Text Datasets Research Articles

Development and Evaluation of Gold Standard Dataset for Sentiment Analysis of Tweets

Pre-labeled data is typically required for supervised machine learning. A limited number of object classes in the majority of open access and pre-annotated datasets make them unsuitable for certain tasks, even though they are readily available for training machine learning algorithms. For custom models, previously available pre-annotated data is typically insufficient, so gathering and preparing training data is necessary for the majority of real-world applications. The quantity and quality of annotations clearly trade-off with one another. Either more annotated data can be produced or better data quality can be guaranteed by allocating time and resources. Development of the gold standard by annotating textual information is an essential part of the “Text Analytics” domain in the field of “Natural Language Processing-NLP”. In “Text Analytics”, annotation can be done by adopting a manual, semi-automatic or automatic approach. In the case of the manual approach, annotators often work with partial parts of the corpus, and the results are generalized by automated text classification which may affect the final classification results. Annotations reliability and suitability of assigned labels are particularly important in the NLP applications related to opinion mining or sentiment analysis. In this research study, we have evaluated the significance of the annotation process on a novel dataset that contained multiple languages (English, Roman Urdu), a free text dataset that was extracted from Twitter. This unique dataset contained multiple languages which makes this annotation process essential for researching this data. Using this multi-language dataset, we examine the inter-annotator agreement in multiclass and multi-label sentiment annotation. To scrutinize the reliability of this research work, several annotation agreement metrics, statistical analysis, and Machine Learning methods have been considered to evaluate the accuracy of resulting annotations. It was observed that the annotation process is significant and a complex step that is essential for the proper implementation of Natural Language processing tasks for text analytics in machine learning. During this research, different gaps were identified and resolved which can impact the overall reliability of the annotation process which are reported in this paper. We conclude that while inaccurate annotations worsen the results, the impact is minimal, at least when using text data. The advantages of the larger annotated data set (obtained by employing subpar auto-annotation techniques) surpass the degradation resulting from the use of annotated data.

Hierarchical Understanding in Robotic Manipulation: A Knowledge-Based Framework

In the quest for intelligent robots, it is essential to enable them to understand tasks beyond mere manipulation. Achieving this requires a robust parsing mode that can be used to understand human cognition and semantics. However, the existing methods for task and motion planning lack generalization and interpretability, while robotic knowledge bases primarily focus on static manipulation objects, neglecting the dynamic tasks and skills. To address these limitations, we present a knowledge-based framework for hierarchically understanding various factors and knowledge types in robotic manipulation. Using this framework as a foundation, we collect a knowledge graph dataset describing manipulation tasks from text datasets and an external knowledge base with the assistance of large language models and construct the knowledge base. The reasoning tasks of entity alignment and link prediction are accomplished using a graph embedding method. A robot in real-world environments can infer new task execution plans based on experience and knowledge, thereby achieving manipulation skill transfer.

Enhanced industrial text classification via hyper variational graph-guided global context integration.

Joint local context that is primarily processed by pre-trained models has emerged as a prevailing technique for text classification. Nevertheless, there are relatively few classification applications on small sample of industrial text datasets. In this study, an approach of employing global enhanced context representation of the pre-trained model to classify industrial domain text is proposed. To achieve the application of the proposed technique, we extract primary text representations and local context information as embeddings by leveraging the BERT pre-trained model. Moreover, we create a text information entropy matrix through statistical computation, which fuses features to construct the matrix. Subsequently, we adopt BERT embedding and hyper variational graph to guide the updating of the existing text information entropy matrix. This process is subjected to iteration three times. It produces a hypergraph primary text representation that includes global context information. Additionally, we feed the primary BERT text feature representation into capsule networks for purification and expansion as well. Finally, the above two representations are fused to obtain the final text representation and apply it to text classification through feature fusion module. The effectiveness of this method is validated through experiments on multiple datasets. Specifically, on the CHIP-CTC dataset, it achieves an accuracy of 86.82% and an F1 score of 82.87%. On the CLUEEmotion2020 dataset, the proposed model obtains an accuracy of 61.22% and an F1 score of 51.56%. On the N15News dataset, the accuracy and F1 score are 72.21% and 69.06% respectively. Furthermore, when applied to an industrial patent dataset, the model produced promising results with an accuracy of 91.84% and F1 score of 79.71%. All four datasets are significantly improved by using the proposed model compared to the baselines. The evaluation result of the four dataset indicates that our proposed model effectively solves the classification problem.

