Depression Detection on Mandarin Text through Bert Model

This study introduces a novel methodology for classifying engine sounds into distinct categories, including Exhaust Leak, Rod Knock, and Lifter Tick, utilizing advanced signal processing and machine learning techniques. Despite diligent efforts, the classification model achieved a modest accuracy of 60%. One contributing factor to this performance could be the limited size and diversity of the training dataset, potentially failing to capture the full range of real-world engine sound variations. Additionally, the feature representation and selection process may not have fully captured the unique characteristics of each engine sound category. The classification model was developed using a Convolutional Neural Network (CNN) architecture. The CNN model comprises multiple convolutional layers followed by fully connected layers, enabling spatial feature extraction from spectrogram-based representations of engine sound recordings. Hyperparameters such as the number of convolutional filters, filter sizes, dropout rates, and learning rate were fine-tuned through empirical testing and cross-validation. To enhance training effectiveness, 500 files per class were utilized for training data, while 50 files per class were reserved for testing purposes. This expanded dataset aims to provide a more comprehensive representation of engine sound variations across different categories. Future research directions could involve augmenting the training dataset with a wider range of samples and implementing advanced feature engineering techniques like deep feature learning or transfer learning to extract more informative representations from raw audio data. Additionally, ensemble learning methods that combine multiple classifiers or models could boost classification performance by leveraging diverse insights. Furthermore, integrating domain-specific knowledge and expert annotations into the model training process may enhance its ability to generalize across various engine sound scenarios. In conclusion, while the CNN model presented in this study shows promise for engine sound classification, there is potential for improvement in model architecture, feature representation, dataset augmentation, and domain knowledge integration. Addressing these aspects could lead to more accurate automotive diagnostic systems capable of identifying mechanical issues based on acoustic signatures, thereby enhancing vehicle safety and reliability.

BackgroundThe COVID-19 pandemic has created a global health crisis that is affecting economies and societies worldwide. During times of uncertainty and unexpected change, people have turned to social media platforms as communication tools and primary information sources. Platforms such as Twitter and Sina Weibo have allowed communities to share discussion and emotional support; they also play important roles for individuals, governments, and organizations in exchanging information and expressing opinions. However, research that studies the main concerns expressed by social media users during the pandemic is limited.ObjectiveThe aim of this study was to examine the main concerns raised and discussed by citizens on Sina Weibo, the largest social media platform in China, during the COVID-19 pandemic.MethodsWe used a web crawler tool and a set of predefined search terms (New Coronavirus Pneumonia, New Coronavirus, and COVID-19) to investigate concerns raised by Sina Weibo users. Textual information and metadata (number of likes, comments, retweets, publishing time, and publishing location) of microblog posts published between December 1, 2019, and July 32, 2020, were collected. After segmenting the words of the collected text, we used a topic modeling technique, latent Dirichlet allocation (LDA), to identify the most common topics posted by users. We analyzed the emotional tendencies of the topics, calculated the proportional distribution of the topics, performed user behavior analysis on the topics using data collected from the number of likes, comments, and retweets, and studied the changes in user concerns and differences in participation between citizens living in different regions of mainland China.ResultsBased on the 203,191 eligible microblog posts collected, we identified 17 topics and grouped them into 8 themes. These topics were pandemic statistics, domestic epidemic, epidemics in other countries worldwide, COVID-19 treatments, medical resources, economic shock, quarantine and investigation, patients’ outcry for help, work and production resumption, psychological influence, joint prevention and control, material donation, epidemics in neighboring countries, vaccine development, fueling and saluting antiepidemic action, detection, and study resumption. The mean sentiment was positive for 11 topics and negative for 6 topics. The topic with the highest mean of retweets was domestic epidemic, while the topic with the highest mean of likes was quarantine and investigation.ConclusionsConcerns expressed by social media users are highly correlated with the evolution of the global pandemic. During the COVID-19 pandemic, social media has provided a platform for Chinese government departments and organizations to better understand public concerns and demands. Similarly, social media has provided channels to disseminate information about epidemic prevention and has influenced public attitudes and behaviors. Government departments, especially those related to health, can create appropriate policies in a timely manner through monitoring social media platforms to guide public opinion and behavior during epidemics.

