Tourism intelligent interaction platform based on AI

Abstract

The invention provides an AI-based tourism intelligent interaction platform, and relates to the field of tourism intelligent interaction platforms, and the platform comprises the following units: a user portrait construction unit which is used for synthesizing basic information, historical behaviors and social media activity data of a user, and constructing an omnibearing user portrait; the dynamic itinerary planning unit dynamically plans an optimal itinerary according to the current position of the user, the weather, the traffic condition and the preference of the user; and the personalized recommendation unit is used for providing customized tourism recommendation for the user based on the user portrait and the real-time situation. And the intelligent travel photographing unit integrates an AI photographing guidance function and provides photographing position, angle and light suggestions for the user. According to the invention, omnibearing and intelligent tourism service is provided for the user; the user experience and the flexibility and accuracy of travel planning are improved, and the functions of travel partners with emotional resonance, intelligent travel photography guidance and the like are provided; meanwhile, data management and utilization are optimized.

Description

Technical Field

[0001] The present invention relates to the field of tourism intelligent interactive platform, and in particular to an AI-based tourism intelligent interactive platform. Background Art

[0002] With the rapid development of artificial intelligence technology, its application in the tourism industry is becoming more and more extensive. Traditional tourism service platforms can only provide basic itinerary planning, scenic spot introduction and other services, lacking personalized and intelligent elements. However, modern tourists have higher and higher requirements for tourism experience, and they hope to obtain more personalized and intelligent tourism services to meet their diverse needs.

[0003] In this context, the AI-based tourism intelligent interactive platform came into being; this platform provides users with comprehensive and intelligent tourism services by integrating multiple AI technologies, such as deep learning, natural language processing, sentiment analysis, etc.; these technologies can not only process and analyze large amounts of user data, but also perceive users' emotions and needs, thereby providing more intimate and personalized services.

[0004] Therefore, we made improvements to this and proposed an AI-based tourism intelligent interactive platform. Summary of the invention

[0005] The purpose of the present invention is to address the fact that the current traditional tourism service platform can only provide basic itinerary planning, scenic spot introduction and other services, and lacks personalized and intelligent elements.

[0006] In order to achieve the above-mentioned purpose of the invention, the present invention provides an AI-based tourism intelligent interactive platform to improve the above-mentioned problems.

[0007] The specific application is as follows:

[0008] The following units are included:

[0009] User portrait construction unit, which is used to integrate basic user information, historical behavior, and social media activity data to build a comprehensive user portrait;

[0010] Dynamic itinerary planning unit, which dynamically plans the optimal itinerary based on the user's current location, weather, traffic conditions and user preferences;

[0011] Personalized recommendation unit, which provides users with customized travel recommendations based on user portraits and real-time context;

[0012] The voice interaction assistant unit supports users to directly request recommendations and services through voice, realizing intelligent interaction with the platform;

[0013] Real-time tourism information unit, providing real-time information on destination weather, transportation, and opening of scenic spots;

[0014] The AI ​​emotional resonance travel companion unit uses sentiment analysis technology to enable the voice interaction assistant to perceive user emotions and provide corresponding comfort, encouragement and relaxation suggestions.

[0015] Smart travel photography unit, with integrated AI photography guidance function, providing users with shooting location, angle and lighting suggestions;

[0016] The AI ​​emotional resonance travel companion unit specifically includes the following modules:

[0017] Sentiment analysis module: deep learning models process sequence data and capture semantic information; divide the labeled sentiment data into training set, validation set and test set; use the training set data to train the model and update the model parameters through the back propagation algorithm; evaluate the performance of the model on the validation set; and adjust the model parameters, optimize the algorithm and try different model structures based on the evaluation results;

[0018] Dialogue system module: uses natural language understanding technology to identify the user's input intention; records the context information of the dialogue so that it can be referenced and updated in subsequent dialogues; generates responses based on the user's intention and dialogue status using natural language generation technology;

[0019] Emotional resonance module: Identify the user's emotional state through the sentiment analysis model; generate corresponding resonance responses based on the user's emotional state; guide users to express emotions and share experiences through dialogue.

[0020] As a preferred technical solution of the present application, in the dynamic itinerary planning unit, the D algorithm is selected; its data structure is defined as follows:

[0021] OPEN_LIST: A priority queue used to store nodes to be expanded, sorted by the k value of the node;

[0022] CLOSED_LIST: A collection used to store nodes that have been expanded to avoid repeated calculations;

[0023] g(X): the actual path cost from the starting point to node X;

[0024] h(X): heuristic estimated cost from node X to the end point;

[0025] k(X): the minimum h(X) value after node X is added to OPEN_LIST;

[0026] parent(X): records the predecessor node of node X, used for path backtracking.

