Intelligent travel public service system based on digital travel

Abstract

The invention discloses a smart travel public service system based on digital travel, particularly relates to the field of digital travel, and is used for solving the problems of dynamic grading and synchronous regulation and control of language difficulty of existing digital travel services. The tourist portrait construction module extracts the age of the tourist based on the reservation information, and matches a corresponding education stage; the cognition grade evaluation module is used for extracting a text travel knowledge point keyword, calculating a cognition coverage degree based on a teaching material corpus in an education stage and generating a cognition grade label; the model training module is used for training a generative language model according to corpora corresponding to different cognitive levels; the multi-difficulty version generation module is used for constructing a semantic map of explanation contents of the exhibition items and generating a multi-difficulty version; the accompanying tourist division module is used for binding explanation equipment and tourist identifiers and dividing tourist groups according to reservation numbers; and the voice rhythm synchronization module identifies the cognitive level of the main group, performs semantic number alignment and rhythm adjustment on non-main group equipment, and realizes group synchronous explanation.

Description

Technical Field

[0001] The present invention relates to the field of smart cultural tourism technology, and more specifically, to a smart cultural tourism public service system based on digital tourism. Background Art

[0002] Intelligent voice commentary is a common feature of digital tourism. However, in multi-person tours, the tourist mix is ​​complex, often encompassing people of varying ages, cultural backgrounds, and cognitive abilities. Their understanding of the content presented by the commentary varies significantly. For example, children are more suited to concise, visually appealing, slower speech, and more accessible content; whereas adults or senior students may prefer professional commentary with complete structure, clear terminology, and a higher degree of semantic depth. Therefore, when voice commentary fails to dynamically grade and synchronize language difficulty across multiple participants, it is very easy to create a disconnected experience: children don't understand, and adults only listen superficially.

[0003] In addition, in the actual travel process, especially in multi-person travel modes such as family trips, parent-child tours, friends' companions, and group studies, voice interpretation systems of different difficulty levels often lack the necessary coordination mechanism, resulting in misaligned voice interpretation progress among fellow tourists and a split in the overall experience. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a smart cultural and tourism public service system based on digital tourism to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A smart cultural tourism public service system based on digital tourism, including a tourist portrait construction module, a cognitive level assessment module, a model training module, a multi-difficulty version generation module, a group tourist classification module, and a voice rhythm synchronization module;

[0007] The tourist portrait construction module creates a tourist portrait based on the platform's reservation registration information and matches the tourist's age to the corresponding education stage;

[0008] The cognitive level assessment module extracts keywords related to regional cultural and tourism knowledge points, calculates the cognitive coverage of the knowledge point keywords at the educational stage of the tourist profile, and establishes a cognitive level label for the tourist;

[0009] The model training module obtains the textbook corpus of the corresponding education stage based on the cognitive level labels and trains the multi-difficulty language generation model;

[0010] The multi-difficulty version generation module structures the audio explanation content of the exhibits into a sequence of semantic units and constructs a semantic map. Based on the multi-difficulty language generation model, it generates audio explanation difficulty versions corresponding to different cognitive level labels.

[0011] The group tourist classification module initializes the voice explanation difficulty level of the voice explanation device, binds the voice explanation device number with the tourist ID, and divides the group of tourists into groups;

[0012] The voice rhythm synchronization module identifies the main groups in the group of tourists traveling together, and synchronizes the voice rhythm of the voice explanations of all devices based on the difficulty level of the voice explanations of the main group.

[0013] In a preferred embodiment, the tourist portrait building module builds a tourist portrait based on the platform reservation registration information, and matches the tourist's age to the corresponding education stage, specifically including:

[0014] Call the reservation platform interface to obtain tourist registration information, extract the tourist ID, date of birth and reservation number, and build a tourist profile for the current reservation batch;

[0015] Set the corresponding relationship between age ranges and education stage labels, and establish a mapping rule table between age and education stage;

[0016] The tourist's age is calculated based on the birth date of the tourist portrait, and the age is matched and marked with the corresponding education stage according to the age and education stage mapping rule table to generate the education stage distribution of the tourists in the current reservation batch.

