Museum self-service explanation system based on artificial intelligence

Abstract

The invention relates to a museum self-service explanation system based on artificial intelligence, and the system comprises an exhibit information storage module which is used for obtaining exhibit information in a museum, storing the exhibit information in a structured data form, and constructing an exhibit information database; the positioning module is used for acquiring a position positioning result of the visitor and matching exhibit data according to the position positioning result; the artificial intelligence processing module is used for calling the exhibit information corresponding to the position of the visitor from the exhibit information database according to the position positioning result, generating an explanation text, combining the explanation text with user preference, and outputting a final explanation text; the voice synthesis module is used for converting the final explanation text into voice and outputting the voice; and the interaction module is used for receiving an instruction of a visitor and feeding back the instruction to the artificial intelligence processing module. According to the self-service explanation system, the museum visiting experience is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of self-guided interpretation, and particularly to a museum self-guided interpretation system based on artificial intelligence. Background Art

[0002] As an emerging research field, museum intelligent interpretation technology is crucial in enhancing the dissemination efficiency of cultural heritage and the visitor experience. With the popularization of digital technology, museums are gradually shifting from traditional static displays to dynamic interactive experiences, which not only enhances visitors' understanding of exhibits but also promotes the modernization of cultural education. However, the rapid development of this field has also exposed many problems that urgently need to be broken through by technological innovation.

[0003] Currently, many museums' interpretation solutions rely mostly on static audio devices or simple mobile applications. Although these methods have low costs, they have significant limitations in practical applications: insufficient positioning accuracy leads to a disconnection between the interpreted content and the visitor's location, the information push lacks personalization and cannot be dynamically adjusted according to the visitor's interests, and the poor interactivity results in a single user experience. These defects directly affect the practicality and attractiveness of the interpretation system and limit the potential of museums in digital transformation.

[0004] In this context, the core challenges in this field are gradually emerging, among which the three technical factors of positioning accuracy, dynamic content generation, and personalized adaptation are particularly prominent. First, the positioning module needs to achieve high-precision real-time positioning in a complex indoor environment, but due to signal interference and building structures, it is difficult for existing technologies to stably output accurate positions. Second, the intelligent processing of exhibit information needs to generate adapted interpretation content according to the location and user preferences. However, current systems lack flexibility in dynamically adjusting the depth and breadth of the content. Finally, user preference learning relies on historical behavior data, but how to quickly and accurately predict interest points in limited interactions remains an unsolved problem. The deficiencies of these technical factors prevent the interpretation system from fully meeting the diverse needs of visitors and thus weaken the realization of the immersive experience.

[0005] Therefore, how to integrate high-precision positioning, dynamic content generation, and personalized preference learning to build an intelligent interpretation system that can provide accurate and flexible interpretations in real time according to the visitor's location and interests has become the key problem that this research urgently needs to solve. The solution to this problem will directly promote the transformation of museum interpretation technology from extensive services to refined and intelligent experiences. Summary of the Invention

[0006] The purpose of the present invention is to provide a museum self-guided interpretation system based on artificial intelligence, which significantly improves the museum visit experience.

[0007] To achieve the above purpose, the present invention provides the following solutions:

[0008] An AI-based self-guided museum tour system, comprising: an exhibit information storage module, a positioning module, an AI processing module, a speech synthesis module, and an interaction module;

[0009] The exhibit information storage module is used to obtain exhibit information in the museum, store the exhibit information in a structured data form, and construct an exhibit information database;

[0010] The positioning module is used to obtain the location positioning result of the visitor and match the exhibit data according to the location positioning result;

[0011] The AI processing module is used to retrieve the exhibit information corresponding to the location where the visitor is located from the exhibit information database according to the location positioning result, generate an explanation text, and output the final explanation text by combining the explanation text with the user preference;

[0012] The speech synthesis module is used to convert the final explanation text into speech for output;

[0013] The interaction module is used to receive the instructions of the visitor and feedback the instructions to the AI processing module.