[0027] As a preferred technical solution of this application, the user portrait construction unit specifically includes the following units:

[0028] Tag classification module: divides user data into different tag categories according to business needs and data characteristics;

[0029] Tag definition and calculation module: define clear meaning and calculation rules for each tag;

[0030] Label system optimization module: add new label categories, refine the granularity of existing labels, and adjust label calculation rules;

[0031] User portrait modeling module: extract useful features from the integrated data, perform feature selection and feature scaling;

[0032] Model selection and training module: select K-means clustering model, logistic regression classification model and linear regression model for training according to business needs and data characteristics;

[0033] Model evaluation and optimization module: Evaluate the model through cross-validation, confusion matrix, and ROC curve methods, and optimize the model based on the evaluation results.

[0034] As the preferred technical solution of this application, the specific formula of the K-means clustering model algorithm is:

[0035] J=∑ i＝1 k ∑ x∈Ci ||x-μ i || 2

[0036] where k is the number of clusters, C i is the i-th cluster, x is the cluster C i Point in, μ i It is cluster C i the center of

[0037] The specific formula of the logistic regression classification algorithm is:

[0038] J(θ)=(-1 / m)∑ i＝1 m [y i log(h θ (x i ))+(1-y i )log(1-h θ (x i ))]

[0039] Where m is the number of samples, y i is the true label of the i-th sample, h θ (x i )=σ(θ T xi ) is the model's predicted probability for the i-th sample;

[0040] The specific formula of the linear regression algorithm is:

[0041] J(θ)=(1 / 2m)∑ i＝1 m (h θ (x i )-y i ) 2

[0042] Where m is the number of samples, h θ (x i )=θ T x i is the model's predicted value for the i-th sample, y i is the true value of the i-th sample.

[0043] As a preferred technical solution of this application, the voice interaction assistant unit specifically includes the following modules:

[0044] Acoustic model matching module: The extracted feature vector is input into the pre-trained acoustic model, and the feature vector is mapped to the corresponding phonemes and morphemes through the deep neural network method. The specific algorithm formula of the deep neural network method is:

[0045] h l =f l (W l h l-1 +b l )

[0046] where h l is the hidden state of layer l, f l is a nonlinear activation function, W l and b l are the weights and biases of the lth layer;

[0047] Language model decoding module: Based on the output of the acoustic model and combined with the pre-trained language model, the most likely text sequence is found through the Viterbi algorithm and beam search algorithm decoding strategy.

[0048] As a preferred technical solution of this application, the personalized recommendation unit specifically includes the following modules:

[0049] Model building and training module: Use collaborative filtering algorithm to build a model and train it; the specific algorithm formula of the collaborative filtering algorithm can be expressed as:

[0050] User-to-user collaborative filtering:

[0051] textsimilarity(u,v)=∑ i∈Iuv (R ui -Rˉ u )(R vi -Rˉ v ) / (√∑ i∈Iuv (R ui -Rˉ u ) 2 √∑

[0052] i∈Iuv (R vi -Rˉ v ) 2 )

[0053] Among them, I uv is the set of products that user u and user v have evaluated together, R ui is the rating of user u on product i, Rˉ u is the average rating of user u;

[0054] Collaborative filtering between items:

[0055] Algorithm formula (taking cosine similarity as an example to calculate item similarity):

[0056] textsimilarity(i,j)=∑ u∈Uij R ui R uj / (√∑ u∈Uij R ui 2 √∑ u∈Uij R uj 2 )

[0057] Among them, U ij is the set of users who have evaluated both product i and product j, R ui is the rating of product i by user u.

[0058] As the preferred technical solution of the present application, the personalized recommendation unit specifically also includes a model training module: dividing the collected data into a training set, a validation set and a test set; adjusting the hyperparameters of the model through cross-validation method, learning rate, number of iterations, and regularization terms, using the training set data to train the recommendation model, and updating the model parameters through the gradient descent optimization algorithm.

[0059] As the preferred technical solution of the present application, the real-time tourism information unit: extracts entities from structured data, semi-structured data and unstructured data; identifies the relationships between entities and constructs triples; extracts attribute information of entities; solves the problem of inconsistent entity names in data from different sources, and merges the same and similar entities; merges and integrates the same and similar relationships in data from different sources; and stores and represents the extracted and fused knowledge in the form of a graph structure; performs semantic search in the knowledge graph according to the user's query request to find relevant entities and relationships; uses the rules and relationships in the knowledge graph for reasoning and completion, generates more complete information and displays the query results to the user in a structured manner.

