Intelligent corpus data query system based on multivariate cultural background

Abstract

The invention discloses an intelligent corpus data query system based on a multivariate cultural background. According to the invention, through the application of the accent classification sub-module, the efficiency and accuracy of ELF accent research are significantly improved. Firstly, a random forest model integrates multi-dimensional data such as audio features, speaker information and environmental factors in a data preparation stage, and comprehensiveness of input features is ensured. Then, the reasonability of model training and verification is ensured through a data division strategy, key features are screened out through importance evaluation in the feature selection step, the data complexity is reduced, and the model training efficiency is improved. In the model construction process, the robustness of the model is enhanced through random training of multiple decision trees, and the overfitting problem is avoided. And finally, in a parameter optimization link, an optimal parameter combination is found through grid search or random search, so that the classification accuracy is further improved, and reliable data support is provided for researchers.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent corpus, and in particular relates to an intelligent corpus data query system based on a multicultural background. Background Art

[0002] The ELF accent, or English as a Lingua Franca, is a variant of English widely used in international communication. It is not based on the English of any particular country, but rather incorporates characteristics of various English dialects, aiming to improve the efficiency and comprehensibility of cross-cultural communication. In the context of globalization, it is increasingly valued and considered a universal language tool in international communication. The ELF Accent Corpus is a language database specifically collected and organized for the study of the ELF accent. This corpus aims to provide researchers with a comprehensive and diverse sample of ELF language usage, covering the phonetic, lexical, and grammatical characteristics of speakers from different native language backgrounds when using English internationally. The ELF Accent Corpus Query System is an online tool designed specifically for researchers, allowing users to efficiently search and analyze data from the ELF Accent Corpus. The system offers a variety of query functions, including keyword search, phonetic feature filtering, and contextual clarification, allowing researchers to quickly locate relevant data based on their specific needs.

[0003] However, existing techniques fail to consider the time series characteristics of accent data and are unable to effectively analyze the trends of accent changes over time. Furthermore, in trend analysis, existing techniques may fail to accurately identify and fit time series data, resulting in distorted analysis results. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent corpus data query system based on a multicultural background in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: an intelligent corpus data query system based on a multicultural background, the system comprising: an audio collection module, a metadata annotation module, a data preprocessing module, a corpus construction module, a query interface module, a data analysis module, a user management module and a system maintenance module;

[0006] The data analysis module is internally provided with an accent classification submodule, a trend analysis submodule, a visualization display submodule and a statistical analysis submodule;

[0007] The output end of the audio collection module is connected to the input end of the metadata annotation module; the output end of the metadata annotation module is connected to the input end of the data preprocessing module;

[0008] The output end of the data preprocessing module is connected to the input end of the corpus construction module;

[0009] The output of the corpus construction module is connected to the input of the query interface module

[0010] The output end of the query interface module can be connected to the input end of the data analysis module;

[0011] The user management module is connected with the audio collection module, metadata annotation module, data preprocessing module, corpus construction module, query interface module and data analysis module through authority management and user identity authentication;

[0012] The system maintenance module is connected with all other modules through system monitoring and log management.

[0013] In a preferred embodiment, the audio collection module is the data entry of the system, responsible for collecting audio data of ELF accent from various channels and devices. The composition of this module includes: 1) data collection tools, such as recording equipment, online audio crawlers, etc., for real-time or batch collection of audio data; 2) data format converter, which converts audio files of different formats (such as WAV, MP3, etc.) into a unified format of the system; 3) metadata preliminary extractor, which extracts basic metadata information from audio files, such as recording time, location, device model, etc.; 4) data storage interface, which stores the collected audio data and preliminary metadata in a central database or distributed storage system. The module can be integrated with a smartphone application, allowing users to directly record and upload ELF accent conversations, while automatically recording relevant information about the recording environment;

[0014] The metadata annotation module is responsible for performing detailed metadata annotation on the collected audio data to enhance the validity and queryability of the data. This module consists of: 1) an annotation tool, which provides a user-friendly interface for annotators to input and edit metadata, such as speaker information, accent characteristics, and conversation scenes; 2) an automated annotation engine, which uses speech recognition and natural language processing technologies to automatically extract some metadata, such as keywords and speaking speed; 3) an annotation rule library, which stores annotation standards and rules to ensure the consistency and accuracy of annotation; and 4) a quality review mechanism, which conducts spot checks and reviews of annotation results to ensure data quality. The module can automatically identify the language type in the audio and prompt annotators to further refine the accent type and regional characteristics.