MCICT: Graph convolutional network-based end-to-end model for multi-label classification of imbalanced clinical text

The rapid growth of clinical text data requires accurate and powerful automated classification methods to support medical decision making and personalized healthcare. The multi-label classification task for clinical texts is designed to assign the most relevant set of labels to each clinical text. However, this task presents two significant challenges: (1) how to accurately extract fine-grained semantic features from complex clinical texts, and (2) how to effectively mitigate the issue of label class imbalance. To overcome these problems, we innovatively propose a novel Multi-label Classification of Imbalanced Clinical Text (MCICIT) model. In order to obtain fine-grained semantic features from clinical texts, we utilize the specialized pre-trained language model BioBERT, tailored for biomedical texts. To tackle the challenge of label class imbalance, we present a Co-occurrence Based and Embeddings with Additional Information Enhanced Graph Convolutional Network (CoEAI-GCN) module. On one hand, we enrich the label content by incorporating additional information to acquire more accurate word embeddings as the feature matrix. On the other hand, we combine the co-occurrence relationship of labels to construct a correlation matrix. Ultimately, label representations are learned through a graph convolutional network. By conducting multi-label classification experiments on two clinical text datasets extracted from real medical systems, our model achieves a 3.2% and 0.5% improvement in F1 scores, respectively, compared to state-of-the-art deep learning models. Additionally, we conduct ablation studies to explore the behaviors of the proposed model. These results together demonstrate that our proposed MCICT effectively enhances the classification performance of imbalanced clinical texts.

Text classification algorithm of tourist attractions subcategories with modified TF-IDF and Word2Vec.

Text classification, as an important research area of text mining, can quickly and effectively extract valuable information to address the challenges of organizing and managing large-scale text data in the era of big data. Currently, the related research on text classification tends to focus on the application in fields such as information filtering, information retrieval, public opinion monitoring, and library and information, with few studies applying text classification methods to the field of tourist attractions. In light of this, a corpus of tourist attraction description texts is constructed using web crawler technology in this paper. We propose a novel text representation method that combines Word2Vec word embeddings with TF-IDF-CRF-POS weighting, optimizing traditional TF-IDF by incorporating total relative term frequency, category discriminability, and part-of-speech information. Subsequently, the proposed algorithm respectively combines seven commonly used classifiers (DT, SVM, LR, NB, MLP, RF, and KNN), known for their good performance, to achieve multi-class text classification for six subcategories of national A-level tourist attractions. The effectiveness and superiority of this algorithm are validated by comparing the overall performance, specific category performance, and model stability against several commonly used text representation methods. The results demonstrate that the newly proposed algorithm achieves higher accuracy and F1-measure on this type of professional dataset, and even outperforms the high-performance BERT classification model currently favored by the industry. Acc, marco-F1, and mirco-F1 values are respectively 2.29%, 5.55%, and 2.90% higher. Moreover, the algorithm can identify rare categories in the imbalanced dataset and exhibit better stability across datasets of different sizes. Overall, the algorithm presented in this paper exhibits superior classification performance and robustness. In addition, the conclusions obtained by the predicted value and the true value are consistent, indicating that this algorithm is practical. The professional domain text dataset used in this paper poses higher challenges due to its complexity (uneven text length, relatively imbalanced categories), and a high degree of similarity between categories. However, this proposed algorithm can efficiently implement the classification of multiple subcategories of this type of text set, which is a beneficial exploration of the application research of complex Chinese text datasets in specific fields, and provides a useful reference for the vector expression and classification of text datasets with similar content.

Applied Linguistics driven Artificial Intelligence for Automated Sentiment Detection and Classification

The widespread dissemination of World Wide Web has paved the way to express individual sentiments. Also, it is a medium with a massive quantity of data where the user can view the opinions of other users that are categorized into dissimilar sentimental classes and are growing increasingly as a major aspect in decision making. Sentiment analysis (SA) is a method utilized in natural language processing (NLP) that defines the emotion or sentiment formulated in the text portion. SA method is often performed on text datasets to assist in accepting client requirements, businesses monitoring brands, and product sentiment in customer feedback. SA is the challenging and most common complication in artificial intelligence (AI). It applies automated mechanisms to identify physiological information namely feelings, thoughts, and attitudes shown in text and indicated through blogs, social networks, and news. This manuscript develops Applied Linguistics driven Artificial Intelligence for Automated Sentiment Detection and Classification (ALAI-ASDC) technique. The preprocessing stage includes tokenizing and cleaning textual information, followed by encoder words into vector representation using pretrained GloVe embeddings. This embedding captures semantic similarities between words, which provides an abundant depiction of textual information for SA. Integrating single-valued neutrosophic fuzzy soft expert set (SVNFSES) improves the SA method by addressing imprecision, uncertainty, and ambiguity inherent in text sentiment expression. FNS enables the representation of linguistic variables with degrees of truth, falsity, and indeterminacy, allowing a nuanced understanding of sentiment polarity. Moreover, the Hybrid Jelly Particle Swarm Optimization (HJPSO) is applied for the parameter tuning of the SA technique. Enhancing the performance of the SA model. Empirical analysis illustrates the efficiency of the presented technique in precisely categorizing sentiment polarity in different textual datasets