[0060] As a preferred technical solution of the present application, the smart travel photography unit specifically includes the following units:

[0061] Scene recognition module: Convolutional neural network is used as the core model; CNN automatically extracts feature information from images through multi-layer convolution and pooling operations for scene classification; collects a large number of annotated travel photography image data sets to ensure the diversity and representativeness of the data sets; the data sets should include various travel scenes; uses supervised learning methods to train the CNN model; inputs the annotated image data into the model, and adjusts the model parameters through the back propagation algorithm so that the model can accurately identify different scenes;

[0062] Intelligent composition module: Use semantic segmentation technology to divide the image into different areas; combine photographic aesthetic principles to evaluate the current composition; based on the composition evaluation results, guide the user to adjust the shooting angle and distance parameters through visual feedback and voice prompts.

[0063] As the preferred technical solution of this application, it also includes a data collection and processing unit: collecting travel photography works and related information shared by users through social media, travel forums, and blog platforms; subscribing to and purchasing professional databases related to travel photography; establishing a data sharing mechanism with travel agencies, hotels, and airline partners to obtain their customers' travel photography data; using sensors built into smartphones and cameras to collect environmental data during shooting; deleting duplicate data records to ensure the uniqueness of the data, filling in missing data fields based on contextual information and deleting the entire record, and identifying and processing outliers.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] In the scheme of this application:

[0066] 1. In order to solve the problem that traditional tourism service platforms in the prior art lack personalized and intelligent elements, this application realizes a comprehensive understanding of users' preferences and needs through the user portrait construction unit, thereby providing more personalized services; the personalized recommendation unit can recommend suitable travel routes, attractions, restaurants, etc. for users based on their preferences and real-time situations; at the same time, the voice interaction assistant unit enables users to directly interact with the platform through voice, which improves the convenience and fun of use;

[0067] 2. Through the dynamic itinerary planning unit, the optimal itinerary can be dynamically planned according to the user's current location, weather, traffic conditions and preferences. When the real-time traffic information changes, the platform can re-run the algorithm and dynamically adjust the optimal path to ensure the smooth progress of the itinerary, solving the problems of flexibility and accuracy of itinerary planning in the existing technology;

[0068] 3. The intelligent travel photography unit is equipped with an integrated AI photography guidance function to provide users with suggestions on shooting position, angle, and light; this enables even photography novices to easily take high-quality photos, solving the problem of low travel fun and commemorative value in existing technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A block diagram of the AI-based tourism intelligent interactive platform provided for this application;

[0070] Figure 2 A block diagram of the AI ​​emotional resonance travel companion unit of the AI-based tourism intelligent interactive platform provided for this application;

[0071] Figure 3 A block diagram of a user portrait construction unit for the AI-based tourism intelligent interactive platform provided in this application;

[0072] Figure 4 A block diagram of the voice interaction assistant unit of the AI-based tourism intelligent interactive platform provided in this application;

[0073] Figure 5 A block diagram of the personalized recommendation unit of the AI-based tourism intelligent interactive platform provided in this application;

[0074] Figure 6 A block diagram of the intelligent travel photography unit of the AI-based tourism intelligent interactive platform provided in this application. DETAILED DESCRIPTION

[0075] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0076] As mentioned in the background art, traditional tourism service platforms can often only provide basic itinerary planning, scenic spot introduction and other services, lacking personalized and intelligent elements.

[0077] In order to solve this technical problem, the present invention provides an AI-based tourism intelligent interactive platform, which is applied to all-round and intelligent tourism services.

[0078] Specifically, please refer to Figure 1-Figure 6 The AI-based tourism intelligent interactive platform specifically includes the following units:

[0079] User portrait construction unit, which is used to integrate basic user information, historical behavior, and social media activity data to build a comprehensive user portrait;

[0080] Dynamic itinerary planning unit, which dynamically plans the optimal itinerary based on the user's current location, weather, traffic conditions and user preferences;

[0081] Personalized recommendation unit, which provides users with customized travel recommendations based on user portraits and real-time context;

[0082] The voice interaction assistant unit supports users to directly request recommendations and services through voice, realizing intelligent interaction with the platform;

[0083] Real-time tourism information unit, providing real-time information on destination weather, transportation, and opening of scenic spots;

[0084] The AI ​​emotional resonance travel companion unit uses sentiment analysis technology to enable the voice interaction assistant to perceive user emotions and provide corresponding comfort, encouragement and relaxation suggestions.

