Intelligent travel system based on python deep learning

Abstract

The invention discloses an intelligent travel text system based on python deep learning, and relates to the technical field of python deep learning, and the system comprises the following components: a data collection module which is used for automatically collecting comment data about a specific travel destination from various channels of various travel websites, social media platforms and online travel forums, according to the method, massive comments and evaluation data about a specific tourism destination can be automatically collected from multiple channels, the accuracy and consistency of analysis data are ensured through fine data preprocessing, fine adjustment is performed by applying an advanced deep learning sentiment analysis algorithm in combination with a specific corpus in the tourism field, and the accuracy and consistency of data analysis are ensured. The system can accurately judge the emotional tendency of tourists and deeply mine key opinions and suggestions in comments, and the deep emotional mining ability not only helps tourism management departments and scenic spots to quickly understand real feedback of the tourists, but also can accurately identify satisfaction distribution and main concerns of the tourists on services.

Description

Technical Field

[0001] The present invention relates to the field of Python deep learning technology, and specifically to a smart cultural tourism system based on Python deep learning. Background Art

[0002] With the popularization of the Internet and the development of social media, the tourism industry is undergoing unprecedented changes. During their travels, tourists share their experiences, comments and suggestions through online platforms, forming a huge amount of tourism public opinion data. This data not only reflects tourists' feelings and opinions on tourist destinations, services, facilities and activities, but also contains huge market potential and room for improvement. Therefore, how to efficiently collect, process and deeply analyze these public opinion data to guide decision-making and service optimization in the tourism industry has become a problem that needs to be solved urgently.

[0003] Although traditional technologies have supported the collection and basic analysis of tourism public opinion to a certain extent, they have obvious shortcomings in processing large amounts of data, complex emotional expressions, and in-depth exploration of tourist needs. First, traditional methods often rely on manual screening and simple statistics, which are difficult to cope with massive and rapidly growing public opinion data. Secondly, for the judgment of emotional tendencies, traditional technologies mostly use keyword matching and simple rules, which are difficult to accurately capture the subtle differences in emotions in complex contexts, resulting in deviations in analysis results. In addition, traditional systems lack an intelligent mechanism for generating improvement suggestions and are unable to provide personalized service optimization solutions based on analysis results, limiting the application effect of public opinion data in improving service quality.

[0004] Therefore, it is necessary to develop a smart cultural tourism system based on Python deep learning to solve the above problems. Summary of the invention

[0005] The purpose of this invention is to make up for the shortcomings of the existing technology and provide a smart cultural tourism system based on python deep learning. It can realize intelligent processing and analysis of tourist destination reviews through data collection, preprocessing, sentiment analysis, result display and feedback improvement modules, and can accurately judge the emotional tendencies of tourists, extract key opinions and suggestions, and display them to management departments in the form of intuitive charts, helping decision makers to quickly grasp tourist needs and market trends.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a smart cultural tourism system based on Python deep learning, which includes the following components: Data collection module: Automatically collects review data about specific tourist destinations from various channels such as major travel websites, social media platforms and online travel forums; Data preprocessing module: performs a series of preprocessing operations on the collected raw data, including data cleaning, denoising, and word segmentation, to generate a clean and standardized data set suitable for subsequent sentiment analysis; Sentiment analysis module: conducts deep sentiment mining on pre-processed data, uses language models combined with tourism-specific corpora for adjustment, achieves accurate judgment of sentiment tendencies, and extracts key opinions and suggestions from reviews, providing a specific basis for improving services; Results display module: The results of sentiment analysis are presented to tourism management departments and scenic spots in the form of intuitive charts and reports. Through visualization, the distribution of tourists’ satisfaction, main concerns and improvement suggestions are clearly displayed; Feedback improvement module: Based on the results of sentiment analysis, through intelligent recommendation algorithms, combined with historical data and current market conditions, it generates personalized improvement plans, tracks the implementation effects of improvement measures, and forms a closed-loop management.