Bayesian Optimization with Tree Ensembles to Improve Depression Screening on Textual Datasets

Bayesian Optimization with Tree Ensembles to Improve Depression Screening on Textual Datasets

Learning Multiresolution Features for Unsupervised Anomaly Localization on Industrial Textured Surfaces

In industrial quality assessment, monitoring whether the textured product contains defects is a critical step. Compared to a large number of defect-free images that are easy to obtain, anomaly samples are limited and vary randomly in size and type. It is challenging to develop an automatic and accurate texture defect localization system that only uses normal images for training. In this paper, a multi-resolution feature learning network is proposed to detect various texture defects in an unsupervised manner. A robust pre-trained model is first employed to extract the perceptual features from the input image, then the perceptual features of various layers are fed to the corresponding multi-scale autoencoder framework. This hierarchical alignment strategy aids in receiving multi-level information for locating anomalies of various sizes. Moreover, a residual attention module (RAM) is embedded in the architecture to further improve the detection performance. Our proposed method has achieved state-of-the-art performance on the texture dataset of MVTecAD. We also extended the experiment to the real industrial texture datasets, and its detection result is better than the major existing advanced techniques.

Towards AI-Enabled Approach for Urdu Text Recognition: A Legacy for Urdu Image Apprehension

Recognizing Urdu text in natural images is more challenging as compared to other languages, such as English, due to the cursive nature of Urdu script. However, Urdu scene text has not received enough attention from both industry and academia due to the lack of the dataset of Urdu text. We propose a large-scale Urdu Scene Text Dataset (USTD) to address this problem, which is designed for Urdu scene text detection and recognition. The proposed dataset contains 29674 text annotations (17877 Urdu and 11797 English), 749725 characters in 6389 images. It covers a wide variety of text images with both Nastaleeq and Naskh writing styles, taken from different streets and roads of Pakistan. The vast diversity of this dataset makes it a benchmark to work on and train robust neural networks for the detection and recognition of cursive text. Besides, baseline results are also provided with several state-of-the-art networks, including TextBoxes++, Seglink, DB(ResNet-50) and EAST for text localization and Convolutional Recurrent Neural Network (CRNN) for text recognition. To further evaluate the performance of these models, we have used the most popular evaluation matrices of precision, recall, and F-measure. Our experimental outputs reveal that an end-to-end combination of DB(ResNet-50) and CRNN provides the best results with precision, recall, and F-measure of 0.7526, 0.5974, and 0.6660, respectively.

Образцы текстов сгенерированных нейросетями по простому и комплексному промптам. Приложение к статье «Текстовые генеративные нейросети в исследовательской деятельности студентов».

The appendix contains a set of text data generated by the Gemini, Claude and Perplexity neural networks about outstanding teachers. The data was used to detect artificially created text in the Antiplagiat and Scribbr detectors.

Large Language Model-aided Life Cycle Analysis for Circular Bio-manufacturing

This paper explores the pivotal role of circular manufacturing methods and tools for biobased materials through an assessment tool built on a large language model (LLM) embedded mobile application. The LLMs, trained on extensive textual datasets, can provide precision and efficiency for the life cycle assessment (LCA) for biobased manufacturing. The tool automates data collection, categorises information from diverse sources, and supports ecological decision-making in material and manufacturing method selection. The research follows a methodology based on three main workflows, including database development, LCA specifications, and the LLM-embedded mobile interface development. Furthermore, the study recognizes the intricacy involved in utilising newly developed materials, especially considering the abundance of available formulations. It questions how the tool adapts to various material and manufacturing options and how it could augment practicality and applicability in real-life design and manufacturing stages. The study demonstrates that the tool offers a comprehensive and critical evaluation of biobased materials for manufacturing towards product development across design domains, unlocking new possibilities at the intersection of ecology, circular design, and digital tools.