[0085] Smart travel photography unit, with integrated AI photography guidance function, providing users with shooting location, angle and lighting suggestions;

[0086] The AI ​​emotional resonance travel companion unit specifically includes the following modules:

[0087] Sentiment analysis module: deep learning models process sequence data and capture semantic information; divide the labeled sentiment data into training set, validation set and test set; use the training set data to train the model and update the model parameters through the back propagation algorithm; evaluate the performance of the model on the validation set; and adjust the model parameters, optimize the algorithm and try different model structures based on the evaluation results;

[0088] Dialogue system module: Use natural language understanding technology to identify the user's input intention; record the context information of the dialogue so that it can be referenced and updated in subsequent dialogues; use natural language generation technology to generate responses based on the user's intention and dialogue status; for common dialogue scenarios, response templates can be pre-defined and filled in according to actual conditions; for more complex dialogue scenarios, the GPT generation model can be used to generate responses; generate more personalized responses based on the user's portrait and current emotional state;

[0089] Emotional resonance module: Identify the user's emotional state through the sentiment analysis model; generate corresponding resonance responses based on the user's emotional state; guide users to express emotions and share experiences through dialogue, and provide psychological counseling and suggestions when necessary.

[0090] The AI-based tourism intelligent interactive platform provided by the present invention provides users with all-round and intelligent tourism services; this not only improves the flexibility and accuracy of user experience and itinerary planning, but also provides functions such as an emotionally resonant travel companion and intelligent travel photography guidance; at the same time, by optimizing data management and utilization, the platform can also continuously optimize and improve itself to better meet user needs.

[0091] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0092] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.

[0093] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0094] Example 1, please refer to Figure 1 , an AI-based tourism intelligent interactive platform, which selects the D algorithm in the dynamic itinerary planning unit; its data structure is defined as follows:

[0095] OPEN_LIST: A priority queue used to store nodes to be expanded, sorted by the k value of the node;

[0096] CLOSED_LIST: A collection used to store nodes that have been expanded to avoid repeated calculations;

[0097] g(X): the actual path cost from the starting point to node X;

[0098] h(X): heuristic estimated cost from node X to the end point;

[0099] k(X): the minimum h(X) value after node X is added to OPEN_LIST;

[0100] parent(X): records the predecessor node of node X, used for path backtracking;

[0101] Core logic:

[0102] During initialization, add the starting point to OPEN_LIST, set its g(start) to 0, and h(start) to the heuristic estimated cost;

[0103] Loop through the following steps until the end is found or OPEN_LIST is empty:

[0104] Take the node X with the smallest k value from OPEN_LIST;

[0105] If X is the end point, the path planning is successful, and the optimal path is obtained by backtracking through the parent pointer;

[0106] Otherwise, expand the adjacent nodes of node X, calculate the g value to reach each adjacent node, and update the k value and parent pointer;

[0107] If an adjacent node is not in OPEN_LIST and CLOSED_LIST, add it to OPEN_LIST;

[0108] If an adjacent node is already in OPEN_LIST, but the g value of reaching the node through X is smaller, update its g value, k value and parent pointer;

[0109] If an adjacent node is already in CLOSED_LIST and the g value of reaching the node through X is not smaller, skip the node;

[0110] In dynamic trip planning, when real-time traffic information changes, the g and h values ​​of the affected nodes are recalculated, and the OPEN_LIST and CLOSED_LIST are updated;

[0111] The result is:

[0112] According to the output of the D* algorithm, the optimal path is obtained by backtracking through the parent pointer;

[0113] Calculate the total cost of the optimal path, including time and cost;

[0114] Taking into account user preferences and destination constraints, fine-tune the optimal path;

[0115] Results:

[0116] Display the itinerary planning results to the user through the user interface (such as mobile APP, web page);

[0117] Provide detailed route information, including the distance, estimated travel time, and cost of each route segment;

[0118] Provide interactive functions to allow users to make adjustments as needed, such as adding waypoints and modifying departure times;

[0119] Real-time updates:

[0120] Continuously collect real-time traffic information during the user's journey;

[0121] When real-time traffic information changes, the D* algorithm is re-run to dynamically adjust the optimal path;

[0122] User feedback:

[0123] Provide a user feedback mechanism to collect users' evaluation and suggestions on itinerary planning results;

[0124] Continuously optimize algorithms and module performance based on user feedback to improve itinerary planning accuracy and user experience.

[0125] The dynamic itinerary planning unit can dynamically plan the optimal itinerary based on the user's current location, weather, traffic conditions and preferences; when real-time traffic information changes, the platform can re-run the algorithm and dynamically adjust the optimal path to ensure the smooth progress of the itinerary; this greatly improves the flexibility and accuracy of itinerary planning.