[0014] Optionally, the exhibit information storage module includes: a parameter acquisition unit, an image optimization unit, and a database construction unit;

[0015] The parameter acquisition unit is used to obtain the exhibit material and form data, judge the category to which the exhibit belongs and determine the initial shooting parameter range by matching with the form feature database;

[0016] The image optimization unit is used to, if the results of the reflection detection or shadow detection exceed the preset threshold, regenerate the shooting picture by adjusting the light source angle and exposure parameters, and obtain the optimized picture data;

[0017] The database construction unit is used to extract the key detail features from the optimized picture data and construct an exhibit information database according to the key detail features, exhibit background, category, and process information.

[0018] Optionally, the positioning module includes: a positioning unit and a matching unit;

[0019] The positioning unit is used to obtain the sensor data and indoor map information, calculate the real-time position coordinates by fusing multi-source positioning signals, and perform smoothing processing on the signal interference by using the particle filter algorithm to obtain the location positioning result;

[0020] The matching unit is configured to perform position matching on the structured information in the exhibit information data according to the spatial relationship between the position positioning result and the exhibit, and obtain the exhibit data within the target range.

[0021] Optionally, calculating the real-time position coordinates by fusing multi-source positioning signals includes: interacting with sensors in the exhibition area through Bluetooth signals or WiFi signals, where the sensors detect changes in signal strength and upload data to generate the real-time position coordinates.

[0022] Optionally, the artificial intelligence processing module includes: an explanation text generation unit, a user preference learning unit, and a personalized adaptation unit;

[0023] The explanation text generation unit is configured to obtain the exhibit data within the target range, input the exhibit data into a large language model for semantic parsing and expansion, and generate dynamic explanation text that matches the position.

[0024] The user preference learning unit is configured to extract the user operation sequence from the historical interaction behavior data, classify the user operation sequence using a clustering algorithm, and obtain the preference feature vector.

[0025] The personalized adaptation unit is configured to perform weighted adjustment on the dynamic explanation text through an attention mechanism according to the preference feature vector and the dynamic explanation text, and obtain the final explanation text.

[0026] Optionally, obtaining the preference feature vector includes: using the K-means algorithm to convert the user operation sequence into a preference feature vector.

[0027] Optionally, performing weighted adjustment on the explanation text through an attention mechanism includes:

[0028] Performing preliminary fusion on the preference feature vector and the dynamic explanation text through an attention mechanism to obtain a preliminary weighted result;

[0029] According to the preliminary weighted result, obtain the distribution characteristics of content depth and content breadth, and determine the weight ratio of depth weighting and breadth adjustment;

[0030] According to the weight ratio of depth weighting and breadth adjustment, obtain the text content with enhanced depth and the output with expanded breadth, and obtain the final explanation text.

[0031] Optionally, converting the final explanation text into voice output includes: converting the final explanation text into corresponding explanation voice, and using a soft sound source parameter controller to convert the explanation voice into virtual voice data with preset sound characteristics.

[0032] The beneficial effects of the present invention are as follows: The method of the present invention obtains the position of the visitor through high-precision positioning technology, matches it with the structured information in the exhibit database, and determines the current exhibit and related content. Semantic parsing and expansion are performed on the exhibit information to generate dynamic explanation text. At the same time, the present invention analyzes the user's historical interaction behavior, uses a clustering algorithm to judge the interest tendency, and updates the preference feature vector. Combining the user's preferences and the dynamic explanation text, the depth and breadth of the content are adjusted through an attention mechanism to achieve personalized explanation output. The present invention can also receive the instructions of the visitor to further optimize the explanation content. This intelligent and personalized self-guided explanation method significantly improves the museum visit experience, enables visitors to obtain in-depth and comprehensive exhibit information according to their own interests, and effectively promotes cultural dissemination and knowledge popularization. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0034] Figure 1 It is a framework diagram of a museum self-guided explanation system based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0036] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0037] As Figure 1 shown, this embodiment provides a museum self-guided explanation system based on artificial intelligence, including: an exhibit information storage module, a positioning module, an artificial intelligence processing module, a voice synthesis module, and an interaction module;

[0038] The exhibit information storage module is used to obtain the exhibit information in the museum and store the exhibit information in a structured data form to construct an exhibit information database;

[0039] The positioning module is used to obtain the position positioning result of the visitor and match the exhibit data according to the position positioning result;

[0040] An artificial intelligence processing module, which is used to retrieve the exhibit information corresponding to the location where the visitor is from the exhibit information database according to the location positioning result, generate an explanation text, and combine the explanation text with the user preferences to output the final explanation text;

[0041] A speech synthesis module, which is used to convert the final explanation text into speech for output;

[0042] An interaction module, which is used to receive the instructions of the visitor and feedback the instructions to the artificial intelligence processing module.

