intelligent Q&A database of the scenic area. On the one hand, the scenic area knowledge graph contains specific spatio-temporal information; on the other hand, use an agent to call the natural language model and the spatio-temporal intelligent Q&A database of this scenic area, and automatically generate new Q&A pairs through the retrieval enhancement mode. During the generation process, the pre-trained natural language model will combine the spatio-temporal location, needs of the tourists, and the real-time information of the scenic area to automatically generate Q&A pairs containing spatio-temporal information. The format of all the final Q&A pairs is as follows: longitude and latitude + time + question, longitude and latitude + time + answer. For example:

[0043] Question: "(36°N, 110°E (current location) + 19:00) I want to watch the fireworks show."

[0044] Answer: "(36°N, 110°E (current location) + 19:00) The opening time of the fireworks show is from 19:30 to 20:00. Starting from your current location, walk 1000 meters north along the main road to reach the fireworks show viewing area."

[0045] Step 3. Use the spatio-temporal intelligent Q&A dataset to fine-tune the pre-trained natural language model to obtain a general natural language model applied to the cultural and tourism industry. The fine-tuning method of the general natural language model is as follows:

[0046] Step 3.1 First, construct an evaluation benchmark for the general natural language model applied to the cultural and tourism industry, which serves as a reference standard for the fine-tuning results of the general natural language model applied to the cultural and tourism industry. The evaluation benchmark mainly consists of two parts: subjective scores and objective scores. Subjective scores include: 1. Manual evaluation: Arrange scenic area guides to score the results of each model; 2. Model mutual scoring: Use the results of Tongyi Qianwen model for scoring, including accuracy, integrity, relevance, clarity, and politeness. Objective scores are evaluated using general AI models: cosine similarity, METEOR, BLEU, and ROUGE. These metrics are commonly used in text generation tasks and can effectively evaluate the similarity, generation quality, and coverage between the text content and the reference answer.

[0047] Step 3.2 Select a pre-trained natural language model for fine-tuning. Taking Llama3-8b as an example, this large-scale natural language model performs well in various natural language processing tasks and can effectively process various text information. Combining with the spatio-temporal intelligent Q&A database of the scenic area constructed in Step 2, and guided by the evaluation benchmark of the scenic area, fine-tune the pre-trained natural language model. During fine-tuning, the pre-trained natural language model can learn how to generate answers in a specific time and space context and will be more sensitive to questions related to cultural and tourism and scenic areas. For example, the model will learn how to answer queries about scenic spots based on the real-time location of tourists and how to provide appropriate activity recommendations for tourists according to the current time. When the scenic area evaluation of the natural language model reaches the evaluation benchmark in Step 3.1, the fine-tuning of the pre-trained natural language model is completed, and a general natural language model applied to the cultural and tourism industry is obtained for the spatio-temporal intelligent guided tour module of the scenic area, and updates are realized through the update module of the spatio-temporal intelligent guided tour of the scenic area.

[0048] The specific operations of the spatio-temporal intelligent guided tour module of the tourist scenic area are as follows:

[0049] Step A1. Configure the general natural language model applied to the cultural and tourism industry as a service on the server, and provide a simple and intuitive user interface through the H5 page of the browser side. Tourists can enter the guided tour system and quickly obtain a personalized guided tour experience by simply scanning the QR code or accessing the web page through other means.

[0050] Step A2. After the user logs in, interact through two methods: voice input or text input. The spatio-temporal intelligent guided tour module of the tourist scenic area uses speech recognition technology to convert speech into text to achieve voice input, which is realized with the help of existing speech recognition platforms (such as Google Speech-to-Text, iFlytek Hearing, etc.).