[0126] Please refer to Figure 3 , an AI-based tourism intelligent interactive platform, whose user portrait construction unit specifically includes the following units:

[0127] Tag classification module: divides user data into different tag categories according to business needs and data characteristics; common tag categories include basic information tags (such as gender, age, region), behavior feature tags (such as browsing time, purchase frequency, interactive behavior), and preference feature tags (such as product preference, brand preference, content preference);

[0128] Tag definition and calculation module: Define clear meanings and calculation rules for each tag; for example, basic information tags can be directly extracted from user registration information; behavioral feature tags can be obtained by counting users' browsing and purchasing behavior data; preference feature tags may need to be mined and predicted through machine learning algorithms;

[0129] Label system optimization module: With the development of business and the accumulation of data, the label system needs to be continuously optimized and expanded; this includes adding new label categories, refining the granularity of existing labels, and adjusting the label calculation rules;

[0130] User portrait modeling module: extract useful features from the integrated data, perform feature selection and feature scaling; feature selection aims to select features that have an important impact on modeling and reduce model complexity; feature scaling helps speed up model training and improve model performance;

[0131] Model selection and training module: K-means clustering model, logistic regression classification model and linear regression model are selected for training according to business needs and data characteristics; during model training, it is necessary to consider the tuning of hyperparameters and the handling of model overfitting and underfitting;

[0132] Model evaluation and optimization module: Evaluate the model through cross-validation, confusion matrix, and ROC curve methods, and optimize the model based on the evaluation results; optimization measures may include adjusting model parameters, increasing the amount of data, and improving feature engineering;

[0133] User portrait generation and update module: Generate user portraits based on the trained model and user feature data; user portraits can be presented in various forms such as tables and graphs, including the user's basic information, behavioral characteristics, and preference characteristics;

[0134] Portrait update module: As user data continues to increase and change, user portraits need to be updated regularly. During updates, incremental learning methods can be used to process and model only newly added or changed data to improve efficiency and accuracy. At the same time, the label system also needs to be reviewed and updated regularly to ensure the timeliness and accuracy of the portraits.

[0135] Through the user portrait construction unit, the platform can fully understand the user's preferences and needs, so as to provide more personalized services; the personalized recommendation unit can recommend suitable travel routes, attractions, restaurants, etc. based on the user's preferences and real-time context; at the same time, the voice interaction assistant unit allows users to interact directly with the platform through voice, which improves the convenience and fun of use.

[0136] Further, such as Figure 3 As shown, the specific formula of the K-means clustering model algorithm is:

[0137] J=∑ i＝1 k ∑ x∈Ci ||x-μ i || 2

[0138] where k is the number of clusters, C i is the i-th cluster, x is the cluster C i Point in, μ i It is cluster C i the center of

[0139] The specific formula of the logistic regression classification algorithm is:

[0140] J(θ)=(-1 / m)∑ i＝1 m [y i log(h θ (x i ))+(1-y i )log(1-h θ (x i ))]

[0141] Where m is the number of samples, y i is the true label of the i-th sample, h θ (x i )=σ(θ T x i ) is the model's predicted probability for the i-th sample (using the sigmoid function to map the linear regression result to the [0,1] interval);

[0142] The specific formula of the linear regression algorithm is:

[0143] J(θ)=(1 / 2m)∑ i＝1 m (h θ (x i )-y i ) 2

[0144] Where m is the number of samples, h θ (x i )=θ T x i is the model's predicted value for the i-th sample, y i is the true value of the i-th sample;

[0145] In the process of building user portraits, it is sometimes necessary to calculate the similarity between different user portraits; commonly used similarity calculation methods include cosine similarity and Pearson similarity; taking cosine similarity as an example, its formula is:

[0146] textsimilarity(A,B)=(A·B) / (||A||||B||)

[0147] Among them, A and B represent the feature vectors of two user portraits respectively, · represents the dot product of the vectors, ||A|| and ||B|| represent the modulus lengths of vectors A and B respectively; the value range of cosine similarity is between [-1,1], and the closer the value is to 1, the more similar the two user portraits are.

[0148] Embodiment 2 further optimizes the AI-based tourism intelligent interactive platform provided in Embodiment 1 or 2. Specifically, Figure 4 As shown, the voice interaction assistant unit specifically includes the following modules:

[0149] Acoustic model matching module: The extracted feature vector is input into the pre-trained acoustic model, and the feature vector is mapped to the corresponding phonemes and morphemes through the deep neural network method. The specific algorithm formula of the deep neural network method is:

[0150] h l =f l (W l h l-1 +b l )

[0151] where h l is the hidden state of layer l, f l is a nonlinear activation function, W l and b l are the weights and biases of the lth layer;

[0152] Language model decoding module: Based on the output of the acoustic model and combined with the pre-trained language model, the most likely text sequence is found through the Viterbi algorithm and beam search algorithm decoding strategy.