[0015] In a preferred embodiment, the data preprocessing module is responsible for cleaning, standardizing, and formatting audio data and metadata in preparation for subsequent analysis. This module comprises: 1) a data cleaning tool to remove noise, silence, and invalid data; 2) an audio enhancer to enhance audio with poor sound quality, such as denoising and volume equalization; 3) a metadata normalizer to unify metadata from different sources into a unified format and standard; and 4) a data segmenter to divide long audio into smaller segments as needed for analysis and annotation. The module can automatically detect and remove long silences in audio and segment conversational audio into multiple segments based on speaker switching points, facilitating subsequent detailed analysis.

[0016] In a preferred embodiment, the corpus construction module is responsible for integrating preprocessed audio data and metadata into a structured corpus. This module comprises: 1) a database designer, which defines the structure, fields, and relationships of the corpus, such as audio file tables, metadata tables, and index tables; 2) a data import tool, which imports preprocessed data into the corpus in batches; 3) an index builder, which creates efficient data indexes, such as full-text indexes and keyword indexes, to accelerate queries; and 4) a corpus management system, which provides maintenance, updates, and backup functions for the corpus. The module can design a multi-level indexing system, allowing users to quickly retrieve audio data based on multiple dimensions, such as accent type, region, and speaker characteristics.

[0017] In a preferred embodiment, the query interface module provides users with flexible and convenient corpus query capabilities. This module comprises: 1) a query interface, providing an intuitive query input box, filter criteria, and search options; 2) a query parser, which parses user-entered queries into query statements understandable to the system; 3) a result sorter and filter, which sorts, filters, and paginates query results based on user needs; and 4) a result displayer, which presents query results in a list, summary, or detailed format, and provides audio playback and metadata viewing capabilities. The module supports natural language queries, such as "find all ELF-accented dialogues from Southeast Asia," and returns a list of relevant audio and a metadata summary.

[0018] In a preferred embodiment, the accent classification submodule is trained using a random forest model, with the following specific steps:

[0019] Model training: Data preparation: Input features: including audio features of accent (such as pitch, duration, intensity, etc.), speaker information (such as age, gender, etc.), environmental factors (such as background noise, etc.);

[0020] Data partitioning: Divide the dataset into training set and validation set with a ratio of 8:2 or 7:3;

[0021] Feature selection: Based on feature importance assessment, select features that have a greater impact on accent;

[0022] Model construction: Build multiple decision trees, randomly select some features and samples for each tree during training; each tree is fully grown without pruning;

[0023] Parameter optimization: Optimize the parameters of random forest, including the number of trees, tree depth, and number of feature selections, through grid search or random search methods;

[0024] The feature importance evaluation formula is:

[0025] Where:

[0026] Importance(f) represents the importance of feature f;

[0027] N represents the number of decision trees;

[0028] ΔIi(f) represents the reduction in impurity of feature f in the i-th tree; the reduction in impurity is calculated using the Gini impurity and information gain indicators;

[0029] Grid search finds the optimal parameter settings by traversing all possible parameter combinations. The calculation formula is:

[0030] BestParams=argmaxθ∈ΘPerformance(Model(θ,X train ,y train ),X val ,y val )

[0031] In the formula

[0032] BestParamsBestParams represents the optimal parameter combination;

[0033] Θ represents the set of all possible parameter combinations;

[0034] Model(θ,X train ,y train ) means using parameter θ in the training set (X train ,y train )The model obtained by training;

[0035] Performance represents the model performance evaluation function, including accuracy, recall, and F1 value;

[0036] (X val ,y val ) represents the validation set.

[0037] In a preferred embodiment, the trend analysis submodule is internally provided with a trend detection module that uses methods such as ARIMA and seasonal decomposition to analyze time series data of accent features;

[0038] The trend analysis submodule is internally provided with a moving average model algorithm to predict the trend of emotion changes over time. The calculation content includes:

[0039] ①. Time series data preparation: Arrange the emotion classification results in chronological order to form an emotion time series.

[0040] ②. Model identification: Determine the parameters of the ARIMA model, including the number of autoregressive terms (p), the order of differencing (d), and the number of moving average terms (q).

[0041] ③. Model estimation: Use historical sentiment data to estimate the parameters of the ARIMA model.