[0007] Furthermore, the data collection module sets the goal of data collection, specifies the specific tourist destinations and their related comment data that need to be collected, designs and develops a customized web crawler program, uses the crawler program to extract comments and evaluation data about the specific tourist destinations from various channels, and stores the extracted data in a local database.

[0008] Furthermore, the data preprocessing module receives the original data transmitted by the data acquisition module, performs preliminary integration and cleaning of the data, removes invalid information, processes missing values, converts it into a unified format, performs word segmentation for Chinese, performs word form restoration for English, and performs data standardization and converts it into a format suitable for the sentiment analysis algorithm.

[0009] Furthermore, the sentiment analysis module loads the preprocessed data set from the data storage system, creates a BERT language model to judge the sentiment tendency, constructs a specific corpus about the tourism field, uses the conditional random field CRF to annotate the part of speech of each word in the corpus, uses the annotated tourism field specific corpus to adjust the BERT language model, inputs the preprocessed data into the adjusted model, predicts the sentiment tendency, and based on the judgment of the sentiment tendency, uses natural language processing technology to extract the key information in the comments, integrates the sentiment tendency judgment results and the key information and suggestions to form a complete sentiment analysis report.

[0010] Furthermore, the sentiment analysis module creates a BERT language model. BERT is a language model based on the Transformer architecture. BERT consists of multiple layers of Transformer encoders. Each Transformer encoder contains a multi-head self-attention mechanism and a feedforward neural network. The calculation of the multi-head self-attention mechanism is expressed as: Softmax , where Q is the query vector used to focus on different parts of the input sequence, K is the key vector that interacts with the query vector to calculate the attention weights, and V is the value vector that is weighted summed according to the attention weights. is the dimension of the key. BERT uses a masked language model. In the masked language model, some words in the input text are randomly masked, and the model needs to predict the masked words. The input of BERT includes word embedding, segment embedding, and position embedding. The input word sequence is x, the word embedding is represented by E, the position embedding is represented by P, and the segment embedding is represented by. The final input is represented by , add a specific output layer on top according to the specific task, and adjust the parameters of the weight matrix W and bias vector b on the new data.

[0011] Furthermore, the natural language processing module uses the conditional random field CRF to label each word in the corpus with its part of speech, and records a sentence about scenic spot information as , where x is a single word in the sentence, and the output sequence is recorded as , where y is the part-of-speech tag of x, and a set of feature functions are defined ,in, is the index of the characteristic function, is the position index of the word in the sentence, is the part-of-speech tag at position i, It's location The part-of-speech tagging is used for each feature function Assign a weight , for a given sentence x and a part-of-speech tag sequence y, use the feature function and weight to calculate the score of the tag sequence: , by normalizing the scores, we get the conditional probability of annotating sequence Y given sentence X: During the training process, the weights of the feature functions are learned using the training data In the testing and application phases, for a given scenic environment sentence X, the scores of all part-of-speech tag sequences are calculated, and the one with the highest score is selected as the tag.

[0012] Furthermore, the sentiment analysis module uses natural language processing technology to extract key information from the comments. After completing the sentiment tendency judgment, the classification results are analyzed. The sentiment tendency is divided into three categories: positive, negative and neutral. Represents, decomposing the review C into a sequence of words , let word The part of speech is ,but Represents the probability of predicting its part of speech for a given word, identifies the named entities in the comments, and sets the named entity set as E. For each word in the comment , to determine whether it belongs to a named entity , using binary variables express, Indicates belonging to, Indicates that it does not belong to, extract keywords by calculating word frequency and inverse document frequency, words Frequency of words Calculated as: ,in, It's a word The number of times it appears in the comment, N is the total number of words in the comment, the word Inverse document frequency Calculated as: , where M is the total number of documents, Contains the word The importance of keywords is expressed by TF-IDF value, and the formula is: .

