A big data-based tourism information pushing method, system, device and medium

Abstract

The application discloses a big data-based tourism information pushing method, system, device and medium. Voice feedback data submitted by a user after virtual reality tourism experience is collected, and core text data expressing tourism demand is extracted therefrom. The core text data is input into a semantic understanding model, which maps the core text data, theme keywords of a scene theme library and interest point labels of an interest point knowledge base to a same semantic vector space, and calculates first and second semantic correlations between the core text data vector and the theme keywords and the interest point labels. At least one target tourism theme is identified according to the first semantic correlation, a candidate interest point label set is determined in combination with a predefined mapping relationship between the tourism theme and the interest point labels, at least one target interest point label with the highest correlation is screened out according to the second semantic correlation, and pre-stored user preference data is combined to retrieve associated tourism recommendation content from a tourism resource database and directly push the content to a user terminal, so that the information pushing accuracy can be improved.

Description

Technical Field

[0001] This application relates to the field of information processing technology, and more specifically, to a method, system, device, and medium for pushing tourism information based on big data. Background Technology

[0002] With the rapid development of big data and virtual reality (VR) technologies, tourism information push has become an important means of enhancing users' travel experience. Traditional tourism information push methods mainly rely on manual user searches or platform-preset recommendation algorithms, such as static pushes based on browsing history or location data. However, these methods have significant limitations in handling real-time user feedback, failing to effectively capture users' immediate needs after a VR travel experience, resulting in low matching rates between recommended content and users' actual interests and low push efficiency. With the digital transformation of the tourism industry, users' expectations for personalized and immersive tourism services are increasing, and intelligent push systems are gradually becoming a core technological requirement to meet diverse user needs.

[0003] Existing tourism information push technologies generally suffer from static and generalized data processing issues, making it difficult to adapt to the dynamic semantics of user voice feedback. This is especially true in VR tourism scenarios, where user feedback often involves multimodal expressions (such as emotions and preferences). Traditional methods fail to introduce semantic models that reflect the non-linear relationship between user needs and tourism resources, and lack analysis and optimization strategies for key semantic relevance. Consequently, push solutions struggle to adapt to user changes and cannot fully realize their potential to improve user satisfaction. Therefore, how to optimize tourism information push based on big data to improve accuracy has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method, system, device, and medium for pushing tourism information based on big data, which can optimize the push of tourism information and improve the accuracy of the push.

[0005] Firstly, this application provides a method for pushing tourism information based on big data, the method comprising:

[0006] The system collects voice feedback data submitted by users through designated terminals after their virtual reality tourism experiences via a big data platform, and extracts core text data expressing users' tourism needs from the voice feedback data.

[0007] The core text data is input into the semantic understanding model, which maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space.

[0008] In the semantic vector space, the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance with each interest point tag vector are calculated simultaneously.

[0009] Based on the first semantic relevance, at least one target tourism theme is identified from the scene theme thesaurus. Based on the predefined mapping relationship between tourism themes and point of interest tags, a set of candidate point of interest tags associated with the at least one target tourism theme is determined. Based on the second semantic relevance, at least one target point of interest tag with the highest relevance is selected from the set of candidate point of interest tags.

[0010] Based on the at least one target point of interest tag and combined with pre-stored user preference data, relevant tourism recommendation content is retrieved from the tourism resource database and directly pushed to the user's terminal device.

[0011] Preferably, the process involves collecting voice feedback data submitted by users through designated terminals after their virtual reality tourism experience via a big data platform, and extracting core text data expressing the user's tourism needs from the voice feedback data, specifically including:

[0012] The voice feedback data is converted into text data using speech recognition technology;

[0013] Stop words and noise words are removed from the text data, retaining key phrases that express users' travel needs;

[0014] The key phrases are combined into the core text data.

[0015] Preferably, the core text data is input into a semantic understanding model, which maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space. Specifically, this includes:

[0016] The core text data, the topic keywords, and the point of interest tags are each input as an independent text unit into the same semantic encoder;

[0017] The semantic encoder encodes each individual text unit into a fixed-dimensional semantic vector.

[0018] The core text data vector, the topic keyword vector, and the interest point tag vector are placed in the same semantic vector space.

[0019] Preferably, in the semantic vector space, simultaneously calculating the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance with each interest point tag vector specifically includes:

[0020] For the core text data vector and each topic keyword vector, calculate the cosine similarity between them, which is used as the first semantic relevance.