[0042] ④. Trend prediction: Use the estimated model to predict future sentiment trends;

[0043] The basic formula of the ARIMA model is:

[0044] (1-φ1L-φ2L 2 -...-φ p L p )(1-L) d X t =c+θ1LZ t +θ2L 2 Z t +...+θ q L q Z t ;

[0045] in:

[0046] X_t is the value of the time series data at time point t.

[0047] L is the lag operator, L^k X_t=X_{tk}.

[0048] φ1,φ2,...,φp are the coefficients of the autoregressive term.

[0049] d is the difference order.

[0050] θ1,θ2,...,θq are the coefficients of the moving average term.

[0051] Z_t is a white noise sequence.

[0052] c is a constant term.

[0053] In a preferred embodiment, the visualization submodule consists of the following components: 1) a data visualization engine, responsible for converting the processed ELF accent data into visualization elements; 2) a graphical interface design, including map display, chart generation (such as bar charts, line charts, pie charts, etc.), and interactive controls for displaying accent distribution and characteristics; 3) a dynamic update mechanism to ensure that the visualization content changes in real time as data is updated; and 4) user customization options, providing personalized settings such as color schemes and chart type switching to meet the needs of different users.

[0054] The statistical analysis submodule includes the following components: 1) a data preprocessing unit that cleans, normalizes, and standardizes raw data; 2) a statistical calculation engine that performs statistical tests, correlation analysis, regression analysis, and more; 3) a result output and interpretation unit that presents analysis results concisely and provides professional interpretation; 4) multidimensional data analysis tools, such as principal component analysis and cluster analysis, that discover underlying patterns and structures in the data; and 5) a report generator that automatically generates detailed statistical analysis reports. Together, these components form a powerful and flexible data analysis platform.

[0055] In a preferred embodiment, the user management module is responsible for user registration, authentication, permission management, and personal information maintenance within the system. This module comprises: 1) a user registration system, which allows new users to create accounts and enter basic information; 2) an identity authentication system, which verifies user identities through usernames and passwords, verification codes, or third-party authentication; 3) a permission management system, which defines the roles and permissions of different users, such as administrators, annotators, and regular users; and 4) a personal information center, which allows users to view and edit their personal information, query history, and preferences. The module can set different permission levels to ensure that only certified annotators can access metadata annotation functions, while regular users can only query and browse.

[0056] In a preferred embodiment, the system maintenance module is responsible for the stable operation, performance monitoring, and troubleshooting of the system. This module comprises: 1) a system monitor, which monitors the system's operating status, resource usage, and performance indicators, such as CPU usage, memory usage, and disk space, in real time; 2) a log management system, which records the system's operation logs, error logs, and audit logs to facilitate problem tracking and fault analysis; 3) a backup and recovery tool, which regularly backs up system data and configuration information and restores them when needed to ensure data security and integrity; 4) a system update and upgrade mechanism, which is responsible for system version updates, feature upgrades, and security patch installation to adapt to new needs and address potential security threats; 5) a fault alarm system, which automatically sends an alarm notification to the administrator when a system anomaly or performance degradation is detected; and 6) a performance optimization tool, which analyzes system bottlenecks and provides optimization suggestions, such as database index optimization and cache strategy adjustment. The module can set an automatic backup strategy to perform incremental backups of the corpus every morning, and automatically trigger the performance optimization tool to perform analysis and adjustments when system performance degradation is detected, ensuring the efficient and stable operation of the system.

[0057] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0058] 1. The present invention significantly improves the efficiency and accuracy of ELF accent research through the application of the accent classification submodule. First, the random forest model integrates multi-dimensional data such as audio features, speaker information, and environmental factors during the data preparation phase, ensuring the comprehensiveness of the input features. Next, the data partitioning strategy ensures the rationality of model training and verification, while the feature selection step screens key features through importance assessment, reducing data complexity and improving the efficiency of model training. During the model construction process, random training of multiple decision trees enhances the robustness of the model and avoids overfitting. Finally, the parameter optimization step finds the optimal parameter combination through grid search or random search, further improving classification accuracy. The organic combination of these steps enables the system to efficiently and accurately complete the accent classification task, providing researchers with reliable data support.