An improved multi-stage framework for large-scale hierarchical text classification problems using a modified feature hashing and bi-filtering strategy

The classification of large-scale textual dataset is associated with a huge number of instances and millions of features which must be discriminated between large numbers of categories. The task requires the utilization of a defined hierarchy structure and tools that automatically classify instances within the hierarchy known as Large Scale Hierarchical Text Classification (LSHTC). Predicting the labels of instances by the employed classifiers is challenging due to the high number of features. Furthermore, the existing Dimensional Reduction (DR) approaches in cooperation with the LSHTC framework are still quite inefficient. In such a problem, an effective Hierarchical Dimensional Reduction approach can be advantageous in improving the performance of the LSHTC. Therefore, in this paper, we enhance the performance of LSHTC by proposing a Multi-stage Hierarchical Dimensional Reduction (MHDR) approach based on Modified Feature Hashing (MFH) and Hierarchical Bi-Filtering (HBF) method. In addition to alleviating bad collision and result discrepancy, experimental results show that the proposed approach has achieve the best performance in terms of micro-f1 and macro-f1 by recording average scores of 58.47% and 54.77% using TD-SVM, and average scores of 51.14% and 48.70% using TD-LR, respectively. The method also achieved 11% speed-up than the approaches compared.

Disease Discourse through Sentiment and Network Analysis

In the realm of public health, the discourse surrounding diseases significantly impacts awareness, funding, and stigma. Traditional methods of analyzing disease mentions often overlook the intricate relationships and sentiments expressed in text, which can provide deeper insights into public perception and information dissemination. Existing literature on disease discourse primarily focuses on quantitative metrics such as disease frequency counts or associations based on semantic similarity. However, such analyses rarely account for the sentiment of the discourse, which can play a critical role in shaping public perception and response to health information. Moreover, the complexity of textual data and the subtleties of language usage present significant challenges for extracting meaningful patterns, especially when considering the emotional context of disease-related discussions. This study introduces a novel approach that integrates sentiment analysis and network visualization to examine the discourse on diseases within a comprehensive textual dataset. By applying the TextBlob library, we analyze sentence-level sentiment and categorize it as positive or negative. Utilizing these sentiment scores, we construct two distinct network graphs to depict the relationships between diseases based on their co-occurrence within sentiment-laden sentences. The analysis reveals contrasting landscapes of disease discourse: the positive sentiment network highlights diseases frequently mentioned in the context of successful treatment or optimistic outcomes, forming distinct clusters around well-managed conditions. In contrast, the negative sentiment network elucidates diseases that coalesce around shared public concerns, fears, or complications, marking out critical areas for public health intervention. The findings underscore the importance of sentiment in understanding disease discourse, offering a novel perspective that can assist public health officials in tailoring communication strategies. By recognizing patterns in disease associations within negative contexts, interventions can be more effectively targeted to address misconceptions and alleviate public health anxieties. Conversely, identifying diseases with strong positive connections can guide campaigns to bolster preventive measures and highlight medical advancements. This methodology serves as a blueprint for leveraging sentiment analysis in the exploration of health-related textual data, paving the way for more nuanced and actionable insights into public health narratives.

Inverse-Like Antagonistic Scene Text Spotting via Reading-Order Estimation and Dynamic Sampling.

Scene text spotting is a challenging task, especially for inverse-like scene text, which has complex layouts, e.g., mirrored, symmetrical, or retro-flexed. In this paper, we propose a unified end-to-end trainable inverse-like antagonistic text spotting framework dubbed IATS, which can effectively spot inverse-like scene texts without sacrificing general ones. Specifically, we propose an innovative reading-order estimation module (REM) that extracts reading-order information from the initial text boundary generated by an initial boundary module (IBM). To optimize and train REM, we propose a joint reading-order estimation loss ( LRE ) consisting of a classification loss, an orthogonality loss, and a distribution loss. With the help of IBM, we can divide the initial text boundary into two symmetric control points and iteratively refine the new text boundary using a lightweight boundary refinement module (BRM) for adapting to various shapes and scales. To alleviate the incompatibility between text detection and recognition, we propose a dynamic sampling module (DSM) with a thin-plate spline that can dynamically sample appropriate features for recognition in the detected text region. Without extra supervision, the DSM can proactively learn to sample appropriate features for text recognition through the gradient returned by the recognition module. Extensive experiments on both challenging scene text and inverse-like scene text datasets demonstrate that our method achieves superior performance both on irregular and inverse-like text spotting.