[0051] Step A3. The acquisition of time and space is as follows:

[0052] Step A3.1 Real-time GPS positioning: Modern devices usually have high-precision GPS functions, which can achieve positioning accuracy within 3-5 meters in outdoor environments. Therefore, using GPS technology, the latitude and longitude of tourists are obtained in real time through mobile phones or other devices, and the location data of tourists will be dynamically updated at each interaction;

[0053] Step A3.2 Time synchronization: In addition to location data, the current timestamp is also obtained each time the visitor interacts, ensuring that the time information in the questions and answers matches the visitor's real-time activities. When a visitor enters a scenic spot, the system adds the location and timestamp of each visitor's interaction to the visitor's question-and-answer interaction. The intelligent question-and-answer model can then identify the visitor's current location and time, and perform accurate scene recognition, tour suggestions, and route recommendations.

[0054] Step A4. After the tourists interact, the answer given by the general natural language model applied to the cultural tourism industry can not only be displayed on the screen in text, but also read out through speech synthesis technology, which is the same technology as step A2.

[0055] Step A5. The user interaction module not only supports instant question-and-answer services, but also has learning and optimization capabilities. It can continuously adjust the content of the guided tour service based on tourists' feedback to enhance tourists' personalized experience.

[0056] Step A6. In order to improve the system response speed and stability, the backend is based on the RESTful API architecture and adds a cache mechanism to ensure efficient interaction between the front-end H5 page and the back-end service. At the same time, the system adopts a multi-layer data processing architecture to ensure smooth operation under high concurrency.

[0057] The specific operations and applications of the operation and maintenance update module are as follows:

[0058] Step B1. As the scenic area changes (such as adding new attractions, updating facilities or adjusting routes, etc.), the digital twin needs to be updated synchronously. The updated information is input into the spatial information of the scenic area through the API interface to ensure that the virtual model is synchronized with the real world and update the digital twin model of the scenic area.

[0059] Step B2. On the one hand, as the digital twin model of the scenic spot is updated, the spatiotemporal intelligent question-answering database should also be updated synchronously. Using natural language processing technology, repeat steps 2.1 and 2.2. to generate new question-answer pairs. On the other hand, based on the real information feedback from tourists after the tour, feedback is given to the tour guide and other scenic spot staff through the external interactive interface. After the staff sorts out the answers, they manually update the spatiotemporal intelligent question-answering database;

[0060] Step B3. According to the internal adjustments of the scenic area (such as ticket changes, performance arrangements, maintenance and repair, etc.), the scenic area staff updates the scenic area knowledge base. After the update, Step 2.3 is used to update the spatio-temporal intelligent Q&A database;

[0061] Step B4. Through continuous updates in Steps 5.2 and 5.3, the new data will ultimately be used to retrain the scenic area private model, ensuring that the model can respond to the dynamic changes of the scenic area in real time and provide personalized services.

[0062] The present invention proposes a spatial intelligent guiding technology for tourist scenic areas based on digital twins, aiming to provide tourists with a more convenient, intelligent, and personalized scenic area guiding experience by using digital twin technology and a spatial intelligent interaction system. First, the digital twin technology is used to construct a virtual space model of the scenic area, and data such as the geographical information of the scenic area, the locations of infrastructure, and the spatial coordinates of each scenic spot are obtained. Then, the spatio-temporal intelligent Q&A data of the scenic area containing spatio-temporal information is integrated, and a new Q&A pair dataset is generated through an intelligent agent and a natural language model. Based on the spatio-temporal intelligent Q&A dataset, the training of a private model for tourist scenic areas with spatio-temporal intelligence is carried out, and the pre-trained language model is fine-tuned to ensure that the natural language model can accurately respond to the specific needs of the scenic area, especially when dealing with personalized questions with spatio-temporal information. After the training of the natural language model is completed, it is configured as a service on the server, and a user interface is provided through the H5 page of the browser. This guiding service supports multi-modal interaction methods such as voice and text input, and provides functions such as voice navigation and text Q&A, enhancing the guiding experience of tourists.

[0063] Generally speaking, the present invention can not only update the scenic area information in real time, but also provide accurate guiding services according to the time, location, and needs of tourists, greatly improving the satisfaction and overall experience of tourists, and providing a new method for the development of the intelligent tourism industry.