[0021] For the core text data vector and each interest point label vector, the cosine similarity between them is calculated as the second semantic relevance.

[0022] Preferably, based on the first semantic relevance, at least one target tourism theme is identified from the scene topic thesaurus, and based on the predefined mapping relationship between tourism themes and point-of-interest tags, the set of candidate point-of-interest tags associated with the at least one target tourism theme is determined, specifically including:

[0023] The topic keywords are sorted from high to low based on the first semantic relevance, and the top N topic keywords are selected as the target tourism topic, where N is a preset positive integer;

[0024] Based on the predefined mapping relationship between tourism themes and point of interest tags, the mapping relationship table is queried to obtain the point of interest tags corresponding to the target tourism theme, forming the candidate point of interest tag set.

[0025] Preferably, based on the second semantic relevance, selecting at least one target interest point label with the highest relevance from the candidate interest point label set specifically includes:

[0026] The top K interest point labels with the highest second semantic relevance are selected from the candidate interest point label set as the target interest point labels, where K is a preset positive integer.

[0027] Preferably, based on the at least one target point of interest tag and combined with pre-stored user preference data, retrieving associated tourism recommendation content from the tourism resource database and directly pushing the tourism recommendation content to the user's terminal device specifically includes:

[0028] The target point of interest tags are weighted and fused with the historical preference tags in the user preference data to obtain the matched point of interest tags.

[0029] Using the matched point-of-interest tags as query conditions, a search operation is performed in the tourism resource database to obtain the tourism recommendation content;

[0030] The travel recommendations are sent to the terminal device via a designated network interface.

[0031] Secondly, this application provides a tourism information push system based on big data, the system comprising:

[0032] The data collection module is used to collect voice feedback data submitted by users through a designated terminal after their virtual reality tourism experience via a big data platform, and to extract core text data expressing the user's tourism needs from the voice feedback data.

[0033] The processing module is used to input the core text data into the semantic understanding model, which maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space.

[0034] The processing module is also used to simultaneously calculate, in the semantic vector space, the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance between the core text data vector and each interest point tag vector;

[0035] The processing module is further configured to identify at least one target tourism theme from the scene theme lexicon based on the first semantic relevance, determine a set of candidate point of interest tags associated with the at least one target tourism theme based on the predefined mapping relationship between tourism themes and point of interest tags, and filter at least one target point of interest tag with the highest relevance from the set of candidate point of interest tags based on the second semantic relevance.

[0036] The execution module is used to retrieve relevant tourism recommendation content from the tourism resource database based on the at least one target point of interest tag and pre-stored user preference data, and to directly push the tourism recommendation content to the terminal device used by the user.

[0037] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described method for pushing tourism information based on big data.

[0038] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-mentioned big data-based tourism information push method.

[0039] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0040] In this embodiment, a big data platform collects voice feedback data submitted by users through a designated terminal after a virtual reality tourism experience. Core text data expressing the user's tourism needs is extracted from this voice feedback data. This core text data is input into a semantic understanding model, which maps the core text data, keywords from a scene-theme lexicon, and interest point tags from an interest point knowledge base to the same semantic vector space. In this semantic vector space, a first semantic relevance between the core text data vector and each keyword vector, and a second semantic relevance with each interest point tag vector, are simultaneously calculated. Based on the first semantic relevance, at least one target tourism theme is identified from the scene-theme lexicon. According to a predefined mapping relationship between tourism themes and interest point tags, a set of candidate interest point tags associated with the at least one target tourism theme is determined. Based on the second semantic relevance, at least one target interest point tag with the highest relevance is selected from the candidate interest point tag set. Based on the at least one target interest point tag and pre-stored user preference data, related tourism recommendations are retrieved from a tourism resource database and directly pushed to the user's terminal device.