[0013] Furthermore, the result display module receives complete analysis data from the sentiment analysis system, uses data visualization technology to generate a chart of the distribution of tourist satisfaction, and compiles a report with a clear structure and strict logic based on the analysis results. It provides dynamic interactive functions for charts and reports, allowing users to customize the displayed chart type, color, and font according to their own needs, and upload the generated charts and reports to the online platform for decision makers to view at any time. It provides a user feedback channel, collects feedback from tourism management departments and scenic spots on the result display module, and continuously optimizes the function and performance of the result display module based on user feedback and market demand.

[0014] Furthermore, the feedback improvement module receives and integrates the output results from the sentiment analysis module, uses the data warehouse to collect and analyze historical service data, market trends and external information of competitor dynamics, builds an intelligent recommendation model based on the sentiment analysis results and historical data and the logistic regression model, conducts a multi-dimensional evaluation of the generated improvement plans, converts the evaluated improvement plans into actual action plans, and assigns them to relevant departments and responsible persons for execution. After the improvement measures have been implemented for a period of time, relevant data is collected, the implementation effect is comprehensively evaluated, and the effect evaluation results are fed back to the sentiment analysis module as new input data for continuous optimization of the sentiment analysis model and generation of improvement suggestions.

[0015] Furthermore, the feedback improvement module builds an intelligent recommendation model based on the logistic regression model, collects sentiment analysis result data and related historical data, extracts meaningful features from the original data, and forms a feature set. ,in, The average length of stay of tourists, the number of service personnel, and the score of scenic spot facilities are used to construct a logistic regression model. The basic form of the logistic regression model is: ,in, is a binary variable, is the intercept term, are the regression coefficients to be estimated. These parameters are estimated through maximum likelihood estimation. The model is trained using feature data and regularization parameters are adjusted to prevent overfitting. New data is input into the trained model to identify potential problems in the service. Based on the output of the model, personalized improvement suggestions and solutions are generated for specific potential problems.

[0016] Compared with the existing technology, this intelligent cultural tourism system based on python deep learning has the following beneficial effects: 1. The present invention can automatically collect massive comments and evaluation data about specific tourist destinations from multiple channels, and after sophisticated data preprocessing, ensure the accuracy and consistency of the analyzed data. By using advanced deep learning sentiment analysis algorithms and fine-tuning in combination with specific corpora in the tourism field, the system can accurately judge the emotional tendencies of tourists and deeply explore the key opinions and suggestions in the comments. This deep sentiment mining capability not only helps tourism management departments and scenic spots to quickly understand the real feedback from tourists, but also accurately identifies the distribution of tourists' satisfaction with services and their main concerns, providing strong data support for public opinion monitoring and service quality improvement in the tourism industry.

[0017] 2. The sentiment analysis report generated by the sentiment analysis module of the present invention is presented in the form of intuitive charts and reports, so that decision makers with non-technical backgrounds can also quickly understand tourist needs and market trends. In addition, the system also integrates a feedback improvement module, which can intelligently recommend personalized improvement plans based on sentiment analysis results, combined with historical data and current market conditions, and track the implementation effect of improvement measures to form a closed-loop management. This intelligent public opinion analysis and management mechanism not only promotes the continuous optimization of tourism services, but also significantly improves tourist satisfaction and loyalty, laying a solid foundation for the sustainable development of the tourism industry.

[0018] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 This is a flowchart of a smart cultural tourism system based on Python deep learning. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0022] This embodiment provides a smart cultural tourism system based on Python deep learning. The system integrates five modules: data collection, data preprocessing, sentiment analysis, result display and feedback improvement. Through multi-channel data collection, comprehensive collection of tourism review data is achieved. After fine data preprocessing, a standardized data set is generated. The deep learning sentiment analysis algorithm is used in combination with a specific corpus in the tourism field to accurately judge sentiment tendencies and extract key opinions. The results are displayed in the form of intuitive charts and reports to assist tourism management departments and scenic spots to quickly understand tourists' needs. The feedback improvement module intelligently recommends improvement plans based on the analysis results, and tracks the implementation effect to form a closed-loop management and continuously improve service quality.