[0064] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. An intelligent spatio-temporal guided tour method for scenic spots based on digital twins, characterized in that, Use digital twin technology to build a virtual three-dimensional space model of the scenic area, construct a spatio-temporal intelligent Q&A dataset containing spatio-temporal location information, user questions and answers through retrieval-augmented generation technology, and use the spatio-temporal intelligent Q&A dataset to fine-tune a pre-trained natural language model to ensure that the natural language model responds accurately to the specific needs of the scenic area. The trained natural language model is then applied as a general natural language model for the cultural and tourism industry to the scenic area server, and voice navigation and text Q&A are realized through the browser terminal user interface. The specific steps are as follows: Step 1. Use digital twin technology to build a virtual three-dimensional space model of the scenic area, and obtain the geographical information, infrastructure locations, and spatial coordinate data of each scenic spot in the scenic area; Step 2. Based on the general Q&A pairs of the scenic area, personalized Q&A pairs, and text content about the scenic area crawled from the Internet, use an agent that can move within the digital twin model of the scenic area to link the natural language model to generate Q&A pairs containing different spatio-temporal information, and finally integrate them into a spatio-temporal intelligent Q&A dataset; Step 3. Use the spatio-temporal intelligent Q&A dataset to fine-tune the pre-trained natural language model to obtain a general natural language model for the cultural and tourism industry, apply the general natural language model for the cultural and tourism industry to the spatio-temporal intelligent guided tour module of the scenic area, and realize the update through the update module of the spatio-temporal intelligent guided tour of the scenic area.

2. The intelligent spatio-temporal scenic area navigation method based on digital twin according to claim 1, wherein, In step 1, the information of the scenic area and the spatial coordinate data of each scenic spot are obtained according to the following method: Step 1.1 Use satellite remote sensing technology to collect images of the scenic area, extract the geographical information and features of the scenic area through image matching technology, and generate a three-dimensional space model with surface geometry for the processed scenic area images through structured light method, dense reconstruction technology, and RealityCapture software, and convert the features of the terrain, buildings, and roads in the scenic area into polygon meshes; Step 1.2 Use geographic information system technology to complete the spatial information annotation of the three-dimensional model obtained in step 1.1 to ensure that tourists can find the corresponding location in the real world according to the virtual tour guide.

3. The intelligent spatio-temporal guided tour method for scenic spots based on digital twin according to claim 1, wherein, In step 2, the steps for obtaining the spatio-temporal intelligent Q&A dataset are as follows: Step 2.1 Create a knowledge base of the scenic area as the basic data source, and the dynamic information in the basic data source is updated in a timely manner by the staff of the scenic area; Step 2.2 Use natural language technology to build a knowledge graph of all scenic spots in the scenic area, and associate the basic data sources of the scenic spots through nodes, edges, and attributes in the knowledge graph; Step 2.3 Convert all the information in the knowledge graph of the scenic area constructed in step 2.2 into natural language corpus pairs to obtain personalized Q&A pairs for the scenic area, and then combine the general Q&A pairs of the scenic area and the text content of the scenic area crawled from the Internet to build a spatio-temporal intelligent Q&A database of the scenic area.

4. The scenic area intelligent spatio-temporal navigation method based on digital twin according to claim 1, wherein, In step 3, the method for fine-tuning the general natural language model for the cultural and tourism industry is as follows: Step 3.1 Build a general natural language model evaluation benchmark for the cultural and tourism industry as a reference standard for the fine-tuning results; Step 3.2 Select a pre-trained natural language model for fine-tuning. Fine-tune the pre-trained natural language model according to the spatio-temporal intelligent Q&A database of the scenic area obtained in Step 2. When the fine-tuning result of the natural language model meets the general natural language model evaluation benchmark for application in the cultural and tourism industry, the fine-tuning of the general natural language model for application in the cultural and tourism industry is completed.

Images