[0043] Furthermore, the exhibit information storage module includes: a parameter acquisition unit, an image optimization unit, and a database construction unit;

[0044] The parameter acquisition unit is used to acquire the exhibit material and form data, and by matching with the form feature database, determine the category to which the exhibit belongs and determine the initial shooting parameter range;

[0045] The image optimization unit is used to, if the results of reflection detection or shadow detection exceed the preset threshold, regenerate the shooting picture by adjusting the light source angle and exposure parameters to obtain the optimized picture data;

[0046] The database construction unit is used to extract the key detail features from the optimized picture data, and construct an exhibit information database according to the key detail features, exhibit background, category, and process information.

[0047] Specifically, the exhibit information storage module is the core component of the museum's digital management, and its structured storage method provides a basis for the efficient retrieval and utilization of exhibit information. Specifically, each exhibit in the database is assigned a unique identifier, and the associated information fields are multi-dimensional. The historical background of the exhibit can be broken down into sub-items such as the era, the place of origin, and the culture to which it belongs. The benefit of refinement is that it can support more accurate queries and classifications.

[0048] Furthermore, a positioning unit and a matching unit;

[0049] The positioning unit is used to acquire the sensor data and indoor map information, calculate the real-time position coordinates by fusing multi-source positioning signals, and use the particle filter algorithm to smooth the signal interference to obtain the location positioning result;

[0050] The matching unit is used to perform position matching on the structured information in the exhibit information data according to the spatial relationship between the location positioning result and the exhibit, and obtain the exhibit data within the target range.

[0051] Furthermore, calculating the real-time position coordinates by fusing multi-source positioning signals includes: interacting with sensors in the exhibition area through Bluetooth signals or WiFi signals, where the sensors detect changes in signal strength and upload data to generate real-time position coordinates.