[0023] The data collection module sets the data collection goals, obtains rich user comments and rating data from online travel platforms such as Ctrip, Qunar, and Mafengwo, obtains data on travel experiences, pictures, and videos shared by users on these platforms from social media such as Weibo, Xiaohongshu, and Douyin, obtains official information and important channels for tourists to communicate from official websites and forums, designs and develops customized web crawlers, uses crawlers to extract comments and evaluation data about specific tourist destinations from various channels, and stores the extracted data in a local database in a preset format.

[0024] The data preprocessing module receives the original data transmitted by the data acquisition module, deletes records containing advertisements, spam, duplicate comments and invalid data, fills in the missing data according to the actual situation, unifies the format of the comment text, uses the Chinese word segmentation tool to perform word segmentation processing for Chinese comment text, and divides the continuous text into words and phrases. For English comment text, it performs word form restoration processing to restore different forms of words to basic forms to reduce the impact of lexical diversity on sentiment analysis, converts the processed text data into a format suitable for the sentiment analysis algorithm, and stores the processed data in a format suitable for subsequent processing and analysis.

[0025] The sentiment analysis module loads the preprocessed data set from the data storage system and creates a BERT language model to judge the sentiment tendency. The BERT model can capture rich semantic information, collect corpus related to the tourism field from multiple channels, use the conditional random field model to perform part-of-speech tagging on the corpus, organize the annotated corpus into a format suitable for model training, adjust the BERT model using a specific corpus in the tourism field, input the preprocessed comment data into the fine-tuned BERT model, and the model outputs the sentiment tendency prediction results for each comment. Based on the judgment of sentiment tendency, natural language processing technology is used to extract key information from the comments, such as specific attractions, services, problems and suggestions mentioned by tourists, integrate the sentiment tendency prediction results and key information, and generate a complete sentiment analysis report.

[0026] The result display module receives complete analysis data from the sentiment analysis system, including the distribution of tourist satisfaction, a list of major concerns, and improvement suggestions generated for these concerns. It uses data visualization technology to generate a chart of tourist satisfaction distribution, intuitively displaying the satisfaction levels of different service links and the overall tourism experience, and displays the most frequently mentioned issues, suggestions, and expectations in tourist feedback in the form of word clouds, bar charts, and heat maps. Based on the analysis results, it compiles a report with a clear structure and rigorous logic, provides dynamic interactive functions for charts and reports, and allows users to customize the displayed chart type, color, and font according to their own needs. The generated charts and reports are uploaded to the official website, internal management system, and mobile application online platform of the tourism management department and scenic spot, so that decision makers can view them at any time. It provides an offline document download function so that decision makers can obtain analysis results when it is inconvenient to access the Internet. It sets up user feedback channels to collect feedback from tourism management departments and scenic spots on the result display module, understand user needs and room for improvement, and continuously optimize the functions and performance of the result display module based on user feedback and market demand to improve user experience and display effects.

[0027] The feedback improvement module receives the output results from the sentiment analysis module, integrates the sentiment analysis results with the historical service data, market trend reports and competitor dynamic information in the data warehouse to form a comprehensive analysis basis, and extracts meaningful features from the integrated data, such as the average length of stay of tourists, the number of service personnel, and the ratings of scenic spot facilities to form a feature set. Based on the feature set, the logistic regression model is selected as the basic framework. The basic form of the logistic regression model is: ,in, is a binary variable, is the intercept term, is the regression coefficient to be estimated. The model is trained using feature data. Model parameters are estimated through maximum likelihood estimation. Regularization parameters are adjusted to prevent overfitting. New data is input into the trained model. Potential problems in the service are identified. Based on the output of the model, personalized improvement suggestions are generated for specific potential problems. The generated improvement plans are evaluated in multiple dimensions. The evaluated improvement plans are converted into actual action plans and assigned to relevant departments and responsible persons for implementation. The implementation progress of the improvement measures is tracked to ensure that they are carried out as planned. After the improvement measures have been implemented for a period of time, relevant data is collected, and a comprehensive evaluation of the implementation effect is conducted. The effect evaluation results are fed back to the sentiment analysis module as new input data for continuous optimization of the sentiment analysis model and the generation of improvement suggestions. According to the feedback results, the logistic regression model is iteratively optimized to improve the prediction accuracy and the pertinence of the improvement suggestions. Example