TaeC: A manually annotated text dataset for trait and phenotype extraction and entity linking in wheat breeding literature.

Wheat varieties show a large diversity of traits and phenotypes. Linking them to genetic variability is essential for shorter and more efficient wheat breeding programs. A growing number of plant molecular information networks provide interlinked interoperable data to support the discovery of gene-phenotype interactions. A large body of scientific literature and observational data obtained in-field and under controlled conditions document wheat breeding experiments. The cross-referencing of this complementary information is essential. Text from databases and scientific publications has been identified early on as a relevant source of information. However, the wide variety of terms used to refer to traits and phenotype values makes it difficult to find and cross-reference the textual information, e.g. simple dictionary lookup methods miss relevant terms. Corpora with manually annotated examples are thus needed to evaluate and train textual information extraction methods. While several corpora contain annotations of human and animal phenotypes, no corpus is available for plant traits. This hinders the evaluation of text mining-based crop knowledge graphs (e.g. AgroLD, KnetMiner, WheatIS-FAIDARE) and limits the ability to train machine learning methods and improve the quality of information. The Triticum aestivum trait Corpus is a new gold standard for traits and phenotypes of wheat. It consists of 528 PubMed references that are fully annotated by trait, phenotype, and species. We address the interoperability challenge of crossing sparse assay data and publications by using the Wheat Trait and Phenotype Ontology to normalize trait mentions and the species taxonomy of the National Center for Biotechnology Information to normalize species. The paper describes the construction of the corpus. A study of the performance of state-of-the-art language models for both named entity recognition and linking tasks trained on the corpus shows that it is suitable for training and evaluation. This corpus is currently the most comprehensive manually annotated corpus for natural language processing studies on crop phenotype information from the literature.

Tapping the Economic Potential and Optimizing Strategies of Intangible Cultural Heritage under Digital Transformation

Abstract Intangible cultural heritage has important historical, cultural, and spiritual values. This paper selects the intangible cultural heritage of Yunnan Painting, the national architectural decoration of Dali, as the case study object. We use web crawler technology to extract the consumer’s comment text data, establishing the Yunnan Painting consumer comment text dataset. Secondly, we extract high-frequency words from the dataset using text mining technology, TF-IDF keyword extraction, and LDA themes, and cluster them using the K-means method. Finally, we evaluate the economic potential of Yunnan Painting in terms of consumers’ cognitive abilities and the emotional appeal of its products, based on the extracted results. The study reveals that the dressmaking, painting, and tourism industries are the main economic activities of Yunnan color painting, with the tourism industry requiring further strengthening. According to the text mining of high-frequency words, the frequency of “product” in Yunnan painting was the highest at 296 times, followed by “culture,” “technology,” “cultural creativity,” and “theme.” The analysis of consumers’ emotional text features reveals that most consumers have positive emotions towards Yunnan-colored paintings, while very few have negative emotions. The economic optimization strategy that is proposed is based on the results and provides a reference value for the economic development of other intangible cultural heritage.

Co-occurrence word model for news media hotspot mining-text mining method design.

Currently, with the rapid growth of online media, more people are obtaining information from it. However, traditional hotspot mining algorithms cannot achieve precise and fast control of hot topics. Aiming at the problem of poor accuracy and timeliness in current news media hotspot mining methods, this paper proposes a hotspot mining method based on the co-occurrence word model. First, a new co-occurrence word model based on word weight is proposed. Then, for key phrase extraction, a hotspot mining algorithm based on the co-occurrence word model and improved smooth inverse frequency rank (SIFRANK) is designed. Finally, the Spark computing framework is introduced to improve the computing efficiency. The experimental outcomes expresses that the new word discovery algorithm discovered 16871 and 17921 new words in the Weibo Short News and Weibo Short Text datasets respectively. The heat weight values of the keywords obtained by the improved SIFRANK reaches 0.9356, 0.9991, and 0.6117. In the Covid19 Tweets dataset, the accuracy is 0.6223, the recall is 0.7015, and the F1 value is 0.6605. In the President-elects Tweets dataset, the accuracy is 0.6418, the recall is 0.7162, and the F1 value is 0.6767. After applying the Spark computing framework, the running speed has significantly improved. The text mining news media hotspot mining method based on the co-occurrence word model proposed in this study has improved the accuracy and efficiency of mining hot topics, and has great practical significance.