The prevalence of depression has increased significantly over the past few years both in developed and developing countries. However, many people with symptoms of depression still remain untreated or undiagnosed. Social media may be a tool to help researchers and clinicians to identify and support individuals who experience depression. More than 394,000,000 postings were collected from China’s most popular social media website, Sina Weibo. 1000 randomly selected depression-related postings was coded and analyzed to learn the themes of these postings, and a text classifier was built to identify the postings indicating depression. The identified depressed users were compared with the general population on demographic characteristics, diurnal patterns, and patterns of emoticon usage. We found that disclosure of depression was the most popular theme; depression displayers were more engaged with social media compared to non-depression displayers, the depression postings showed geographical variations, depression displayers tended to be active during periods of leisure and sleep, and depression displayers used negative emoticons more frequently than non-depression displayers. This study offers a broad picture of depression references on China’s social media, which may be cost effectively developed to detect and help individuals who may suffer from depression disorders.

Dementia is a syndrome chronic or progressive that usually affects the cognitive functioning of the subjects. Alzheimer’s, a neurodegenerative disorder, is the leading cause of dementia. One of the many symptoms of Alzheimer’s Dementia is the inability to speak and understand language clearly. The last decade has seen a surge in the research done in Alzheimer’s Dementia detection using Linguistics and acoustic features. This paper takes up the Alzheimer’s Dementia classification task of ADReSS INTERSPEECH-2020 challenge, ”Alzheimer’s Dementia Recognition through Spontaneous Speech: The ADReSS Challenge”. It uses eight different acoustic features to find the attributes in the human speech production system (vocal track and excitation source) affected by Alzheimer’s Dementia. In this study, the Alzheimer’s dementia classification is performed using five different Machine Learning models using ADReSS INTERSPEECH-2020 challenge dataset. Since most of the studies in the previous literature have used linguistic features successfully for Alzheimer’s dementia classification, the current study also demonstrates the performance of the BERT model for the dementia classification task. The maximum accuracy obtained by the acoustic feature is 64.5%, and the BERT Model provides a classification accuracy of 79.1% over the test dataset. Finally, the score-level fusion of the acoustic model with the BERT Model shows an improvement of 6.1% classification accuracy over the BERT Model, which indicates the complementary nature of acoustic features to linguistic features.

Introduction. In today’s globalized world, social media has become an integral part of daily life, with most people being active on at least one platform. It is no surprise that media companies have recognized the value of social media as a powerful tool for information dissemination and audience engagement. In particular, social media platforms are actively used by media outlets that disseminate tourism-related information. Thanks to their large audience, quick access to information, and lightning-fast speed of dissemination, social media is becoming an indispensable platform for promoting tourist attractions, popularizing cultural heritage, and attracting tourists. Relevance of the study. Understanding how tourism media can effectively operate within digital platforms is essential for increasing the country’s tourism competitiveness. The aim of the article is to determine the features of the transformation of traditional Chinese tourism media in the era of social media, assess their achievements and challenges on domestic and international platforms, and analyze the publicity effects and challenges these traditional tourism media face in the context of globalization. Methodology. A critical analysis of existing literature was conducted to assess the level of understanding of the issue within global journalism studies. Through comparative analysis, we examined the distinct characteristics of various platforms and their influence on the transformation strategies of traditional Chinese tourism media. Additionally, by analyzing several representative cases of Chinese travel media, we explored the key factors contributing to successful transformation as well as the major challenges encountered in the process. Results. Given the specifics of the national policy that applies to China’s media system, it is worth distinguishing between domestic and global social media platforms. For example, Facebook, X, Instagram, YouTube, and TikTok are outside the Chinese media market. In China, the platforms WeChat, Weibo, RED, Bilibili, and Douyin are widespread. An analysis of traditional Chinese media pages revealed that some actively use international digital platforms to distribute content in addition to domestic Chinese social media. For example, media iPanda. By consistently delivering high-quality panda-related content on various digital platforms, including international social media, iPanda creates a strong emotional connection with its audience. Conclusions. This study offers a new perspective on the transformation of traditional Chinese tourism media in the era of social media by conducting a cross-platform comparative analysis and combining quantitative and qualitative methods. Unlike previous research, it specifically examines the adaptation of tourism media materials to global digital platforms, identifying key challenges, opportunities, and best practices. The findings provide practical recommendations for optimizing content dissemination through AI, big data, and multilingual strategies, offering valuable insights for media organizations and researchers.