[0052] Specifically, when a user enters a museum with a smartphone, the phone interacts with sensors in the exhibition area via Bluetooth or Wi-Fi signals. When the user approaches a certain exhibit, the sensor detects a change in signal strength and uploads data to generate coordinates accurate to the meter level, such as "2 meters away from Exhibit A". This method ensures the real-time nature and geographical relevance of the location information, providing a basis for subsequent matching. To extract the spatial relationship between the current location and the exhibit from the high-precision location results, it should be noted first that location data usually contains three-dimensional coordinate information, such as x, y, and z values, which can be used to determine the direction and distance of the user relative to the exhibit. Exemplarily, in a museum, assuming the current location coordinates are (5, 3, 1) and the coordinates of a certain exhibit are (6, 3, 1), it can be calculated that the user is 1 meter away from the exhibit on the horizontal plane and at the same height. The extraction of this spatial relationship can be determined by a preset distance threshold. For example, setting 2 meters as the effective range, if the distance exceeds this value, it is considered that the user is not close to the exhibit. The advantage of this method is that it can quickly screen out the range of exhibits that the user is really interested in, avoiding interference from irrelevant information. For position matching with the structured information in the exhibit database, specifically, the database usually stores the identification, location coordinates, and associated content of the exhibits. For example, the name of the exhibit is "Bronze Tripod", the coordinates are (6, 3, 1), and the content includes historical background and manufacturing process, etc. In one possible implementation, the system will compare the current location with the coordinates of all exhibits in the database one by one to find the exhibit identification that is the closest and within the threshold. For example, it matches "Bronze Tripod". Preferably, a directional judgment can be introduced, combined with the user's orientation data (such as the information provided by the phone's gyroscope), to further confirm whether the user is facing the exhibit directly. The advantage of this matching method is that it not only improves the accuracy but also can dynamically adapt to the position changes brought about by the user's movement. After determining the current exhibit identification and the range of associated content, it should be noted that the range of associated content is not limited to single-exhibit information and may also include extended knowledge related to the exhibit. For example, the associated content of the "Bronze Tripod" can cover the cultural background of the dynasty to which it belongs, its connection with other exhibits, etc. For the implementation method of spatial relationship extraction, the calculation can be simplified by a preset spatial grid. Assume that the museum exhibition hall is divided into small areas of 1 meter × 1 meter, and several exhibit information is bound to each area. When the user enters a certain grid, the system directly calls the exhibit data of the corresponding area. The advantage of this grid method is to reduce the real-time calculation burden. Especially in areas with dense exhibits, it can quickly respond to the user's position changes. In one embodiment, if the user stands near (5, 3, 1), the system identifies that they are in "Area A" and directly pushes the content related to the "Bronze Tripod", rather than calculating the distance to all exhibits one by one. For the implementation of position matching, another way is to optimize by combining historical location data. For example, when the user moves from (3, 3, 1) to (5, 3, 1), the system can infer their points of interest based on the movement trajectory and give priority to matching the exhibits in front rather than on the side.This method can effectively improve the intelligence of matching. Especially in the scenario where multiple exhibits coexist, it can avoid pushing irrelevant information. Preferably, priority rules can also be set. For example, the priority of exhibits closer in distance is higher than that of those farther away, ensuring that users always obtain the most relevant explanations. When determining the scope of associated content, the experience can be enriched through content stratification. For example, the basic layer provides the exhibit name and brief introduction, such as "Bronze Ding, a cultural relic of the Shang Dynasty"; the extended layer adds details, such as "Used for sacrifices, with cloud and thunder patterns as decorations". In a possible implementation, the system dynamically adjusts the pushed content according to the user's stay time. If the stay exceeds 30 seconds, the extended layer is pushed. The advantage of this hierarchical design is that it not only meets the needs of quick browsing but also provides in-depth information for interested users, greatly enhancing the flexibility of the explanation.

[0053] Furthermore, the artificial intelligence processing module includes: an explanation text generation unit, a user preference learning unit, and a personalized adaptation unit;

[0054] The explanation text generation unit is used to obtain the exhibit data within the target range, input the exhibit data into the large language model for semantic parsing and expansion, and generate dynamic explanation text that matches the location;

[0055] The user preference learning unit is used to extract the user operation sequence from the historical interaction behavior data, classify the user operation sequence using a clustering algorithm, and obtain the preference feature vector;

[0056] The personalized adaptation unit is used to weight-adjust the dynamic explanation text through the attention mechanism according to the preference feature vector and the dynamic explanation text, and obtain the final explanation text.

[0057] Furthermore, obtaining the preference feature vector includes: using the K-means algorithm to convert the user operation sequence into a preference feature vector.

[0058] Specifically, for the requirement of obtaining the basic content corresponding to the current exhibit identifier and generating dynamic explanatory texts that match the location through a preset content generation model for semantic parsing and expansion of exhibit information, analysis and examples can be carried out from multiple technical themes. The exhibit identifier serves as the core index in the database, and its basic content may include simple fields such as name, age, and excavation location. For example, for an exhibit labeled "X-001", its basic content may be "Name: Bronze Ding, Age: Shang Dynasty, Excavation Location: Anyang City, Henan Province". The value of this basic content lies in providing the original material for subsequent semantic parsing. Specifically, the process of semantic parsing can be understood as disassembling these basic fields and endowing them with richer semantic associations. In one possible implementation, for the age information of the "Shang Dynasty", the model can expand relevant historical backgrounds, such as information on the political structure and the development level of bronze smelting technology in the Shang Dynasty. For example, "Shang Dynasty" may be parsed as "Approximately 3,600 to 3,100 years ago, with a highly developed material culture represented by bronze ware". This expansion not only increases the depth of information but also provides a historical basis for the explanatory text. It should be noted that during parsing, the category of the exhibit can be combined. For example, a bronze ding may be related to the function of sacrifice, and the model will preferentially extract semantic content related to religious use. In one embodiment, the generation of dynamic explanatory texts that match the location depends on the spatial context where the exhibit is located. For example, assuming that the "X-001" bronze ding is located near the entrance of the "Shang Dynasty Civilization Exhibition Hall" in the museum, the model will generate a guiding explanation based on this location, such as "What you are seeing now is a bronze ding from the Shang Dynasty. It is not only a heavy sacrificial vessel but also a symbol of the majesty of the royal power at that time". Preferably, if the exhibit is located deep in the exhibition hall, the text may be adjusted to "After seeing the previous display of Shang Dynasty jade wares, the bronze ding you are seeing now demonstrates another peak of smelting technology". This dynamic adjustment can closely match the explanatory text with the visitor's viewing path and enhance the sense of immersion. It can be understood that the generation model can also be expanded from the perspective of manufacturing techniques. For example, for the basic content of the bronze ding, the model may parse out "Casting technique: Lost-wax casting, Feature: The body of the vessel has intricate patterns". Based on this, the explanatory text can be generated as "This ding was cast using the lost-wax casting technique. The craftsmen shaped these vivid patterns through complex molds, demonstrating the superb skills of the Shang Dynasty craftsmen". This method not only enriches the text content but also stimulates the audience's interest in the process details.