[0059] 2. In this invention, the introduction of the trend analysis submodule greatly deepens the system's ability to analyze and predict trends in ELF accent changes. During the time series data preparation phase, the system arranges the emotion classification results in chronological order, laying the foundation for trend analysis. The model identification phase achieves accurate fitting of time series data by determining the parameters of the ARIMA model. The model estimation process optimizes the model parameters using historical data, improving the accuracy of predictions. Finally, the trend prediction function uses the estimated model to predict future emotion trends, providing researchers with a forward-looking perspective. In addition, the module also uses methods such as seasonal decomposition to further reveal the cyclical fluctuations and long-term trends of accent characteristics. Through these analytical methods, the system can not only describe the current accent distribution but also predict future trends, providing strong support for language evolution research, cross-cultural communication, and the development of teaching strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a block diagram of the overall system of the present invention;

[0061] Figure 2 This is a system block diagram of the data analysis module in the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] Example:

[0064] Reference Figure 1-2 , an intelligent corpus data query system based on multicultural background, the system includes: audio collection module, metadata annotation module, data preprocessing module, corpus construction module, query interface module, data analysis module, user management module and system maintenance module;

[0065] The data analysis module is internally configured with an accent classification submodule, a trend analysis submodule, a visualization display submodule, and a statistical analysis submodule;

[0066] The output of the audio collection module (the collected raw audio data) is connected to the input of the metadata annotation module; the output of the metadata annotation module (the annotated audio data and metadata) is connected to the input of the data preprocessing module;

[0067] The output of the data preprocessing module (preprocessed audio data and metadata) is connected to the input of the corpus construction module;

[0068] The output of the corpus construction module (the constructed corpus) is connected to the input of the query interface module. The output of the query interface module (the user query result) can be connected to the input of the data analysis module (for further analysis).

[0069] The user management module is connected with the audio collection module, metadata annotation module, data preprocessing module, corpus construction module, query interface module and data analysis module through permission management and user authentication;

[0070] The system maintenance module is connected with all other modules through system monitoring and log management.

[0071] The audio collection module is the data entry point of the system, responsible for collecting audio data of the ELF accent from various channels and devices. The module consists of: 1) data collection tools, such as recording equipment, online audio crawlers, etc., for real-time or batch collection of audio data; 2) data format converters, which convert audio files of different formats (such as WAV, MP3, etc.) into a unified format for the system; 3) metadata preliminary extractors, which extract basic metadata information from audio files, such as recording time, location, device model, etc.; 4) data storage interfaces, which store the collected audio data and preliminary metadata in a central database or distributed storage system. The module can be integrated with a smartphone application, allowing users to directly record and upload ELF accent conversations, while automatically recording relevant information about the recording environment;

[0072] The metadata annotation module is responsible for detailed metadata annotation of the collected audio data to enhance the validity and queryability of the data. The composition of this module includes: 1) Annotation tools, which provide a user-friendly interface for annotators to input and edit metadata, such as speaker information, accent characteristics, dialogue scenes, etc.; 2) Automated annotation engine, which uses speech recognition and natural language processing technology to automatically extract some metadata, such as keywords, speaking speed, etc.; 3) Annotation rule library, which stores annotation standards and rules to ensure the consistency and accuracy of annotation; 4) Quality review mechanism, which conducts spot checks and reviews on annotation results to ensure data quality. The module can automatically identify the language type in the audio and prompt the annotator to further refine the accent type and regional characteristics.

[0073] The data preprocessing module is responsible for cleaning, standardizing, and formatting audio data and metadata in preparation for subsequent analysis. This module comprises: 1) a data cleaning tool to remove noise, silence, and invalid data; 2) an audio enhancer to enhance audio with poor sound quality, such as denoising and volume equalization; 3) a metadata normalizer to unify metadata from different sources into a unified format and standard; and 4) a data segmenter to divide long audio into smaller segments as needed for analysis and annotation. The module automatically detects and removes long silences in audio and segments conversational audio into multiple segments based on speaker switching points, facilitating detailed analysis.

[0074] The corpus construction module is responsible for integrating preprocessed audio data and metadata into a structured corpus. This module comprises: 1) a database designer, which defines the corpus's structure, fields, and relationships, such as audio file tables, metadata tables, and index tables; 2) a data import tool, which batch-imports preprocessed data into the corpus; 3) an index builder, which creates efficient data indexes, such as full-text indexes and keyword indexes, to accelerate queries; and 4) a corpus management system, which provides maintenance, updates, and backup functions for the corpus. This module can design a multi-level indexing system, allowing users to quickly retrieve audio data based on various dimensions, such as accent type, region, and speaker characteristics.