[0041] Therefore, this application calculates the first and second semantic relevance through semantic vector space to obtain the target tourism theme and target point of interest tags, and constructs a push strategy based on user preference data. First, by extracting core text data, it can reveal the core semantics of user needs based on the data, thereby establishing a semantically grounded push framework. This data extraction process, through speech recognition and keyword retention, achieves a quantitative expression of feedback data, and performs semantic mapping on this basis, enhancing the model's adaptability to actual user data from the source. Second, by determining the target tourism theme and candidate point of interest tag set through semantic vector space, it can identify the parameter range that significantly affects the push content, effectively avoiding push notifications. This approach avoids the blindness of traditional push strategies, prioritizing semantic relevance optimization to achieve targeted filtering of interest point tags. Then, it matches interest points based on user preferences, using retrieval methods to find relevant travel recommendations, thus outputting quantitatively supported push parameters, significantly improving the accuracy and real-time performance of the push process. Finally, it determines travel recommendations based on the matching results and constructs a push strategy, further strengthening the user adaptation mechanism across multiple needs and minimizing the negative impact of generalized pushes on user experience. This mechanism, through dynamic optimization, constructs an adaptive path for travel information push in big data scenarios, effectively overcoming the problem of unstable matching due to static dependencies in traditional methods. In summary, this application's solution can optimize travel information push, thereby improving its accuracy. Attached Figure Description

[0042] Figure 1 This is an exemplary flowchart of a big data-based tourism information push method according to some embodiments of this application;

[0043] Figure 2 This is a flowchart illustrating the process of extracting core text data expressing user travel needs, based on some embodiments of this application.

[0044] Figure 3 This is a flowchart illustrating the process of uniformly mapping to the same semantic vector space according to some embodiments of this application;

[0045] Figure 4 This is a schematic diagram of the structure of a tourism information push system based on big data, according to some embodiments of this application;

[0046] Figure 5 This is a schematic diagram of the structure of a computer device that implements a big data-based tourism information push method according to some embodiments of this application. Detailed Implementation

[0047] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] refer to Figure 1 The figure is an exemplary flowchart of a big data-based tourism information push method according to some embodiments of this application. The big data-based tourism information push method mainly includes the following steps:

[0049] In step 101, the voice feedback data submitted by users through a designated terminal after their virtual reality tourism experience is collected through a big data platform, and the core text data expressing the user's tourism needs is extracted from the voice feedback data.

[0050] It should be noted that the big data platform in this application is a system based on a distributed storage and computing framework, used for real-time collection and processing of massive amounts of user data. This big data platform can be implemented using the Hadoop or Spark framework. This is only an example and does not constitute a specific limitation on the present invention.

[0051] In some embodiments, reference Figure 2 The figure is a flowchart illustrating the extraction of core text data expressing user travel needs according to some embodiments of this application. The extraction of core text data expressing user travel needs from the voice feedback data can be achieved in the following manner:

[0052] In step 1011, the voice feedback data is converted into text data by using speech recognition technology. Specifically, it can be implemented in the following way: the voice feedback data is input into a pre-trained speech recognition model, such as a speech recognition model based on the Transformer architecture. This speech recognition model converts the audio signal into a corresponding text sequence through temporal feature extraction and decoding. Then, background noise interference is removed from the converted text data and common recognition errors are corrected. For example, homophones are corrected through an n-gram language model. Details are not elaborated here.

[0053] In step 1012, stop words and noise words are removed from the text data, and key phrases expressing the user's travel needs are retained. Specifically, it can be implemented in the following way: First, a pre-defined stop word list and noise word list are loaded. The stop word list contains common words such as "de" (of) and "shi" (is), and the noise word list contains filler words such as "e" and "en". Then, the text data is tokenized. For example, the jieba or NLTK library can be used to split the text into a sequence of words. Each word in the sequence is traversed. If the word is in the stop word list or noise word list, the word is removed. Finally, phrases expressing travel needs in the remaining word sequence, such as "beach vacation" and "historical monuments", are retained. Details are not elaborated here.

[0054] In step 1013, the key phrases are combined into the core text data. Specifically, it can be implemented in the following way: semantic clustering is performed on the retained key phrases, and similar phrases are grouped. For example, the K-means algorithm can be used to group similar phrases based on word vector embedding (such as Word2Vec). This is only for illustrative purposes and does not specifically limit this application. Then, the phrases in the group are sorted in descending order of frequency of occurrence, and the top phrases are selected as representatives. Finally, the representative phrases are concatenated with conjunctions (such as "and" and "or") and punctuation according to the original text semantic logic to form a grammatically coherent core text data, such as "like beach vacation and historical monuments", ensuring that the length of the core text data is controlled within 50 - 200 characters. Details are not elaborated here.