[0059] Extracting the user operation sequence from historical interaction behavior data is the basis for understanding user behavior patterns. For example, in the scenario of intelligent museum tour guides, when users view exhibits through devices, a series of operations will occur, such as clicking on the details of a certain exhibit, the length of stay, or skipping some content. Exemplarily, assume that a certain user clicks on the process introduction of a bronze tripod multiple times in the "Shang Dynasty Bronze Ware Exhibition Area" and stays for an average of 2 minutes, while only stays for 10 seconds on jade exhibits. This operation sequence reflects the user's preference for process details. Specifically, when extracting this data, the timestamp of each user click, the exhibit identifier, and the operation type can be recorded to form a sequence such as "X-001 click - stay for 120 seconds - X-002 click - stay for 10 seconds". In a possible implementation, when using a clustering algorithm to classify behavior patterns, the K-means algorithm can be selected to transform the user operation sequence into a feature vector. For example, taking the stay time, click frequency, and exhibit category as dimensions, the vector of a certain user may be "Bronze ware: 120 seconds / 5 times, Jade: 10 seconds / 1 time". After clustering, it may be found that one type of user prefers process details, and the other type prefers historical backgrounds. It should be noted that the key to clustering lies in feature selection. The length of stay may directly reflect the concentration of interest, while the click frequency implies the depth of exploration. Preferably, if a user repeatedly views the "Casting Process" chapter on the bronze tripod page, they can be classified into the "Process Interest Cluster". When judging the user's interest tendency, it can be understood that the clustering result provides a basis for preference update. In an embodiment, if a certain user is assigned to the "Process Interest Cluster", the system can infer that they are sensitive to technical details. Exemplarily, combined with historical data, such as the user has searched for "lost wax process", this tendency can be further confirmed. Specifically, when updating the preference feature vector, the interest intensity can be represented by a numerical value, such as "Process: 0.9, History: 0.3". The higher the numerical value, the stronger the tendency. This vector can be dynamically adjusted. For example, if the user shows interest in historical backgrounds subsequently, the "History" value may rise to 0.6. After the system analyzes the weighted results, it extracts the distribution characteristics of the content depth and breadth. Depth refers to the degree of detail of a certain topic, such as the time and characters of a specific battle in the "Historical Evolution"; breadth refers to the coverage range of the topic, such as covering multiple aspects of politics and economy. Assuming that the distribution characteristics show that the depth accounts for 60% and the breadth accounts for 40%, the system determines the weight ratio of depth weighting and breadth adjustment accordingly. For example, the depth weighting is 0.6 and the breadth adjustment is 0.4. The implementation of depth weighting can be based on keyword expansion and context supplementation. This supplementation relies on the historical database to ensure accuracy and at the same time allows visitors to feel the richness of the content. Breadth expansion pays more attention to the diversified connection of topics. The advantage of this expansion is that visitors can understand the background of exhibits from multiple perspectives and enhance the breadth of knowledge acquisition.