[0075] The query interface module provides users with flexible and convenient corpus query capabilities. This module comprises: 1) a query interface, which provides an intuitive query input box, filter criteria, and search options; 2) a query parser, which parses user-entered queries into query statements understandable to the system; 3) a result sorter and filter, which sorts, filters, and paginates query results based on user needs; and 4) a result displayer, which presents query results in a list, summary, or detailed format, and provides audio playback and metadata viewing capabilities. The module supports natural language queries, such as "Find all ELF-accented dialogues from Southeast Asia," and returns a list of relevant audio files and a metadata summary.

[0076] The accent classification submodule is trained using a random forest model. The specific steps are as follows:

[0077] Model training: Data preparation: Input features: including audio features of accent (such as pitch, duration, intensity, etc.), speaker information (such as age, gender, etc.), environmental factors (such as background noise, etc.);

[0078] Data partitioning: Divide the dataset into training set and validation set with a ratio of 8:2 or 7:3;

[0079] Feature selection: Based on feature importance assessment, select features that have a greater impact on accent;

[0080] Model construction: Build multiple decision trees, randomly select some features and samples for each tree during training; each tree is fully grown without pruning;

[0081] Parameter optimization: Optimize the parameters of random forest, including the number of trees, tree depth, and number of feature selections, through grid search or random search methods;

[0082] The feature importance evaluation formula is:

[0083] Where:

[0084] Importance(f) represents the importance of feature f;

[0085] N represents the number of decision trees;

[0086] ΔIi(f) represents the reduction in impurity of feature f in the i-th tree; the reduction in impurity is calculated using the Gini impurity and information gain indicators;

[0087] Grid search finds the optimal parameter settings by traversing all possible parameter combinations. The calculation formula is:

[0088] BestParams=argmaxθ∈ΘPerformance(Model(θ,X train ,y train ),X val ,y val )

[0089] In the formula

[0090] BestParamsBestParams represents the optimal parameter combination;

[0091] Θ represents the set of all possible parameter combinations;

[0092] Model(θ,X train ,y train ) means using parameter θ in the training set (X train ,y train )The model obtained by training;

[0093] Performance represents the model performance evaluation function, including accuracy, recall, and F1 value;

[0094] (X val ,y val ) represents the validation set.

[0095] The trend analysis submodule is internally equipped with a trend detection module that uses ARIMA, seasonal decomposition and other methods to analyze time series data of accent characteristics;

[0096] The trend analysis submodule has a moving average model algorithm set up inside to predict the trend of sentiment changes over time. The calculation content includes:

[0097] ①. Time series data preparation: Arrange the emotion classification results in chronological order to form an emotion time series.

[0098] ②. Model identification: Determine the parameters of the ARIMA model, including the number of autoregressive terms (p), the order of differencing (d), and the number of moving average terms (q).

[0099] ③. Model estimation: Use historical sentiment data to estimate the parameters of the ARIMA model.

[0100] ④. Trend prediction: Use the estimated model to predict future sentiment trends;

[0101] The basic formula of the ARIMA model is:

[0102] (1-φ1L-φ2L 2 -...-φ p L p )(1-L) d X t =c+θ1LZ t +θ2L 2 Z t +...+θ q L q Z t ;

[0103] in:

[0104] X_t is the value of the time series data at time point t.

[0105] L is the lag operator, L^k X_t=X_{tk}.

[0106] φ1,φ2,...,φp are the coefficients of the autoregressive term.

[0107] d is the difference order.

[0108] θ1,θ2,...,θq are the coefficients of the moving average term.

[0109] Z_t is a white noise sequence.

[0110] c is a constant term.

[0111] The visualization submodule consists of the following components: 1) a data visualization engine, responsible for converting the processed ELF accent data into visualization elements; 2) a graphical interface design, including map display, chart generation (such as bar charts, line charts, pie charts, etc.), and interactive controls for displaying accent distribution and characteristics; 3) a dynamic update mechanism to ensure that the visualization content changes in real time as data is updated; and 4) user customization options, providing personalized settings such as color schemes and chart type switching to meet the needs of different users.

[0112] The statistical analysis submodule includes the following components: 1) a data preprocessing unit that cleans, normalizes, and standardizes raw data; 2) a statistical calculation engine that performs statistical tests, correlation analysis, regression analysis, and more; 3) a result output and interpretation unit that presents analysis results concisely and provides professional interpretation; 4) multidimensional data analysis tools, such as principal component analysis and cluster analysis, that discover underlying patterns and structures in the data; and 5) a report generator that automatically generates detailed statistical analysis reports. Together, these components form a powerful and flexible data analysis platform.