[0017] In a preferred embodiment, the cognitive level assessment module extracts keywords of regional cultural and tourism-themed knowledge points, calculates the cognitive coverage of the knowledge point keywords at the educational stage to which the tourist profile belongs, and establishes the tourist's cognitive level label, specifically including:

[0018] Extract textual materials from existing exhibits in the target cultural and tourism space, construct an exhibit corpus containing structural paragraphs, and filter the keyword set of exhibit knowledge points based on the set part-of-speech filtering rules;

[0019] Extract the textbook corpus from the regional textbook database and divide it according to the educational stage markers to establish the attribution relationship between the textbook corpus and the educational stage;

[0020] Keyword matching is performed on the keyword set of the knowledge points of the exhibition items and the corpus of textbooks at different educational stages. Based on the semantic similarity results calculated based on word vector mapping, a set of cognitive coverage parameters is established;

[0021] Extract the cognitive coverage parameter value corresponding to each tourist based on the education stage label marked in the tourist portrait of the current reservation batch;

[0022] Based on the preset cognitive level stratification rules and combined with the cognitive coverage parameter values, the tourists' cognitive level labels are divided and marked.

[0023] In a preferred embodiment, the model training module obtains a teaching material corpus corresponding to an educational stage based on cognitive level labels, and training a multi-difficulty language generation model specifically includes:

[0024] Extract the textbook corpus of the educational stage corresponding to the cognitive level label, and use the cognitive level label as the model layer label to build a multi-level textbook corpus;

[0025] Perform unified formatting on each entry in the multi-level textbook corpus to generate a multi-level text training sample set;

[0026] Generative language model training is performed based on a multi-level text training sample set, and the language weight structure after training is recorded;

[0027] An index is created for the results of each hierarchical model that has completed training, and the index is archived according to cognitive levels to generate a multi-difficulty language generation model corresponding to the cognitive level label.

[0028] In a preferred embodiment, the multi-difficulty version generation module structures the audio explanation content of the exhibit into a sequence of semantic units and constructs a semantic map. The generation of audio explanation difficulty versions corresponding to different cognitive level labels based on the multi-difficulty language generation model specifically includes:

[0029] The initial audio explanation of the cultural and tourism space exhibits is obtained and segmented, and the semantic unit sequence is extracted according to the pause structure and semantic coherence constraints between sentences.

[0030] Establish structural dependency relationships and extract core predicates for semantic unit sequences, generate a bidirectional semantic connection graph between language blocks, and encode the semantic connection graph into a semantic graph structure in vector form;

[0031] Based on the tourists' cognitive level labels, a multi-difficulty language generation model is called to generate content of multiple complexity and construct multi-voice explanation difficulty versions;

[0032] The cognitive level label is associated with the corresponding audio explanation difficulty version and loaded into the server cache.

[0033] In a preferred embodiment, the construction of the multi-voice explanation difficulty level version specifically includes:

[0034] Obtain the voice explanation content generated by each level of the multi-difficulty language generation model;

[0035] Compare the semantic graph structure of the voice explanation content generated by each hierarchical model with the initial voice explanation content of the exhibit;

[0036] If the semantic graph structure comparison results are consistent, the voice explanation content generated by the corresponding hierarchical model will be marked as the voice explanation difficulty version of that level;

[0037] If the semantic graph structure comparison results are inconsistent, the corresponding hierarchical model is requested to regenerate the voice explanation content.

[0038] In a preferred embodiment, the tourist group classification module initializes the voice explanation difficulty level of the voice explanation device, binds the voice explanation device number to the tourist ID, and classifies the tourist group specifically including:

[0039] Retrieve tourist sign-in data, activate the interpretation equipment number according to the order of tourist sign-in, and establish a mapping association between the equipment number and the tourist ID;

[0040] Call the server cache based on the tourist profile cognitive level label and write the corresponding audio explanation difficulty version in the cache into the audio resource buffer of the explanation device;

[0041] Obtain the reservation number of the tourist portrait and divide those with the same reservation number into groups of tourists traveling together.

[0042] In a preferred embodiment, the voice rhythm synchronization module identifies the main group of tourists traveling together, and synchronizes the voice rhythm of the semantic explanations of all devices based on the difficulty level of the voice explanation of the main group, specifically including:

[0043] Extract the tourist number from the binding relationship of the tour guide equipment of the group of tourists traveling together, retrieve the cognitive level labels of the tourist portraits in the group of tourists traveling together, and generate the frequency distribution of each cognitive level label;

[0044] Identify the cognitive level label with the highest frequency in the label frequency distribution as the main cognitive level of the corresponding group of tourists;

[0045] Obtain the audio guide difficulty levels corresponding to the main cognitive levels and mark them as the baseline version for the group of tourists traveling together;

[0046] Obtain the voice content in the device's voice resource buffer, extract the semantic unit sequence of the voice content, perform numbering, and uniformly map the numbers to the baseline version semantic sequence;

[0047] Comparing the semantic sequence of the baseline version with the semantic sequence of the non-baseline version, generating a speech rhythm adjustment strategy based on the time dimension difference of the number position;

[0048] The voice rhythm adjustment strategy is used to control the synchronization of the voice explanation rhythm of the tour guide equipment of a group of tourists traveling together.