[0060] Further, converting the final explanatory text into voice output includes: converting the final explanatory text into corresponding explanatory voice, and using a soft sound source parameter controller to convert the explanatory voice into virtual voice data with preset sound characteristics.

[0061] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An AI-based self-guided museum commentary system, characterized in that, Including: An exhibit information storage module, a positioning module, an artificial intelligence processing module, a speech synthesis module, and an interaction module; The exhibit information storage module is used to obtain the exhibit information in the museum and store the exhibit information in a structured data form to construct an exhibit information database; The positioning module is used to obtain the position positioning result of the visitor and match the exhibit data according to the position positioning result; The artificial intelligence processing module is used to retrieve the exhibit information corresponding to the visitor's location from the exhibit information database according to the position positioning result, generate an explanation text, and output a final explanation text by combining the explanation text with user preferences; The speech synthesis module is used to convert the final explanation text into speech for output; The interaction module is used to receive the visitor's instructions and feedback the instructions to the artificial intelligence processing module.

2. The museum self-guided explanation system based on artificial intelligence according to claim 1, wherein The exhibit information storage module includes: a parameter acquisition unit, an image optimization unit, and a database construction unit; The parameter acquisition unit is used to obtain the exhibit material and morphology data, judge the exhibit category by matching with the morphology feature database, and determine the initial shooting parameter range; The image optimization unit is used to, if the results of reflection detection or shadow detection exceed the preset threshold, regenerate the shooting picture by adjusting the light source angle and exposure parameters to obtain the optimized picture data; The database construction unit is used to extract key detail features from the optimized picture data and construct an exhibit information database according to the key detail features, exhibit background, category, and process information.

3. The museum self-guided explanation system based on artificial intelligence according to claim 1, characterized in that, The positioning module includes: a positioning unit and a matching unit; The positioning unit is used to obtain sensor data and indoor map information, calculate real-time position coordinates by fusing multi-source positioning signals, and perform smoothing processing on signal interference using a particle filter algorithm to obtain a position positioning result; The matching unit is used to perform position matching on the structured information in the exhibit information data according to the spatial relationship between the position positioning result and the exhibit to obtain the exhibit data within the target range.

4. The museum self-guided explanation system based on artificial intelligence according to claim 3, characterized in that, Calculating real-time position coordinates by fusing multi-source positioning signals includes: interacting with sensors in the exhibition area through Bluetooth signals or WiFi signals, the sensors detecting changes in signal strength and uploading data to generate the real-time position coordinates.

5. The self-guided museum interpretation system based on artificial intelligence according to claim 1, wherein The artificial intelligence processing module includes: an explanation text generation unit, a user preference learning unit, and a personalized adaptation unit; The explanation text generation unit is used to obtain the exhibit data within the target range, input the exhibit data into a large language model for semantic parsing and expansion, and generate a dynamic explanation text that matches the position; The user preference learning unit is used to extract the user operation sequence from the historical interaction behavior data, classify the user operation sequence using a clustering algorithm, and obtain a preference feature vector; The personalized adaptation unit is used to perform weighted adjustment on the dynamic explanation text through an attention mechanism according to the preference feature vector and the dynamic explanation text to obtain the final explanation text.

6. The museum self-guided explanation system based on artificial intelligence according to claim 5, wherein Obtaining a preference feature vector includes: using the K-means algorithm to convert the user operation sequence into a preference feature vector.

7. The museum self-guided explanation system based on artificial intelligence according to claim 5, characterized in that, Weighted adjustment of the explanation text through the attention mechanism includes: Using the attention mechanism to preliminarily fuse the preference feature vector and the dynamic explanation text to obtain a preliminary weighted result; According to the preliminary weighted result, obtain the distribution characteristics of the content depth and content breadth, and determine the weight ratio of depth weighting and breadth adjustment; According to the weight ratio of depth weighting and breadth adjustment, obtain the text content with enhanced depth and the output with expanded breadth, and obtain the final explanation text.

8. The museum self-guided explanation system based on artificial intelligence according to claim 1, characterized in that, Converting the final explanation text into voice output includes: converting the final explanation text into corresponding explanation voice, and using a soft sound source parameter controller to convert the explanation voice into virtual voice data with preset sound characteristics.

Images