[0113] The user management module is responsible for user registration, authentication, permission management, and personal information maintenance within the system. This module comprises: 1) a user registration system, which allows new users to create accounts and enter basic information; 2) an identity authentication system, which verifies user identities through usernames and passwords, verification codes, or third-party authentication; 3) a permission management system, which defines the roles and permissions of different users, such as administrators, annotators, and regular users; and 4) a personal information center, which allows users to view and edit their personal information, query history, and preferences. The module can set different permission levels, ensuring that only certified annotators can access metadata annotation functions, while regular users can only query and browse.

[0114] The system maintenance module is responsible for the system's stable operation, performance monitoring, and troubleshooting. This module comprises: 1) a system monitor, which monitors the system's operating status, resource usage, and performance indicators, such as CPU utilization, memory usage, and disk space, in real time; 2) a log management system, which records the system's operation logs, error logs, and audit logs to facilitate problem tracking and fault analysis; 3) a backup and recovery tool, which regularly backs up system data and configuration information and restores it when needed to ensure data security and integrity; 4) a system update and upgrade mechanism, which is responsible for system version updates, feature upgrades, and security patch installation to adapt to new requirements and address potential security threats; 5) a fault alarm system, which automatically sends alarm notifications to administrators when system anomalies or performance degradation are detected; and 6) a performance optimization tool, which analyzes system bottlenecks and provides optimization suggestions, such as database index optimization and cache strategy adjustments. The module can set an automatic backup policy to perform incremental backups of the corpus every morning. When system performance degradation is detected, the performance optimization tool is automatically triggered to perform analysis and adjustments to ensure efficient and stable system operation.

[0115] The present invention significantly improves the efficiency and accuracy of ELF accent research through the application of the accent classification submodule. First, during the data preparation phase, the random forest model integrates multi-dimensional data such as audio features, speaker information, and environmental factors, ensuring the comprehensiveness of input features. Next, a data partitioning strategy ensures the rationality of model training and validation, while the feature selection step screens key features through importance assessment, reducing data complexity and improving model training efficiency. During model construction, random training of multiple decision trees enhances model robustness and avoids overfitting. Finally, parameter optimization finds the optimal parameter combination through grid search or random search, further improving classification accuracy. These integrated steps enable the system to efficiently and accurately complete accent classification tasks, providing researchers with reliable data support.

[0116] The introduction of the trend analysis submodule in this paper significantly enhances the system's ability to analyze and predict trends in ELF accents. During the time series data preparation phase, the system arranges emotion classification results in chronological order, laying the foundation for trend analysis. The model identification phase accurately fits the time series data by determining the parameters of the ARIMA model. The model estimation process optimizes the model parameters using historical data, improving prediction accuracy. Finally, the trend prediction function uses the estimated model to predict future emotion trends, providing researchers with a forward-looking perspective. Furthermore, this module employs methods such as seasonal decomposition to further reveal the cyclical fluctuations and long-term trends of accent characteristics. Through these analytical methods, the system not only describes the current accent distribution but also predicts future trends, providing strong support for language evolution research, cross-cultural communication, and the development of teaching strategies. In summary, this system not only improves research efficiency and accuracy, but also significantly enhances trend analysis and prediction capabilities, demonstrating its unique value in the field of ELF accent research.

[0117] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0118] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent corpus data query system based on a multicultural background, characterized by: The system includes: an audio collection module, a metadata annotation module, a data preprocessing module, a corpus construction module, a query interface module, a data analysis module, a user management module and a system maintenance module; The data analysis module is internally provided with an accent classification submodule, a trend analysis submodule, a visualization display submodule and a statistical analysis submodule; The output end of the audio collection module is connected to the input end of the metadata annotation module; the output end of the metadata annotation module is connected to the input end of the data preprocessing module; The output end of the data preprocessing module is connected to the input end of the corpus construction module; The output of the corpus construction module is connected to the input of the query interface module The output end of the query interface module is connected to the input end of the data analysis module; The user management module is connected with the audio collection module, metadata annotation module, data preprocessing module, corpus construction module, query interface module and data analysis module through authority management and user identity authentication; The system maintenance module is connected with all other modules through system monitoring and log management.