[0028] In this embodiment, a smart cultural tourism system based on Python deep learning is used, with particular emphasis on the sentiment analysis module.

[0029] Create a BERT language model to judge sentiment tendency. The BERT model is based on the Transformer architecture and contains multiple layers of encoders. Each encoder has a multi-head self-attention mechanism and a feedforward neural network. The calculation expression of the multi-head self-attention mechanism is: Attention Softmax , where Q is the query vector used to focus on different parts of the input sequence, K is the key vector that interacts with the query vector to calculate the attention weights, and V is the value vector that is weighted summed according to the attention weights. is the dimension of the key. BERT uses a masked language model. In the masked language model, some words in the input text are randomly masked, and the model needs to predict the masked words. The input of BERT includes word embedding, segment embedding, and position embedding. The input word sequence is x, the word embedding is represented by E, the position embedding is represented by P, and the segment embedding is represented by. The final input is represented by , add a specific output layer on top according to the specific task, and adjust the parameters of the weight matrix W and bias vector b on the new data.

[0030] Collect and organize a large amount of text data in the tourism field to form a specific corpus, use conditional random fields (CRF) to perform part-of-speech tagging on each word in the corpus, and record a sentence about scenic spot information as , where x is a single word in the sentence, and the output sequence is recorded as , where y is the part-of-speech tag of x, and a set of feature functions are defined ,in, is the index of the characteristic function is the position index of the word in the sentence, is the part-of-speech tag at position i, It's location The part-of-speech tagging is used for each feature function Assign a weight , for a given sentence x and a part-of-speech tag sequence y, the feature function and weight are used to calculate the score of the tag sequence. The formula is: , by normalizing the scores, we get the conditional probability of annotating sequence Y given sentence X: During the training process, the weights of the feature functions are learned using the training data In the testing and application phases, for a given scenic environment sentence X, the scores of all part-of-speech tagging sequences are calculated, and the sequence with the highest score is selected as the tag.

[0031] The pre-trained BERT model is adjusted using a well-annotated tourism-specific corpus. During the adjustment process, the model parameters are adjusted to adapt to the sentiment analysis task in the tourism field.

[0032] The preprocessed tourism review data is input into the adjusted BERT model, and the model outputs the predicted results of the sentiment tendency of each review, which is usually divided into three categories: positive, negative and neutral. Based on the judgment of sentiment tendency, the classified reviews are further analyzed using natural language processing technology (NLP). Lexical analysis is performed, including word segmentation and part-of-speech tagging, and named entities in the reviews, such as attraction names and service types, are identified. Keywords and words are extracted by calculating the term frequency (TF) and inverse document frequency (IDF). Frequency of words Calculated as: ,in It's a word The number of times it appears in the comment, N is the total number of words in the comment, the word Inverse document frequency Calculated as: , where M is the total number of documents, Contains the word The importance of keywords is expressed by TF-IDF value, and the formula is: , analyze the syntactic dependencies between words in the comments, using a directed graph Represents, where V is a word node and E is a dependency edge. According to the calculation results and combined with the sentiment tendency y, the key information related to the sentiment is screened out, and the screened out key information is integrated and sorted to form the key information extraction result of the comment.