This study aims to verify the wood classification performance of convolutional neural networks (CNNs), such as VGG16, ResNet50, GoogLeNet, and basic CNN architectures, and to investigate the factors affecting classification performance. A dataset from 10 softwood species consisted of 200 cross-sectional micrographs each from the total part, earlywood, and latewood of each species. We used 80% and 20% of each dataset for training and testing, respectively. To improve the performance of the architectures, the dataset was augmented, and the differences in classification performance before and after augmentation were compared. The four architectures showed a high classification accuracy of over 90% between species, and the accuracy increased with increasing epochs. However, the starting points of the accuracy, loss, and training speed increments differed according to the architecture. The latewood dataset showed the highest accuracy. The epochs and augmented datasets also positively affected accuracy, whereas the total part and non-augmented datasets had a negative effect on accuracy. Additionally, the augmented dataset tended to derive stable results and reached a convergence point earlier. In the present study, an augmented latewood dataset was the most important factor affecting classification performance and should be used for training CNNs.

Natural language processing (NLP) has been widely applied in machine translation, speech recognition and other fields. Many elegant results have been proposed such as text classification, sentiment analysis. In recent years, the Bert model has made a great breakthrough in many NLP tasks. Afterwards, using transformer as a feature extractor has been achieving good performance in many scenarios. However, there are too many model parameters in the Bert model. In order to solve this problem, we propose an improved algorithm for the Bert model, which shares the first 11 layers' parameters. The proposed method could largely reduce the number of parameters in the model. We also conduct experiments to evaluate the algorithm's performance; the results also validate that the improved algorithm for the Bert model could achieve good results.

ChatGPT OpenAI, a large-language chatbot model, has gained a lot of attention due to its popularity and impressive performance in many natural language processing tasks. ChatGPT produces superior answers to a wide range of real-world human questions and generates human-like text. The new OpenAI ChatGPT technology may have some strengths and weaknesses at this early stage. Users have reported early opinions about the ChatGPT features, and their feedback is essential to recognize and fix its shortcomings and issues. This study uses the ChatGPT tweets Arabic dataset to automatically find user opinions and sentiments about ChatGPT technology. The dataset is preprocessed and labeled using the TextBlob Arabic Python library into positive, negative, and neutral tweets. Despite extensive works for the English language, languages like Arabic are less studied regarding tweet analysis. Existing literature about Arabic tweet sentiment analysis has mainly focused on machine learning and deep learning models. We collected a total of 27,780 unstructured tweets from Twitter using the Tweepy SNscrape Python library using various hash-tags such as # Chat-GPT, #OpenAI, #Chatbot, Chat-GPT3, and so on. To enhance the model’s performance and reduce computational complexity, unstructured tweets are converted into structured and normalized forms. Tweets contain missing values, URL and HTML tags, stop words, punctuation, diacritics, elongations, and numeric values that have no impact on the model performance; hence, these increase the computational cost. So, these steps are removed with the help of Python preprocessing libraries to enhance text quality and consistency. This study adopts Transformer-based models such as RoBERTa, XLNet, and DistilBERT that automatically classify the tweets. Additionally, a hybrid transformer-based model is proposed to obtain better results. The proposed hybrid model is developed by combining the hidden outputs of the RoBERTA and BERT models using a concatenation layer, then adding dense layers with “Relu” activation employed as a hidden layer to create non-linearity and a “softmax” activation function for multiclass classification. They differ from existing state-of-the-art models due to the enhanced capabilities of both models in text classification. Hybrid models combine the different models to make accurate predictions and reduce bias and enhanced the overall results, while state-of-the-art models are incapable of making accurate predictions. Experiments show that the proposed hybrid model achieves 96.02% accuracy, 100% precision on negative tweets, and 99% recall for neutral tweets. The performance of the proposed model is far better than existing state-of-the-art models.