[0153] Further, such as Figure 5 As shown, the personalized recommendation unit specifically includes the following modules:

[0154] Model building and training module: Use collaborative filtering algorithm to build a model and train it; the specific algorithm formula of collaborative filtering algorithm can be expressed as:

[0155] User-to-user collaborative filtering:

[0156] textsimilarity(u,v)=∑ i∈Iuv (R ui -Rˉ u )(R vi -Rˉ v ) / (√∑ i∈Iuv (R ui-Rˉ u ) 2 √∑

[0157] i∈Iuv (R vi -Rˉ v ) 2 )

[0158] Among them, I uv is the set of products that user u and user v have evaluated together, R ui is the rating of user u on product i, Rˉ u is the average rating of user u;

[0159] Collaborative filtering between items:

[0160] Algorithm formula (taking cosine similarity as an example to calculate item similarity):

[0161] textsimilarity(i,j)=∑ u∈Uij R ui R uj / (√∑ u∈Uij R ui 2 √∑ u∈Uij R uj 2 )

[0162] Among them, U ij is the set of users who have evaluated both product i and product j, R ui is the rating of product i by user u.

[0163] Further, such as Figure 5 As shown, the personalized recommendation unit specifically also includes a model training module: the collected data is divided into a training set, a validation set, and a test set; the hyperparameters of the model are adjusted through the cross-validation method, learning rate, number of iterations, and regularization terms, the recommendation model is trained using the training set data, and the model parameters are updated through the gradient descent optimization algorithm.

[0164] Embodiment 3: The AI-based tourism intelligent interactive platform provided in the above embodiments is further optimized, such as Figure 1As shown in the figure, the real-time tourism information unit: extracts entities from structured data, semi-structured data and unstructured data; identifies the relationship between entities and constructs triples; extracts attribute information of entities; solves the problem of inconsistent entity names in data from different sources, and merges the same and similar entities; merges and integrates the same and similar relationships in data from different sources; and stores and represents the extracted and fused knowledge in the form of a graph structure; performs semantic search in the knowledge graph according to the user's query request to find relevant entities and relationships; uses the rules and relationships in the knowledge graph to reason and complete, generates more complete information and displays the query results to the user in a structured manner.

[0165] Further, such as Figure 6 As shown, the smart travel photography unit specifically includes the following units:

[0166] Scene recognition module: Convolutional neural network is used as the core model; CNN automatically extracts feature information from images through multi-layer convolution and pooling operations for scene classification; collects a large number of annotated travel photography image data sets to ensure the diversity and representativeness of the data sets; the data sets should contain various travel scenes; uses supervised learning methods to train the CNN model; inputs the annotated image data into the model, and adjusts the model parameters through the back propagation algorithm so that the model can accurately identify different scenes; the formula is as follows: For a certain layer in CNN, assuming that the input feature map is X, the convolution kernel is W, the bias is b, and the activation function is ReLU, the output feature map Y can be expressed as:

[0167] Y = ReLU(W*X+b)

[0168] In the entire CNN model, a deep neural network structure is formed by stacking multiple convolutional layers, pooling layers, and fully connected layers. The final output layer uses the softmax function to classify scenes and output the probability distribution of each scene.

[0169] Intelligent composition module: Use semantic segmentation technology to divide the image into different areas; combine the principles of photographic aesthetics to evaluate the current composition; according to the composition evaluation results, guide the user to adjust the shooting angle and distance parameters through visual feedback and voice prompts; image segmentation technology can use a deep learning method based on a fully convolutional network; the specific application of aesthetic principles needs to be adjusted according to the actual scene and user needs; for example, when shooting landscapes, more attention may be paid to the balance and layering of the picture; when shooting people, more attention may be paid to the prominence of the person and the blurring of the background;

[0170] Parameter optimization module: pre-set a set of optimized camera parameter setting schemes for different scenes; these schemes can be obtained through experiments and expert experience; during the shooting process, adjust the camera parameters in real time according to the scene recognition results; for example, automatically increase the exposure time and ISO value in dark scenes; automatically increase the shutter speed when shooting moving objects; camera parameters include but are not limited to exposure time, aperture size, ISO value, shutter speed; the parameter optimization process needs to consider the balance between multiple factors, such as the trade-off between exposure time and ISO value;

[0171] Real-time image enhancement module: Use denoising, sharpening, and color enhancement image processing algorithms to process the captured images in real time; use GPU hardware acceleration equipment to increase the image processing speed and ensure real-time performance; the denoising algorithm can use a method based on deep learning, such as DnCNN; the sharpening algorithm can use a method based on edge detection; the color enhancement algorithm can use a method based on histogram equalization and a color correction network based on deep learning; the real-time image enhancement process needs to consider the balance between computational complexity and real-time performance;

[0172] Personalized recommendation module: Build user portraits by analyzing the user's shooting history and preference information; provide personalized shooting suggestions, filter effects and post-processing solutions based on the current scene information and user portraits; user portrait construction can be based on collaborative filtering, content filtering and hybrid recommendation methods; the personalized recommendation process needs to consider the matching relationship between the user's personalized needs and the actual situation of the current scene.