[0033] Integrate the sentiment tendency judgment results and the extracted key information to form a complete sentiment analysis report. The report should include sentiment tendency statistics, a list of key information and sample comment analysis. Provide the sentiment analysis report to the tourism management department to assist them in formulating improvement measures and improving service quality. Collect feedback from the management department on the report and continuously optimize the sentiment analysis module and report generation process.

[0034] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A smart cultural tourism system based on Python deep learning, characterized in that: The system consists of the following components: Data collection module: Automatically collects review data about specific tourist destinations from various channels such as major travel websites, social media platforms and online travel forums; Data preprocessing module: performs a series of preprocessing operations on the collected raw data, including data cleaning, denoising, and word segmentation, to generate a clean and standardized data set suitable for subsequent sentiment analysis; Sentiment analysis module: conducts deep sentiment mining on pre-processed data, uses language models combined with tourism-specific corpora for adjustment, achieves accurate judgment of sentiment tendencies, and extracts key opinions and suggestions from reviews, providing a specific basis for improving services; Results display module: The results of sentiment analysis are presented to tourism management departments and scenic spots in the form of intuitive charts and reports. Through visualization, the distribution of tourists’ satisfaction, main concerns and improvement suggestions are clearly displayed; Feedback improvement module: Based on the results of sentiment analysis, through intelligent recommendation algorithms, combined with historical data and current market conditions, it generates personalized improvement plans, tracks the implementation effects of improvement measures, and forms a closed-loop management.

2. According to the smart cultural tourism system based on Python deep learning in claim 1, it is characterized in that: The data collection module sets the data collection target, specifies the specific tourist destinations and their related comment data that need to be collected, designs and develops a customized web crawler program, uses the crawler program to extract comments and evaluation data about the specific tourist destinations from various channels, and stores the extracted data in a local database.

3. According to the smart cultural tourism system based on Python deep learning in claim 1, it is characterized in that: The data preprocessing module receives the raw data transmitted by the data acquisition module, performs preliminary integration and cleaning of the data, removes invalid information, processes missing values, converts the data into a unified format, performs word segmentation for Chinese, performs word form restoration for English, and performs data standardization and converts the data into a format suitable for the sentiment analysis algorithm.

4. According to the smart cultural tourism system based on Python deep learning in claim 1, it is characterized in that: The sentiment analysis module loads the preprocessed data set from the data storage system, creates a BERT language model to judge the sentiment tendency, constructs a specific corpus about the tourism field, uses the conditional random field CRF to annotate the part of speech of each word in the corpus, uses the annotated tourism field specific corpus to adjust the BERT language model, inputs the preprocessed data into the adjusted model, predicts the sentiment tendency, and based on the judgment of the sentiment tendency, uses natural language processing technology to extract the key information in the comments, integrates the sentiment tendency judgment results and the key information and suggestions, and forms a complete sentiment analysis report.

5. According to claim 4, a smart cultural tourism system based on Python deep learning is characterized in that: The sentiment analysis module creates a BERT language model. BERT is a language model based on the Transformer architecture. BERT consists of multiple layers of Transformer encoders. Each Transformer encoder contains a multi-head self-attention mechanism and a feedforward neural network. The calculation of the multi-head self-attention mechanism is expressed as: Softmax , where Q is the query vector used to focus on different parts of the input sequence, K is the key vector that interacts with the query vector to calculate the attention weights, and V is the value vector that is weighted summed according to the attention weights. is the dimension of the key. BERT uses a masked language model. In the masked language model, some words in the input text are randomly masked, and the model needs to predict the masked words. The input of BERT includes word embedding, segment embedding, and position embedding. The input word sequence is x, the word embedding is represented by E, the position embedding is represented by P, and the segment embedding is represented by. The final input is represented by , add a specific output layer on top according to the specific task, and adjust the parameters of the weight matrix W and bias vector b on the new data.