[0049] The technical effects and advantages of the present invention's intelligent cultural and tourism public service system based on digital tourism are as follows:

[0050] The tourist profile construction module identifies educational hierarchies and calculates cognitive coverage based on regional cultural and tourism knowledge points, generating cognitive level labels. This in turn drives the model training module to construct language generation models with varying levels of difficulty based on cognitive levels. During the content generation phase, the system incorporates semantic graphs and semantic unit structures to reconstruct content with consistent semantics but varying levels of expression, meeting the comprehension needs of diverse tourists. Furthermore, the system categorizes grouped tourists through the check-in and device initialization processes, identifying key cognitive levels within these groups and shaping the dominant rhythm of the explanation.

[0051] On this basis, the system achieves group-oriented voice rhythm synchronization control, ensuring synchronized knowledge perception and consistent tour experience for tourists as they collaborate. The overall system integrates cognitive modeling, semantic generation, and group synchronization control capabilities, offering advantages such as high content matching accuracy, strong adaptability to tour version changes, and a superior collaborative tour experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a structural diagram of a smart cultural and tourism public service system based on digital tourism in the present invention. DETAILED DESCRIPTION

[0053] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0054] Example 1

[0055] Figure 1 The present invention provides a smart cultural tourism public service system based on digital tourism, which includes a tourist portrait construction module, a cognitive level assessment module, a model training module, a multi-difficulty version generation module, a group tourist classification module, and a voice rhythm synchronization module;

[0056] The tourist portrait construction module creates a tourist portrait based on the platform's reservation registration information and matches the tourist's age to the corresponding education stage;

[0057] The cognitive level assessment module extracts keywords related to regional cultural and tourism knowledge points, calculates the cognitive coverage of the knowledge point keywords at the educational stage of the tourist profile, and establishes a cognitive level label for the tourist;

[0058] The model training module obtains the textbook corpus of the corresponding education stage based on the cognitive level labels and trains the multi-difficulty language generation model;

[0059] The multi-difficulty version generation module structures the audio explanation content of the exhibits into a sequence of semantic units and constructs a semantic map. Based on the multi-difficulty language generation model, it generates audio explanation difficulty versions corresponding to different cognitive level labels.

[0060] The group tourist classification module initializes the voice explanation difficulty level of the voice explanation device, binds the voice explanation device number with the tourist ID, and divides the group of tourists into groups;

[0061] The voice rhythm synchronization module identifies the main groups in the group of tourists traveling together, and synchronizes the voice rhythm of the voice explanations of all devices based on the difficulty level of the voice explanations of the main group.

[0062] The tourist portrait building module builds a tourist portrait based on the platform's reservation registration information and matches the corresponding education stage according to the tourist's age.

[0063] The system obtains visitor reservation registration data through an interface, primarily from data uploaded by the reservation platform. It pulls all valid visitor registration records for the current time period. These original registration records contain multiple fields, with the key fields being visitor ID, date of birth, and reservation number. The visitor ID field uses the platform's unified visitor ID. These three fields are extracted and written into a structured visitor data table to construct a visitor profile index for the current reservation batch. Each visitor profile record uses the visitor ID as the primary key, with date of birth and reservation number as subsidiary fields.

[0064] After the visitor profile structure is constructed, a pre-set age-education stage mapping rule table is connected. This rule table is pre-set by the developer or dynamically loaded by operations and maintenance personnel. The structure is a two-dimensional mapping, with a set of age ranges (e.g., 6–7 years old, 8–9 years old, etc.) on one side and corresponding education stage labels (e.g., lower primary school, upper primary school, junior high school, regular high school, etc.) on the other. A one-to-one or one-to-many mapping relationship is established between each age range and education stage label, forming a complete age stage classification rule.

[0065] After establishing a link between the visitor profile and the mapping rule table, the system calculates each visitor's age. This calculation is based on the difference between the birth year and the current system date, using calendar years by default and rounding to integers. The system performs this calculation on each record in the visitor profile index table, generating an age field and writing it into the visitor profile structure. This age field is then fed into the mapping rule table's logical decision process, where the visitor's educational stage is determined through traversal or interval positioning. This stage label is then written directly into the visitor profile result set, completing the structure of the individual visitor profile.