2. The intelligent corpus data query system based on a multicultural background according to claim 1, characterized in that: The audio collection module includes: 1) a data acquisition tool; 2) a data format converter; 3) a metadata preliminary extractor; 4) a data storage interface; The metadata annotation module includes: 1) annotation tool; 2) automatic annotation engine; 3) annotation rule library; 4) quality review module.

3. The intelligent corpus data query system based on a multicultural background according to claim 1, characterized in that: The data preprocessing module includes: 1) a data cleaning tool to remove noise, silence segments, and invalid data; 2) an audio enhancer to enhance audio with poor sound quality; 3) a metadata normalizer to unify the format and standards of metadata from different sources; and 4) a data segmenter to split long audio into smaller segments as needed for easier analysis and annotation.

4. The intelligent corpus data query system based on a multicultural background according to claim 1, characterized in that: The corpus construction module includes: 1) a database designer, which defines the structure of the corpus; 2) a data import tool, which imports preprocessed data into the corpus in batches; 3) an index builder, which creates an efficient data index; and 4) a corpus management system, which provides maintenance, update, and backup functions for the corpus.

5. The intelligent corpus data query system based on multicultural background according to claim 1, characterized in that: The query interface module includes: 1) a query interface; 2) a query parser; 3) result sorting and filtering; and 4) a result displayer, which displays query results in the form of a list, summary, or detailed information, and provides audio playback and metadata viewing functions.

6. The intelligent corpus data query system based on a multicultural background according to claim 1, characterized in that: The feature importance evaluation formula of the accent classification submodule is: Where: Importance(f) represents the importance of feature f; N represents the number of decision trees; ΔIi(f) represents the reduction in impurity of feature f in the i-th tree; the reduction in impurity is calculated using the Gini impurity and information gain indicators.

7. The intelligent corpus data query system based on multicultural background according to claim 1, characterized in that: The basic formula of the ARIMA model of the trend analysis submodule is: (1-φ1L-φ2L 2 -...-f p L p (1-L) d X t =c+θ1LZ t +θ2L 2 Z t +...+θ q L q Z t ; in: X_t is the value of the time series data at time point t; L is the lag operator, L^k X_t=X_{tk}; φ1,φ2,...,φp are the coefficients of the autoregressive terms; d is the difference order; θ1,θ2,...,θq are the coefficients of the moving average term; Z_t is a white noise sequence; c is a constant term.

8. The intelligent corpus data query system based on multicultural background according to claim 1, characterized in that: The visualization display submodule consists of the following parts: 1) a data visualization engine, responsible for converting the processed ELF accent data into visualization elements; 2) a graphical interface design, including map display and chart generation; 3) a dynamic update mechanism module; 4) User-defined options provide personalized settings; The statistical analysis submodule includes the following components: 1) a data preprocessing unit that cleans, normalizes and standardizes the raw data; 2) Statistical calculation engine, which performs statistical tests, correlation analysis, and regression analysis; 3) Result output and interpretation unit, which presents analysis results in a concise manner and provides professional interpretation; 4) Multidimensional data analysis tools, including principal component analysis and cluster analysis, to discover underlying patterns and structures in data; 5) Report generator, automatically generates detailed statistical analysis reports.

9. The intelligent corpus data query system based on multicultural background according to claim 1, characterized in that: The user management module includes: 1) a user registration system that allows new users to create accounts and fill in basic information; 2) an identity authentication system that verifies user identity through username and password, verification code or third-party authentication; 3) a permission management system that defines the roles and permissions of different users; 4) a personal information center that allows users to view and edit their personal information, query history and preference settings.

10. The intelligent corpus data query system based on multicultural background according to claim 1, characterized in that: The system maintenance module includes: 1) a system monitor, which monitors the system's operating status, resource usage and performance indicators in real time; 2) a log management system, which records the system's operation logs, error logs and audit logs to facilitate problem tracking and fault analysis; 3) a backup and recovery tool, which regularly backs up system data and configuration information and restores them when needed to ensure data security and integrity; 4) a system update and upgrade mechanism, which is responsible for system version updates, function upgrades and security patch installation to adapt to new needs and respond to potential security threats; 5) a fault alarm system, which automatically sends an alarm notification to the administrator when a system anomaly or performance degradation is detected; 6) a performance optimization tool, which analyzes system bottlenecks and makes optimization suggestions.

Images