6. According to claim 4, a smart cultural tourism system based on Python deep learning is characterized in that: The natural language processing module uses the conditional random field CRF to label each word in the corpus with its part of speech. Given a sentence about scenic spot information, it is recorded as , where x is a single word in the sentence, and the output sequence is recorded as , where y is the part-of-speech tag of x, and a set of feature functions are defined ,in, is the index of the characteristic function, is the position index of the word in the sentence, is the part-of-speech tag at position i, It's location The part-of-speech tagging is used for each feature function Assign a weight , for a given sentence x and a part-of-speech tag sequence y, use the feature function and weight to calculate the score of the tag sequence: , by normalizing the scores, we get the conditional probability of annotating sequence Y given sentence X: During the training process, the weights of the feature functions are learned using the training data In the testing and application phases, for a given scenic environment sentence X, the scores of all part-of-speech tag sequences are calculated, and the one with the highest score is selected as the tag.

7. According to claim 4, a smart cultural tourism system based on Python deep learning is characterized in that: The sentiment analysis module uses natural language processing technology to extract key information from comments. After completing the sentiment tendency judgment, the classification results are analyzed. The sentiment tendency is divided into three categories: positive, negative and neutral. Represents, decomposing the review C into a sequence of words , let word The part of speech is ,but Represents the probability of predicting the part of speech of a given word, identifies the named entities in the comments, and sets the named entity set as E. For each word in the comment , to determine whether it belongs to a named entity , using binary variables express, Indicates belonging to, Indicates that it does not belong to, and extracts keywords by calculating word frequency and inverse document frequency. Frequency of words Calculated as: ,in, It's a word The number of times it appears in the comment, N is the total number of words in the comment, the word Inverse document frequency Calculated as: , where M is the total number of documents, Contains the word The importance of keywords is expressed by TF-IDF value, and the formula is: .

8. According to the smart cultural tourism system based on Python deep learning in claim 1, it is characterized in that: The result display module receives complete analysis data from the sentiment analysis system, uses data visualization technology to generate a chart of tourist satisfaction distribution, and compiles a report with a clear structure and strict logic based on the analysis results. It provides dynamic interactive functions for charts and reports, allowing users to customize the displayed chart type, color, and font according to their own needs, and uploads the generated charts and reports to the online platform for decision makers to view at any time. It provides a user feedback channel, collects feedback from tourism management departments and scenic spots on the result display module, and continuously optimizes the function and performance of the result display module based on user feedback and market demand.

9. According to the smart cultural tourism system based on Python deep learning in claim 1, it is characterized in that: The feedback improvement module receives and integrates the output results from the sentiment analysis module, uses the data warehouse to collect and analyze historical service data, market trends and external information of competitor dynamics, builds an intelligent recommendation model based on the sentiment analysis results and historical data and the logistic regression model, conducts a multi-dimensional evaluation of the generated improvement plan, converts the evaluated improvement plan into an actual action plan, and assigns it to relevant departments and responsible persons for execution. After the improvement measures have been implemented for a period of time, relevant data is collected, the implementation effect is comprehensively evaluated, and the effect evaluation results are fed back to the sentiment analysis module as new input data for continuous optimization of the sentiment analysis model and generation of improvement suggestions.

10. The intelligent cultural tourism system based on Python deep learning according to claim 9 is characterized in that: The feedback improvement module builds an intelligent recommendation model based on the logistic regression model, collects sentiment analysis result data and related historical data, extracts meaningful features from the original data, and forms a feature set. ,in, The average length of stay of tourists, the number of service personnel, and the score of scenic spot facilities are used to construct a logistic regression model. The basic form of the logistic regression model is: ,in, is a binary variable, is the intercept term, are the regression coefficients to be estimated. These parameters are estimated through maximum likelihood estimation. The model is trained using feature data and regularization parameters are adjusted to prevent overfitting. New data is input into the trained model to identify potential problems in the service. Based on the output of the model, personalized improvement suggestions and solutions are generated for specific potential problems.

Images