[0055] In step 102, the core text data is input into a semantic understanding model, which maps the core text data, the theme keywords in the scene theme word library, and the interest point labels in the interest point knowledge base to the same semantic vector space.

[0056] It should be noted that the semantic understanding model in this application is based on the framework of a pre-trained language model, such as BERT or RoBERTa. This model is adapted to semantic tasks in the travel domain through fine-tuning. This is only for illustrative purposes and can also be other types of semantic understanding models, which do not specifically limit this application.

[0057] In some embodiments, reference Figure 3 The diagram illustrates a process of uniformly mapping to the same semantic vector space according to some embodiments of this application. The core text data is input into a semantic understanding model, which then maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space. This can be achieved in the following way:

[0058] In step 1021, the core text data, the topic keywords, and the point of interest tags are input as independent text units into the same semantic encoder. Specifically, this can be achieved as follows: First, each independent text unit is preprocessed, including adding [CLS] and [SEP] tags; then, the preprocessed units are input into the semantic encoder in batches. The semantic encoder can be, for example, a Transformer encoder, ensuring a batch size of 32 to optimize computational efficiency. In practice, other types of semantic encoders can also be used. This is only an example and is not intended to limit the scope of this application.

[0059] In step 1022, each independent text unit is encoded into a fixed-dimensional semantic vector by the semantic encoder. Specifically, this can be achieved by the semantic encoder calculating the context representation of each unit through a multi-layer self-attention mechanism; then, the output of the [CLS] tag is taken as the overall semantic vector of the unit, with the dimension consistent with the semantic encoder architecture (e.g., 768 dimensions for BERT-base and 4096 dimensions for ALBERT-base); finally, the vector is L2 normalized to ensure that the vector lies on a unit sphere, which will not be elaborated here.

[0060] In step 1023, the core text data vector, the topic keyword vector, and the interest point tag vector are placed in the same semantic vector space. Specifically, this can be achieved by storing all vectors in the same vector index structure, for example, using the Faiss library to build an inner product index, which supports efficient similarity queries. This will not be elaborated further here.

[0061] In step 103, in the semantic vector space, the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance with each point of interest tag vector are calculated simultaneously.

[0062] In some embodiments, the simultaneous calculation of the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance with each interest point label vector, can be implemented in the following manner: For the core text data vector and each topic keyword vector, the cosine similarity between them is calculated as the first semantic relevance, specifically by dividing the vector dot product by the modulus product; For the core text data vector and each interest point label vector, the cosine similarity between them is calculated as the second semantic relevance; Simultaneously, the data is split into batches (batch size = 64) and allocated to GPU cores to calculate the semantic relevance between the core text vector and each batch vector in parallel to ensure efficiency, which will not be elaborated here.

[0063] In step 104, at least one target tourism theme is identified from the scene theme thesaurus based on the first semantic relevance. Based on the predefined mapping relationship between tourism themes and point of interest tags, a set of candidate point of interest tags associated with the at least one target tourism theme is determined. Based on the second semantic relevance, at least one target point of interest tag with the highest relevance is selected from the set of candidate point of interest tags.

[0064] In some embodiments, at least one target tourism theme is identified from the scenario topic thesaurus based on the first semantic relevance. The set of candidate point-of-interest tags associated with the at least one target tourism theme is determined according to a predefined mapping relationship between tourism themes and point-of-interest tags. This can be achieved by: sorting the topic keywords from high to low based on the first semantic relevance, selecting the top N topic keywords as the target tourism theme, where N is a preset positive integer, for example, N=5; then, querying a predefined mapping relationship table (e.g., a relational table in a database) to obtain the point-of-interest tags corresponding to the target tourism theme, thus forming the set of candidate point-of-interest tags. Further details are omitted here.

[0065] In some embodiments, the selection of at least one target interest point label with the highest relevance from the candidate interest point label set based on the second semantic relevance can be achieved in the following manner: the top K interest point labels with the highest second semantic relevance from the candidate interest point label set are selected as the target interest point labels, where K is a preset positive integer, such as K=3, which will not be elaborated here.

[0066] In step 105, based on the at least one target point of interest tag and combined with pre-stored user preference data, relevant tourism recommendation content is retrieved from the tourism resource database, and the tourism recommendation content is directly pushed to the terminal device used by the user.