[0066] The cognitive level assessment module extracts keywords of regional cultural and tourism knowledge points, calculates the cognitive coverage of the knowledge point keywords at the educational stage to which the tourist profile belongs, and establishes a cognitive level label for the tourist.

[0067] Taking the exhibit group in the target cultural and tourism space (system deployment target) as the target, the text materials provided by each exhibit are centrally organized. The original form of the exhibit material is a segmented description, involving content including but not limited to historical background, product technology, geographical features, and personal deeds. The exhibit material text is preliminarily processed, and each exhibit paragraph is divided into standard structured language blocks to form an exhibit corpus collection. The division method is based on paragraph logic, topic transitions, and language pause symbols. After completion, word segmentation is performed, and part-of-speech filtering rules are used to retain only nouns, verbs, and place names that represent specific things, actions, and domain terms, such as nouns (such as "silk", "city wall", and "terracotta warriors"), verbs (such as "sacrifice", "construction", and "migration"), and place names (such as "Chang'an", "Jiangnan", and "Dunhuang"). After processing, the core words with the most domain characteristics in each exhibit are extracted to form a keyword set of exhibit knowledge points.

[0068] The regional textbook corpus is retrieved from the textbook database of the local education publisher, covering content from primary school to high school. The corpus content is based on units or chapters and is archived according to the academic stage label. Each chapter retains the chapter number, title, paragraph content and the identification of the educational stage to which it belongs. In order to maintain the consistency of the semantic expression style, the text of the textbook paragraphs is uniformly cleaned in format, and non-text notes, abnormal punctuation segments and non-semantic content (such as "Please discuss with students" and "Extension activities" are removed to build an educational stage corpus index library, while ensuring that each educational stage subset in the corpus can be retrieved independently.

[0069] The correlation between the keywords of the exhibits and the textbook corpus is measured by the semantic similarity of word vectors. A general pre-trained word embedding model is used to vectorize and encode the keywords and the terms in the textbook paragraphs. For any exhibit keyword, the cosine similarity between it and the terms in the textbook paragraphs at different educational stages is calculated, and the maximum similarity and the corresponding paragraph number are retained. For example, the keyword "silk" in elementary school textbooks may have a high similarity with "textile" and "trade" in natural geography or historical fragments, while in advanced-level corpora, it may correspond to more complex semantics such as "civilization dissemination" and "China-foreign exchange". By traversing all keywords and summarizing the proportion of knowledge point keywords that match each tourist's educational stage, a mapping table between tourist numbers and cognitive coverage is formed.

[0070] The visitor profiles for the current reservation batch are labeled with their educational level. Using this as a search criterion, we extract the coverage parameter corresponding to each visitor's educational level, creating a corresponding record between the visitor's identifier and cognitive coverage. Each visitor's cognitive coverage parameter represents their semantic adaptability to the set of exhibit keywords at their educational level, reflecting their level of understanding and assimilation of the exhibit's content.

[0071] The cognitive level is divided according to the coverage parameter. The preset level rule table divides the coverage value range into several level segments (based on coverage or similar hit rate, combined with the specific setting of the knowledge point keyword volume). For example, the three levels are set, including low coverage (such as hit rate less than 20%), medium coverage (20% to 60%), and high coverage (greater than 60%). Based on the relationship between the parameter value and the interval, the cognitive level label that should be assigned to each tourist is determined, corresponding to the cognitive level labels of "low cognitive level", "medium cognitive level" and "high cognitive level".

[0072] The model training module obtains a teaching material corpus corresponding to an educational stage based on cognitive level labels and trains a multi-difficulty language generation model.

[0073] Based on the cognitive level tag set, a list of corresponding educational stages is extracted. Each tag item refers to one or more educational stages. For example, "low level" may refer to the middle stage of elementary school, "middle level" to junior high school, and "high level" to high school or university preparatory stage. The system extracts the corresponding paragraph content from the textbook database according to the cognitive tag index structure and constructs the original hierarchical textbook corpus. Each corpus segment records its original source (textbook name, chapter number), paragraph content, and cognitive level tag binding field.

[0074] After the hierarchical corpus structure is established, the corpus format needs to be standardized. Specifically, batch formatting is performed to remove unstructured knowledge-laden text content such as teaching prompts, exercise instructions, and interactive teaching language, while retaining the main body of text with complete semantic expression. Each text segment is stored in a standard structure with fields such as text content, hierarchical label, word count range, and semantic paragraph level.