Purpose Entrepreneurs seek to increase their influence and foster consumer engagement by generating content on social media platforms that expresses certainty. However, the impact of the expressive certainty of entrepreneur-generated content (EGC) on consumer engagement is unclear. Through a field study and controlled experiment, this study aims to reveal how the expressive certainty of EGC affects consumer engagement. Design/methodology/approach The proposed research model was empirically tested by analysing the content generated by 220 entrepreneurs on Sina Weibo over the past four years and conducting a randomised controlled experiment in the laboratory. Negative binomial regression and an analysis of variance were used to examine the proposed research model. Findings The results show that the expressive certainty of EGC increases consumer engagement by influencing consumers’ perceptions of leadership. Specifically, EGC with a higher level of expressive certainty makes entrepreneurs appear more confident and powerful and leads consumers to perceive them as stronger leaders, thereby positively influencing consumer engagement. In addition, EGC with a higher level of certainty stimulates more positive consumer engagement when generated via mobile devices than when generated via non-mobile devices. The authors also found that the positive effect of perceived leadership on consumer engagement is moderated by power distance. Among consumers with high power-distance, the indirect effect of expressive certainty on consumer engagement is stronger. Research limitations/implications This study is subject to several constraints that could be solved through future research. Firstly, the field study only focuses on Sina Weibo, a well-known Chinese social media platform and the experimental study focuses only on Twitter, which may limit the generalisation of this research’s results to other social media platforms. Secondly, a more sophisticated approach to data analysis could be considered for future research to ascertain some of the most effective certainty words in EGC, such as text mining and machine learning. Finally, future research could examine the impact of EGC’s expressive certainty on other significant marketing effectiveness variables, such as consumer attitudes and brand loyalty. Practical implications The results suggest that using certainty in content is a way to help entrepreneurs project a high leadership profile. The positive impact of expressive certainty on perceived leadership suggests that entrepreneurs should use more certainty expressions on social media and other communication channels (e.g. brand launches) to make themselves appear more confident and influential, which is beneficial to stimulate consumer engagement with EGC. Specifically, entrepreneurs should generate more EGC with a higher level of certainty via mobile devices, as doing so will increase consumers’ perceptions of authenticity and trust in certainty information, increasing their engagement. Originality/value This study provides a valuable complement to previous research on information dissemination and consumer engagement on social media by exploring the mechanisms by which the expressive certainty of EGC and generating devices affects consumer engagement from the perspective of perceived leadership and indicates practical guidelines for developing EGC on social media.

Background: The spread of rumors related to COVID-19 on social media has posed substantial challenges to public health governance, and thus exposing rumors and curbing their spread quickly and effectively has become an urgent task. This study aimed to assist in formulating effective strategies to debunk rumors and curb their spread on social media.Methods: A total of 2,053 original postings and 100,348 comments that replied to the postings of five false rumors related to COVID-19 (dated from January 20, 2020, to June 28, 2020) belonging to three categories, authoritative, social, and political, on Sina Weibo in China were randomly selected. To study the effectiveness of different debunking methods, a new annotation scheme was proposed that divides debunking methods into six categories: denial, further fact-checking, refutation, person response, organization response, and combination methods. Text classifiers using deep learning methods were built to automatically identify four user stances in comments that replied to debunking postings: supporting, denying, querying, and commenting stances. Then, based on stance responses, a debunking effectiveness index (DEI) was developed to measure the effectiveness of different debunking methods.Results: The refutation method with cited evidence has the best debunking effect, whether used alone or in combination with other debunking methods. For the social category of Car rumor and political category of Russia rumor, using the refutation method alone can achieve the optimal debunking effect. For authoritative rumors, a combination method has the optimal debunking effect, but the most effective combination method requires avoiding the use of a combination of a debunking method where the person or organization defamed by the authoritative rumor responds personally and the refutation method.Conclusion: The findings provide relevant insights into ways to debunk rumors effectively, support crisis management of false information, and take necessary actions in response to rumors amid public health emergencies.