[0067] In some embodiments, retrieving associated tourism recommendations from a tourism resource database based on the at least one target point of interest tag and pre-stored user preference data, and directly pushing the tourism recommendations to the user's terminal device, can be achieved in the following way: performing a weighted fusion operation between the target point of interest tag and the historical preference tags in the user preference data to obtain a matched point of interest tag; using the matched point of interest tag as a query condition to perform a retrieval operation in the tourism resource database to obtain the tourism recommendations, for example, through SQL query or Elasticsearch index retrieval; selecting the transmission method according to the content size: text files are pushed via HTTP POST, and large files (images / videos) are pushed via HTTP / 2 protocol (supporting breakpoint resumption), which will not be elaborated here.

[0068] This application first uses semantic relevance analysis to filter target tourism themes and interest tags, and then combines this with user preferences for retrieval and push notifications. This process effectively avoids the blindness of traditional static push notifications, significantly improving the matching accuracy of recommended content and user satisfaction. Compared with the generalized recommendation model used in existing technologies, this application, through dynamic semantic mapping and relevance calculation, effectively reduces the negative impact of irrelevant content on user experience, thereby improving overall push efficiency and the economic benefits of tourism services. It can effectively meet the refined needs of tourism information push based on big data.

[0069] On the other hand, in some embodiments, this application provides a tourism information push system based on big data, referencing... Figure 4 The figure is a schematic diagram of the structure of a tourism information push system based on big data according to some embodiments of this application. The tourism information push system 400 based on big data includes: a collection module 401, a processing module 402, and an execution module 403, which are described below:

[0070] The acquisition module 401 in this application is mainly used to collect voice feedback data submitted by users through a designated terminal after experiencing virtual reality tourism through a big data platform, and to extract core text data expressing the user's tourism needs from the voice feedback data.

[0071] Processing module 402, in this application, is used to input the core text data into the semantic understanding model, which maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space.

[0072] The processing module 402 is also used to simultaneously calculate, in the semantic vector space, the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance between the core text data vector and each interest point tag vector.

[0073] The processing module 402 is further configured to identify at least one target tourism theme from the scene theme lexicon based on the first semantic relevance, determine a set of candidate point of interest tags associated with the at least one target tourism theme based on the predefined mapping relationship between tourism themes and point of interest tags, and filter at least one target point of interest tag with the highest relevance from the set of candidate point of interest tags based on the second semantic relevance.

[0074] The execution module 403 in this application is mainly used to retrieve associated tourism recommendation content from the tourism resource database based on the at least one target point of interest tag and in combination with pre-stored user preference data, and to directly push the tourism recommendation content to the terminal device used by the user.

[0075] In addition, this application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the above-described method for pushing tourism information based on big data.

[0076] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device implementing a big data-based tourism information push method according to some embodiments of this application. The big data-based tourism information push method in the above embodiments can... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0077] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0078] The communication bus 502 can be used to transmit information between the aforementioned components. Its core transmission line is covered with a modified polyimide polymer insulation layer. This material possesses excellent dielectric properties and environmental stability, reducing signal transmission loss and isolating electromagnetic interference. The bus sheath is made of reinforced fluoropolymer composite material, which, with its high strength, wear resistance, and corrosion resistance, is suitable for the long-term dynamic operation requirements of the components, while its lightweight design reduces the overall assembly load. The low-temperature flexibility and anti-aging properties of the polymer material ensure that the bus continuously and stably transmits control commands and data signals under complex operating conditions, providing a reliable connection guarantee for the collaborative operation of the components.

[0079] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0080] The memory 503 stores program code for executing the solution of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. In the above embodiment, the tourism information push method based on big data can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0081] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0082] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0083] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0084] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for pushing tourism information based on big data.