[0075] Using a generative language model architecture (e.g., a GPT-like language model based on an autoregressive architecture), the model unifies the vocabulary, positional encoding structure, model depth, and attention layer parameters. Training is performed in batches on sample sets with each cognitive level label. During training, each batch is explicitly bound to a single cognitive label. Upon completion of training, a weight parameter structure is output, which includes an embedding weight matrix, an attention layer parameter tensor, a vocabulary vector, and a control dimension encoding. Each set of training results is encapsulated as a separate snapshot structure for the language generation model.

[0076] After training is complete, each model snapshot structure is numbered and archived. Archiving uses the cognitive level label as the index field to construct a model index table. Each record corresponds to a training version of the model weight parameter set and its input data identifier. For example, the model corresponding to "Low Cognitive Level - Elementary School Corpus Training" is numbered Model V1.0, and the model corresponding to "High Cognitive Level - High School Corpus Training" is numbered Model V3.0. The model index table supports query, retrieval, dynamic loading, and replacement, ensuring that content generation modules can directly point to the corresponding model version based on the cognitive level label when they are called.

[0077] The multi-difficulty version generation module structures the exhibit audio explanation content into a sequence of semantic units and constructs a semantic map, and generates audio explanation difficulty versions corresponding to different cognitive level labels based on a multi-difficulty language generation model.

[0078] The original voice explanation text content bound to the exhibits in the target cultural and tourism space is retrieved. This type of content is pre-recorded by the tour guide design department and stored in the content database in the form of transcribed text. The text is in a paragraph structure as a whole, and each paragraph corresponds to a set of semantic descriptions or narrative fragments in the explanation audio. There are usually sentence pause marks of varying intensities in the text (such as periods, commas, line breaks, etc.), as well as areas of weakened semantic cohesion (such as topic transitions, context switches, etc.). In order to achieve semantic structuring, the explanation text is first subjected to standardized cleaning processing to remove non-language noise (such as transcription markers, emotional markers, etc.), and then primary segmentation is performed based on the voice pause structure to divide the text into basic sentence units.

[0079] Semantic coherence is assessed at the semantic unit level. The degree of semantic linkage between sentences is determined based on semantic referential relationships and subject-verb consistency rules. The sequence is interrupted when strong coherence conditions are not met, ultimately extracting a sequence of sentences with minimal semantic independence and constructing a sequence of semantic units. Each semantic unit is a block of language expressing a central meaning, recording fields such as sequence position, word count, and keywords within the sentence, forming a structured input set. A dependency parser analyzes each semantic unit's subject-verb-object structure, extracting the central predicate, subject, and object components. The predicate semantic relationships between adjacent semantic units are then matched. For example, if the ending predicate of the current semantic unit is "establishment" and the beginning subject of the next unit is "channel," a semantic subject-object cohesion relationship is determined between the two units. Based on these dependency relationships, bidirectional connecting edges are established between the semantic units, and the connection strength and directionality are marked to form a semantic connection graph. This connection graph structure is vectorized and represented as a combination of node-word vectors to construct a semantic graph. This graph structure maintains context, supporting subsequent generation of multi-version content equivalence and local reconstruction control.

[0080] In the multi-difficulty language generation stage, based on the cognitive level label pre-labeled by each tourist, the language generation model version that matches the label is called from the language model index table. This model version is a generative language model trained with a layered corpus in the previous training phase. It can receive structured semantic input and output natural language expression text of the corresponding difficulty level while maintaining the consistency of semantic expression. The system submits content generation requests in sections according to the granularity of semantic units. The content returned by the model uses semantic units as basic blocks and replaces the original semantic sequence with expressions of different complexity. For example, in the output of the "high cognitive level" call version, more abstract words such as "tributary system", "cultural evolution", and "standard replacement" may appear, while the "low cognitive level" version generates replacement text with intuitive expressions such as "handing over gifts", "culture is passed on", and "clothes have changed".

[0081] Each semantic unit generates multiple versions, corresponding to different cognitive level labels. Ultimately, the system establishes a version binding relationship between the cognitive label and the generated content, organizes it into a hierarchical content structure table, and loads this table into the voice interpretation device server cache.

[0082] After extracting semantic units from the audio guide content, constructing a semantic graph, and training a multi-level language generation model based on cognitive level labels, the output of each model version must be verified for consistency in semantic structure with the original content of the exhibit. This ensures that the generated content does not experience semantic drift or contextual misalignment at varying levels of expression difficulty. To this end, a structural semantic comparison method is employed, based on a graph structure matching strategy, to achieve content screening and quality control.