[0173] The smart travel photography unit integrates AI photography guidance function to provide users with shooting position, angle and lighting suggestions; this enables even photography novices to easily take high-quality photos, enhancing the fun and commemorative value of travel.

[0174] Further, such as Figure 1 As shown, it also includes a data collection and processing unit: collecting travel photography works and related information shared by users through social media, travel forums, and blog platforms; subscribing to and purchasing professional databases related to travel photography; establishing a data sharing mechanism with travel agencies, hotels, and airline partners to obtain their customers' travel photography data; using sensors built into smartphones and cameras to collect environmental data during shooting; deleting duplicate data records to ensure the uniqueness of the data, filling in missing data fields based on contextual information and deleting the entire record, and identifying and processing outliers.

[0175] The data collection and processing unit collects and analyzes user data, which not only provides a rich data source for the platform, but also provides a basis for subsequent optimization and improvement. Through continuous data accumulation and analysis, the platform can continuously optimize algorithm and module performance and improve service accuracy and efficiency.

[0176] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection, an electrical connection, or communication with each other; it can be a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0177] Obviously, the embodiments described above are only some embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are given in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to perform equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of the present invention, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. An AI-based tourism intelligent interactive platform, characterized in that: The following units are included: User portrait construction unit, which is used to integrate basic user information, historical behavior, and social media activity data to build a comprehensive user portrait; Dynamic itinerary planning unit, which dynamically plans the optimal itinerary based on the user's current location, weather, traffic conditions and user preferences; Personalized recommendation unit, which provides users with customized travel recommendations based on user portraits and real-time context; The voice interaction assistant unit supports users to directly request recommendations and services through voice, realizing intelligent interaction with the platform; Real-time tourism information unit, providing real-time information on destination weather, transportation, and opening of scenic spots; The AI ​​emotional resonance travel companion unit uses sentiment analysis technology to enable the voice interaction assistant to perceive user emotions and provide corresponding comfort, encouragement and relaxation suggestions. Smart travel photography unit, with integrated AI photography guidance function, providing users with shooting location, angle and lighting suggestions; The AI ​​emotional resonance travel companion unit specifically includes the following modules: Sentiment analysis module: deep learning models process sequence data and capture semantic information; divide the labeled sentiment data into training set, validation set and test set; use the training set data to train the model and update the model parameters through the back propagation algorithm; evaluate the performance of the model on the validation set; and adjust the model parameters, optimize the algorithm and try different model structures based on the evaluation results; Dialogue system module: uses natural language understanding technology to identify the user's input intention; Record the context of the conversation so that it can be referenced and updated in subsequent conversations; generate responses using natural language generation technology based on the user's intent and the state of the conversation; Emotional resonance module: Identify the user's emotional state through the sentiment analysis model; generate corresponding resonance responses based on the user's emotional state; Guide users to express emotions and share experiences through dialogue.

2. According to claim 1, the AI-based tourism intelligent interactive platform is characterized in that: In the dynamic itinerary planning unit, the D algorithm is selected; its data structure is defined as follows: OPEN_LIST: A priority queue used to store nodes to be expanded, sorted by the k value of the node; CLOSED_LIST: A collection used to store nodes that have been expanded to avoid repeated calculations; g(X): the actual path cost from the starting point to node X; h(X): heuristic estimated cost from node X to the end point; k(X): the minimum h(X) value after node X is added to OPEN_LIST; parent(X): records the predecessor node of node X, used for path backtracking.

3. According to claim 2, the AI-based tourism intelligent interactive platform is characterized in that: The user portrait construction unit specifically includes the following units: Tag classification module: divides user data into different tag categories according to business needs and data characteristics; Tag definition and calculation module: define clear meaning and calculation rules for each tag; Label system optimization module: add new label categories, refine the granularity of existing labels, and adjust label calculation rules; User portrait modeling module: extract useful features from the integrated data, perform feature selection and feature scaling; Model selection and training module: select K-means clustering model, logistic regression classification model and linear regression model for training according to business needs and data characteristics; Model evaluation and optimization module: Evaluate the model through cross-validation, confusion matrix, and ROC curve methods, and optimize the model based on the evaluation results.