[0085] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0086] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for pushing tourism information based on big data, characterized in that, include: The system collects voice feedback data submitted by users through designated terminals after their virtual reality tourism experiences via a big data platform, and extracts core text data expressing users' tourism needs from the voice feedback data. The core text data is input into the semantic understanding model, which maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space. In the semantic vector space, the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance with each interest point tag vector are calculated simultaneously. Based on the first semantic relevance, at least one target tourism theme is identified from the scene theme thesaurus. Based on the predefined mapping relationship between tourism themes and point of interest tags, a set of candidate point of interest tags associated with the at least one target tourism theme is determined. Based on the second semantic relevance, at least one target point of interest tag with the highest relevance is selected from the set of candidate point of interest tags. Based on the at least one target point of interest tag and combined with pre-stored user preference data, retrieve relevant tourism recommendation content from the tourism resource database and push the tourism recommendation content directly to the terminal device used by the user. Specifically, based on the first semantic relevance, at least one target tourism theme is identified from the scene topic thesaurus. The set of candidate point-of-interest (POI) tags associated with the at least one target tourism theme is determined according to a predefined mapping relationship between tourism themes and POI tags. The topic keywords are sorted from high to low based on the first semantic relevance, and the top N topic keywords are selected as the target tourism topic, where N is a preset positive integer; Based on the predefined mapping relationship between tourism themes and point of interest tags, the mapping relationship table is queried to obtain the point of interest tags corresponding to the target tourism theme, forming the candidate point of interest tag set.

2. The method according to claim 1, characterized in that, The system collects voice feedback data submitted by users through designated terminals after their virtual reality tourism experiences via a big data platform. The core text data expressing users' tourism needs is extracted from this voice feedback data, specifically including: The voice feedback data is converted into text data using speech recognition technology; Stop words and noise words are removed from the text data, retaining key phrases that express users' travel needs; The key phrases are combined into the core text data.

3. The method according to claim 1, characterized in that, The core text data is input into a semantic understanding model, which then maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space. Specifically, this includes: The core text data, the topic keywords, and the point of interest tags are each input as an independent text unit into the same semantic encoder; The semantic encoder encodes each individual text unit into a fixed-dimensional semantic vector. The core text data vector, the topic keyword vector, and the interest point tag vector are placed in the same semantic vector space.

4. The method according to claim 1, characterized in that, In the semantic vector space, the synchronous calculation of the first semantic relevance between the core text data vector and the vectors of each topic keyword, and the second semantic relevance with the vectors of each point of interest tag, specifically includes: For the core text data vector and each topic keyword vector, calculate the cosine similarity between them, which is used as the first semantic relevance. For the core text data vector and each interest point label vector, the cosine similarity between them is calculated as the second semantic relevance.

5. The method according to claim 1, characterized in that, Based on the second semantic relevance, selecting at least one target interest point label with the highest relevance from the candidate interest point label set specifically includes: The top K interest point labels with the highest second semantic relevance are selected from the candidate interest point label set as the target interest point labels, where K is a preset positive integer.

6. The method according to claim 1, characterized in that, Based on the at least one target point of interest tag and combined with pre-stored user preference data, retrieve relevant tourism recommendation content from the tourism resource database, and directly push the tourism recommendation content to the user's terminal device. Specifically, this includes: The target point of interest tags are weighted and fused with the historical preference tags in the user preference data to obtain the matched point of interest tags. Using the matched point-of-interest tags as query conditions, a search operation is performed in the tourism resource database to obtain the tourism recommendation content; The travel recommendations are sent to the terminal device via a designated network interface.

7. A tourism information push system based on big data, which uses the method described in any one of claims 1 to 6 to push tourism information, characterized in that, The system includes: The data collection module is used to collect voice feedback data submitted by users through a designated terminal after their virtual reality tourism experience via a big data platform, and to extract core text data expressing the user's tourism needs from the voice feedback data. The processing module is used to input the core text data into the semantic understanding model, which maps the core text data, the topic keywords in the scene topic thesaurus, and the interest point tags in the interest point knowledge base to the same semantic vector space. The processing module is also used to simultaneously calculate the first semantic relevance between the core text data vector and each topic keyword vector, and the second semantic relevance with each interest point tag vector in the semantic vector space. The processing module is further configured to identify at least one target tourism theme from the scene theme lexicon based on the first semantic relevance, determine a set of candidate point of interest tags associated with the at least one target tourism theme based on the predefined mapping relationship between tourism themes and point of interest tags, and filter at least one target point of interest tag with the highest relevance from the set of candidate point of interest tags based on the second semantic relevance. The execution module is used to retrieve relevant tourism recommendation content from the tourism resource database based on the at least one target point of interest tag and pre-stored user preference data, and to directly push the tourism recommendation content to the terminal device used by the user.

8. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the big data-based tourism information push method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the tourism information push method based on big data as described in any one of claims 1 to 6.

Images