[0083] Each version of the content is structurally restored, converting it back into a sequence of semantic units. A corresponding semantic connectivity graph is generated based on the same dependency extraction strategy as the original content. This generation method is consistent with the original exhibit content, including subject-verb-object structure recognition, keyword extraction, and semantic dependency strength determination. Ultimately, a graph structure composed of semantic unit nodes and edges is formed. The semantic graph generated from the original audio explanation content serves as the target semantic skeleton, referred to as the baseline graph; the explanation content graphs generated by the cognitive level models are considered candidate graphs.

[0084] The graph comparison adopts a bidirectional graph structure comparison method, and calculates the three indicators of node correspondence rate, edge retention rate and core path overlap between the two graphs in sequence. The node correspondence rate is completed by calculating the matching degree between the predicate nodes and the main noun nodes in the generated version and the original graph; the edge retention rate refers to the retention ratio of the same semantic connection structure in the two graphs; the core path overlap refers to the ratio of the subject-predicate-object path sequence in the graph that is completely consistent in the two graphs. If the above three indicators are all higher than the set threshold (specifically set according to the required model accuracy requirements, such as node correspondence rate>90%, edge retention rate>85%, path overlap>80%), the candidate version is considered to be consistent with the original content in the semantic skeleton. If the consistency conditions are met, the generated content corresponding to the cognitive level is marked as a valid explanation version, and this version is used as the explanation content output at the current cognitive level.

[0085] If any candidate version fails to meet the consistency threshold in the structural comparison, the generated version is deemed to have a semantic structure deviation and is considered invalid. In this case, the current model output is not used. Instead, the semantic unit sequence is resubmitted to the language generation model corresponding to the cognitive level, triggering a content regeneration request.

[0086] The traveling companion classification module initializes the voice explanation difficulty level of the voice explanation device, binds the voice explanation device number with the tourist identification, and divides the traveling companion groups.

[0087] When a visitor enters the park or arrives at a designated exhibit area, the system reads their check-in information. This check-in information originates from the entrance ticket inspection system or from Bluetooth or QR code check-in points within the area. It includes fields for visitor ID, check-in timestamp, and reservation number. The system sorts the check-in timestamp in ascending order to determine the order in which visitors enter. Once sorted, the available audio guide device numbers are activated sequentially. The device number is either a factory-set ID for the physical device or dynamically assigned by the platform based on the order in which the hardware is connected. Each activation operation binds the visitor ID to the assigned audio guide device number.

[0088] The visitor's cognitive level tag is read from the visitor profile structure. This tag is used to call the server cache to retrieve the audio guide version number that matches the tag. The audio guide versions stored in the cache are structured audio data organized into semantic units. Each unit contains a voice file, a text index, and playback control parameters. The system then loads this version content into the audio resource buffer of the visitor's associated audio guide device, in semantic unit order.

[0089] Grouping is performed based on the reservation number field recorded in the visitor profile. The reservation number is a unique identifier generated by the platform's reservation system and represents a single reservation batch. The system calculates the distribution of reservation numbers within the current check-in batch and groups visitors with the same reservation number into a logical group, called a group of traveling companions. This group structure is registered in the group index table, along with fields such as a list of member visitor numbers, a list of device numbers, and information on the distribution of cognitive level tags.

[0090] The voice rhythm synchronization module identifies the main group in the group of tourists traveling together, and synchronizes the voice rhythm of the voice explanations of all devices based on the difficulty level of the voice explanation of the main group.

[0091] Extract the list of all registered tourist numbers from the current group of tourists, and retrieve the corresponding cognitive level label by tourist number. This label is bound to the tourist portrait structure. Extract the labels of all tourists in the group and count the frequencies to construct a cognitive level label frequency distribution table. The frequency distribution table records the number of tourists with each type of label and their proportion in the current group. Perform the maximum frequency label identification operation in the frequency distribution table, and select the cognitive level label with the most occurrences as the main cognitive level label of the group of tourists (the group carrying this cognitive level label is regarded as the main group of the passing crowd). This label serves as the dominant control basis for the semantic expression style of the current group and as the benchmark label for playback scheduling control. Find the voice explanation difficulty version number bound to this label and record it as the benchmark version number mark.