4. The AI-based tourism intelligent interactive platform according to claim 3 is characterized in that: The specific formula of the K-means clustering model algorithm is: J=∑ i＝1 k ∑ x∈Ci ||x-m i || 2 where k is the number of clusters, C i is the i-th cluster, x is the cluster C i Point in, μ i It is cluster C i the center of; The specific formula of the logistic regression classification algorithm is: J(θ)=(-1 / m)∑ i＝1 m [y i log(h θ (x i ))+(1-y i )log(1-h θ (x i ))] Where m is the number of samples, y i is the true label of the i-th sample, h θ (x i )=σ(θ T x i ) is the predicted probability of the model for the i-th sample; The specific formula of the linear regression algorithm is: J(θ)=(1 / 2m)∑ i＝1 m (h θ (x i )-y i ) 2 Where m is the number of samples, h θ (x i )=θ T x i is the model's predicted value for the i-th sample, y i is the true value of the i-th sample.

5. The AI-based tourism intelligent interactive platform according to claim 4 is characterized in that: The voice interaction assistant unit specifically includes the following modules: Acoustic model matching module: The extracted feature vector is input into the pre-trained acoustic model, and the feature vector is mapped to the corresponding phonemes and morphemes through the deep neural network method. The specific algorithm formula of the deep neural network method is: h l =f l (W l h l-1 +b l ) where h l is the hidden state of layer l, f l is a nonlinear activation function, W l and b l are the weights and biases of the lth layer; Language model decoding module: Based on the output of the acoustic model and combined with the pre-trained language model, the most likely text sequence is found through the Viterbi algorithm and beam search algorithm decoding strategy.

6. The AI-based tourism intelligent interactive platform according to claim 5, characterized in that: The personalized recommendation unit specifically includes the following modules: Model building and training module: build a model using collaborative filtering algorithm and train it; the specific algorithm formula of the collaborative filtering algorithm can be expressed as: User-to-user collaborative filtering: textsimilarity(u,v)=∑ i∈Iuv (R ui -Rˉ u )(R vi -Rˉ v ) / (√∑ i∈Iuv (R ui -Rˉ u ) 2 √∑ i∈Iuv (R vi -Rˉ v ) 2 ) Among them, I uv is the set of products that user u and user v have evaluated together, R ui is the rating of user u on product i, Rˉ u is the average rating of user u; Collaborative filtering between items: Algorithm formula (taking cosine similarity as an example to calculate item similarity): textsimilarity(i,j)=∑ u∈Uij R ui R uj / (√∑ u∈Uij R ui 2 √∑ u∈Uij R uj 2 ) Among them, U ij is the set of users who have evaluated both product i and product j, R ui is the rating of product i by user u.

7. The AI-based tourism intelligent interactive platform according to claim 6, characterized in that: The personalized recommendation unit specifically also includes a model training module: dividing the collected data into a training set, a validation set, and a test set; adjusting the hyperparameters of the model through cross-validation method, learning rate, number of iterations, and regularization terms, using the training set data to train the recommendation model, and updating the model parameters through a gradient descent optimization algorithm.

8. The AI-based tourism intelligent interactive platform according to claim 7, characterized in that: The real-time tourism information unit: extracts entities from structured data, semi-structured data and unstructured data; identifies the relationship between entities and constructs triples; extracts attribute information of entities; Solve the problem of inconsistent entity names in data from different sources, merge the same and similar entities; merge and integrate the same and similar relationships in data from different sources; The extracted and fused knowledge is stored and represented in the form of a graph structure; Perform semantic search in the knowledge graph based on the user's query request to find relevant entities and relationships; use the rules and relationships in the knowledge graph to perform reasoning and completion, generate more complete information, and present the query results to the user in a structured manner.

9. The AI-based tourism intelligent interactive platform according to claim 8, characterized in that: The smart travel photography unit specifically includes the following units: Scene recognition module: Convolutional neural network is used as the core model; CNN automatically extracts feature information from images through multi-layer convolution and pooling operations for scene classification; collects a large number of annotated travel photography image data sets to ensure the diversity and representativeness of the data sets; the data sets should include various travel scenes; uses supervised learning methods to train the CNN model; inputs the annotated image data into the model, and adjusts the model parameters through the back propagation algorithm so that the model can accurately identify different scenes; Intelligent composition module: Use semantic segmentation technology to divide the image into different areas; combine photographic aesthetic principles to evaluate the current composition; based on the composition evaluation results, guide the user to adjust the shooting angle and distance parameters through visual feedback and voice prompts.

10. The AI-based tourism intelligent interactive platform according to claim 9, characterized in that: It also includes a data collection and processing unit: collecting travel photography works and related information shared by users through social media, travel forums, and blog platforms; subscribing to and purchasing professional databases related to travel photography; establishing a data sharing mechanism with travel agencies, hotels, and airline partners to obtain their customers' travel photography data; using sensors built into smartphones and cameras to collect environmental data during shooting; deleting duplicate data records to ensure the uniqueness of the data, filling in missing data fields based on contextual information and deleting the entire record, and identifying and processing outliers.

Images