Comparative analysis of machine learning algorithms for biomedical text document classification: A case study on cancer-related publications

Biomedical text document classification is an essential task within Natural Language Processing (NLP), with applications ranging from sentiment analysis to authorship identification. Despite advancements in traditional machine-learning algorithms like Support Vector Machines (SVM) and Logistic Regression, challenges such as data sparsity and high dimensionality persist. Recent years have seen a surge in the use of deep learning models to mitigate these issues. This study aims to conduct a comparative analysis of various machine-learning algorithms for classifying biomedical text documents. The study employs the "Medical Text Dataset - Cancer Doc Classification" from Kaggle, comprising 7570 biomedical text documents labeled into three types of cancer (colon, lung, and thyroid). A preprocessing pipeline involving tokenization, stop-word removal, and Term Frequency-Inverse Document Frequency (TF-IDF) vectorization is applied. Algorithms including Logistic Regression, SVM, and Multinomial Naive Bayes are evaluated through 5-fold cross-validation. Performance metrics like accuracy, precision, recall, F1 score, and area under the ROC curve (AUC ROC) are employed. Logistic Regression outperforms the other algorithms with an accuracy of 78.3% and an AUC ROC of 88.59%. SVM and Multinomial Naive Bayes follow with lower performance metrics. Hyperparameter tuning further enhances the performance of the algorithms, particularly Logistic Regression. The study makes a significant contribution to the field of biomedical text classification by systematically comparing machine-learning algorithms. Logistic Regression emerges as the most effective, emphasizing the importance of algorithm selection and hyperparameter tuning in machine learning applications within this domain.

War, emotions, mental health, and artificial intelligence.

During the war time dysregulation of negative emotions such as fear, anger, hatred, frustration, sadness, humiliation, and hopelessness can overrule normal societal values, culture, and endanger global peace and security, and mental health in affected societies. Therefore, it is understandable that the range and power of negative emotions may play important roles in consideration of human behavior in any armed conflict. The estimation and assessment of dominant negative emotions during war time are crucial but are challenged by the complexity of emotions' neuro-psycho-physiology. Currently available natural language processing (NLP) tools have comprehensive computational methods to analyze and understand the emotional content of related textual data in war-inflicted societies. Innovative AI-driven technologies incorporating machine learning, neuro-linguistic programming, cloud infrastructure, and novel digital therapeutic tools and applications present an immense potential to enhance mental health care worldwide. This advancement could make mental health services more cost-effective and readily accessible. Due to the inadequate number of psychiatrists and limited psychiatric resources in coping with mental health consequences of war and traumas, new digital therapeutic wearable devices supported by AI tools and means might be promising approach in psychiatry of future. Transformation of negative dominant emotional maps might be undertaken by the simultaneous combination of online cognitive behavioral therapy (CBT) on individual level, as well as usage of emotionally based strategic communications (EBSC) on a public level. The proposed positive emotional transformation by means of CBT and EBSC may provide important leverage in efforts to protect mental health of civil population in war-inflicted societies. AI-based tools that can be applied in design of EBSC stimuli, like Open AI Chat GPT or Google Gemini may have great potential to significantly enhance emotionally based strategic communications by more comprehensive understanding of semantic and linguistic analysis of available text datasets of war-traumatized society. Human in the loop enhanced by Chat GPT and Gemini can aid in design and development of emotionally annotated messages that resonate among targeted population, amplifying the impact of strategic communications in shaping human dominant emotional maps into a more positive by CBT and EBCS.

A Study of the Impact of Internet Technology on Minority Languages and Cultures

Abstract Network technology influences and transforms the language use, environment, and cultural practices of ethnic minorities. This paper designs a text classification model for minority languages and cultures, utilizing BERT for word vector modeling, and introduces LSTM and attention mechanisms to capture distant semantic information, solve the gradient vanishing problem, and learn contextual details. The BiLSTM multi-head attention mechanism is then used to filter important semantic features of the input sequence, and finally, all extracted features are linked to complete the text classification and output the categories of minority language and culture texts. The model is applied to text datasets from ethnic minority forums and their news video datasets. It is found that the proportion of traditional ethnic minority vocabulary in forums decreased from 14.6% in 2004 to 10.7% in 2024. Conversely, the ratios of vocabulary related to consumption, marriage, employment, and education increased, with the ratio of marriage-related vocabulary rising from 0.33 to 0.94. This suggests that, despite the influence of Internet technology, traditional cultural concepts of marriage among ethnic minorities remain widely observed. This study provides a feasible solution for categorizing texts in minority languages and offers a unique reflection on the influence of the Internet on the cultural changes of minority languages.