[0092] Read the loaded voice content buffers from all the audio commentary devices in the current group and extract the semantic unit sequence of the audio commentary text on each device. Each sequence is structured content, including fields such as the semantic unit number, unit content, and expected playback duration. Based on the baseline version number, load the semantic unit number sequence under that version to form the baseline version semantic number sequence structure.

[0093] A unified mapping operation is performed on the semantic sequences of devices in non-baseline versions. By matching the semantic numbers of the original semantic units with the logical positions of the semantic units in the baseline version, the semantic numbers are aligned on the timeline. For example, if the baseline version discusses "papermaking technology" in unit 10, and a device is currently in unit 12, "The inventor of papermaking," it can be determined that its playback rhythm has exceeded the main group by two semantic units.

[0094] By calculating the difference in number position and converting it into the difference in playback duration, a table of playback delays for each non-baseline version device relative to the baseline version is generated. This table records the tempo offset for each device, expressed as a mapping of semantic number difference and time difference. A speech rate adjustment strategy is generated for the deviating device, primarily taking the form of "slowed playback + waiting" or "speeded playback + compressed pauses." The playback tempo adjustment strategy structure records parameters such as the target device number, adjustment method (acceleration or deceleration), adjustment tempo factor, and expected return position.

[0095] The pacing adjustment strategy is distributed to all non-baseline devices, which then execute the pacing control commands, adjusting the audio guide playback rate and pause rhythm to align the playback progress of all devices with the semantic position of the baseline version, achieving semantic pacing alignment. This process is repeated at a preset synchronization cycle, forming a dynamic pacing synchronization mechanism, ensuring that groups of tourists experience a consistent tour progress based on cognitive adaptation.

[0096] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0097] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0098] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0099] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0100] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0101] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0102] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0103] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0104] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0105] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A smart cultural tourism public service system based on digital tourism, characterized by: It includes a tourist portrait construction module, a cognitive level assessment module, a model training module, a multi-difficulty version generation module, a group tourist classification module, and a voice rhythm synchronization module; The tourist portrait construction module creates a tourist portrait based on the platform's reservation registration information and matches the tourist's age to the corresponding education stage; The cognitive level assessment module extracts keywords related to regional cultural and tourism knowledge points, calculates the cognitive coverage of the knowledge point keywords at the educational stage of the tourist profile, and establishes a cognitive level label for the tourist; The model training module obtains the textbook corpus of the corresponding education stage based on the cognitive level labels and trains the multi-difficulty language generation model; The multi-difficulty version generation module structures the audio explanation content of the exhibits into a sequence of semantic units and constructs a semantic map. Based on the multi-difficulty language generation model, it generates audio explanation difficulty versions corresponding to different cognitive level labels. The group tourist classification module initializes the voice explanation difficulty level of the voice explanation device, binds the voice explanation device number with the tourist ID, and divides the group of tourists into groups; The voice rhythm synchronization module identifies the main groups in the group of tourists traveling together, and synchronizes the voice rhythm of the voice explanations of all devices based on the difficulty level of the voice explanations of the main group.

2. The intelligent cultural tourism public service system based on digital tourism according to claim 1 is characterized in that: The tourist portrait building module builds a tourist portrait based on the platform's reservation registration information, and matches the tourist's age to the corresponding education stage, specifically including: Call the reservation platform interface to obtain tourist registration information, extract the tourist ID, date of birth and reservation number, and build a tourist profile for the current reservation batch; Set the corresponding relationship between age ranges and education stage labels, and establish a mapping rule table between age and education stage; The tourist's age is calculated based on the birth date of the tourist portrait, and the age is matched and marked with the corresponding education stage according to the age and education stage mapping rule table to generate the education stage distribution of the tourists in the current reservation batch.

3. The intelligent cultural tourism public service system based on digital tourism according to claim 2 is characterized in that: The cognitive level assessment module extracts keywords of regional cultural and tourism-related knowledge points, calculates the cognitive coverage of the knowledge point keywords according to the educational stage of the tourist profile, and establishes the tourist's cognitive level label. Specifically, the module includes: Extract textual materials from existing exhibits in the target cultural and tourism space, construct an exhibit corpus containing structural paragraphs, and filter the keyword set of exhibit knowledge points based on the set part-of-speech filtering rules; Extract the textbook corpus from the regional textbook database and divide it according to the educational stage markers to establish the attribution relationship between the textbook corpus and the educational stage; Keyword matching is performed on the keyword set of the knowledge points of the exhibition items and the corpus of textbooks at different educational stages. Based on the semantic similarity results calculated based on word vector mapping, a set of cognitive coverage parameters is established; Extract the cognitive coverage parameter value corresponding to each tourist based on the education stage label marked in the tourist portrait of the current reservation batch; Based on the preset cognitive level stratification rules and combined with the cognitive coverage parameter values, the tourists' cognitive level labels are divided and marked.

4. The intelligent cultural tourism public service system based on digital tourism according to claim 1 is characterized in that: The model training module obtains a corpus of teaching materials corresponding to an educational stage based on cognitive level labels, and trains a multi-difficulty language generation model, specifically including: Extract the textbook corpus of the educational stage corresponding to the cognitive level label, and use the cognitive level label as the model layer label to build a multi-level textbook corpus; Perform unified formatting on each entry in the multi-level textbook corpus to generate a multi-level text training sample set; Generative language model training is performed based on a multi-level text training sample set, and the language weight structure after training is recorded; An index is created for the results of each hierarchical model that has completed training, and the index is archived according to cognitive levels to generate a multi-difficulty language generation model corresponding to the cognitive level label.

5. The intelligent cultural tourism public service system based on digital tourism according to claim 1 is characterized in that: The multi-difficulty version generation module structures the audio explanation content of the exhibit into a sequence of semantic units and constructs a semantic map. The generation of audio explanation difficulty versions corresponding to different cognitive level labels based on the multi-difficulty language generation model specifically includes: The initial audio explanation of the cultural and tourism space exhibits is obtained and segmented, and the semantic unit sequence is extracted according to the pause structure and semantic coherence constraints between sentences. Establish structural dependency relationships and extract core predicates for semantic unit sequences, generate a bidirectional semantic connection graph between language blocks, and encode the semantic connection graph into a semantic graph structure in vector form; Based on the tourists' cognitive level labels, a multi-difficulty language generation model is called to generate content of multiple complexity and construct multi-voice explanation difficulty versions; The cognitive level label is associated with the corresponding audio explanation difficulty version and loaded into the server cache.

6. The intelligent cultural tourism public service system based on digital tourism according to claim 5 is characterized in that: The construction of the multi-voice explanation difficulty level version specifically includes: Obtain the voice explanation content generated by each level of the multi-difficulty language generation model; Compare the semantic graph structure of the voice explanation content generated by each hierarchical model with the initial voice explanation content of the exhibit; If the semantic graph structure comparison results are consistent, the voice explanation content generated by the corresponding hierarchical model will be marked as the voice explanation difficulty version of that level; If the semantic graph structure comparison results are inconsistent, the corresponding hierarchical model is requested to regenerate the voice explanation content.

7. The intelligent cultural tourism public service system based on digital tourism according to claim 1 is characterized in that: The group tourist classification module initializes the voice explanation difficulty level of the voice explanation device, binds the voice explanation device number with the tourist ID, and classifies the group tourist groups into: Retrieve tourist sign-in data, activate the interpretation equipment number according to the order of tourist sign-in, and establish a mapping association between the equipment number and the tourist ID; Call the server cache based on the tourist profile cognitive level label and write the corresponding audio explanation difficulty version in the cache into the audio resource buffer of the explanation device; Obtain the reservation number of the tourist portrait and divide those with the same reservation number into groups of tourists traveling together.

8. The intelligent cultural tourism public service system based on digital tourism according to claim 1 is characterized in that: The voice rhythm synchronization module identifies the main group of tourists in the group, and synchronizes the voice rhythm of the semantic explanations of all devices based on the difficulty level of the voice explanation of the main group. Specifically, the module includes: Extract the tourist number from the binding relationship of the tour guide equipment of the group of tourists traveling together, retrieve the cognitive level labels of the tourist portraits in the group of tourists traveling together, and generate the frequency distribution of each cognitive level label; Identify the cognitive level label with the highest frequency in the label frequency distribution as the main cognitive level of the corresponding group of tourists; Obtain the audio guide difficulty levels corresponding to the main cognitive levels and mark them as the baseline version for the group of tourists traveling together; Obtain the voice content in the device's voice resource buffer, extract the semantic unit sequence of the voice content, perform numbering, and uniformly map the numbers to the baseline version semantic sequence; Comparing the semantic sequence of the baseline version with the semantic sequence of the non-baseline version, generating a speech rhythm adjustment strategy based on the time dimension difference of the number position; The voice rhythm adjustment strategy is used to control the synchronization of the voice explanation rhythm of the tour guide equipment of a group of tourists traveling together.

Images