ABSTRACTThe purpose of this study is to evaluate hotel companies’ social media marketing efforts using the “6Is” social media monitoring framework. More specifically, in this study we set out to understand how hotels disseminate information, engage with consumers, and distribute products through social media. We investigate the presence of the top 133 hotel brands in China on the top domestic Chinese social media sites in order to identify to what extent the hotel industry in China recognizes the potential of this marketing tool. The findings indicate that the majority of the hotels actively and consistently use the Sina Weibo and Wechat platforms for their social media marketing.

DeepSeek has emerged as a prominent representative of China’s domestic large language models(LLMs), attracting widespread public attention and generating diverse discussions on social media since its release. To systematically examine public sentiment and thematic concerns surrounding DeepSeek, this study analyzes 86,008 Weibo posts collected between January 19 and March 22, 2025. After data cleaning, 59,679 valid entries were retained. A mixed-method approach was adopted, combining Latent Dirichlet Allocation (LDA) for topic modeling and a fine-tuned BERT-based classifier for sentiment analysis. The BERT model, fine-tuned on 10,000 hand-labeled posts, achieved an accuracy of 82% and was subsequently applied to the remaining unlabeled data. Topic modeling revealed nine major themes in public discourse, including chip innovation and industrial transformation, smart device ecosystem, and China–US technological competition. Sentiment analysis showed that 46.3% of the posts expressed a positive attitude, reflecting public endorsement of DeepSeek’s domestic innovation and application potential, although concerns about information security, technical feasibility, and international political dynamics were also evident. Additionally, major events significantly influenced emotional fluctuations, with peaks in public attention observed on January 27 and March 19. These findings offer empirical insights for policymakers, developers, and the broader public into public acceptance, key concerns, and potential risks associated with emerging AI technologies, and support the formulation of informed strategies for technology governance, user engagement, and product optimization.

Text classification is a procedure of categorizing a text into classes or predefined labels, allowing effective analysis and organizing the large volume of textual data. In the growth of online textual data like news, digital platforms, or social media, an effective model for text summarization and classification. However, traditional models of text classification had difficulty capturing deep semantic dependencies and contextual embeddings among texts or words, which led to misclassification and minimized classification accuracy. To overcome these challenges, the research proposed Token Level Class Importance (TLCI) and Layer Wise Learning Rate Decay (LWLRD) named as TLLRD is integrated with Bidirectional Encoder Representations from Transformers (BERT) named as (TLLRD-BERT), which is utilized for multiclass violence text classification and enhanced classification performance. The model integrates LLRD to stabilize fine-tuning across transformer layers and TLCI to emphasize semantically significant words associated with aggressive or violent expressions. The proposed TLLRD-BERT attains an accuracy of 98.24%, 96.54% and 99.71% through YouTube comments dataset, the news headline dataset and the SMS spam detection dataset, respectively.

With the rapid expansion of the Internet and social media, there has been an explosion of text data, containing valuable emotional insights. Sentiment analysis, aimed at discerning and processing people's emotional inclinations, opinions, and perspectives, has emerged as a pivotal area within natural language processing (NLP). This research explores the application of Deep Learning-based BERT models, particularly DistilBERT, in sentiment analysis. It highlights the importance of sentiment analysis in understanding emotional insights from vast text data and introduces BERT models as revolutionary methodologies for enhancing accuracy and efficiency. The study conducts experiments using the SST2 dataset, showcasing the effectiveness of DistilBERT in sentiment classification tasks. Overall, it underscores the transformative potential of BERT models in revolutionizing sentiment analysis methodologies and driving advancements in natural language processing research.

6 - Social media big data analysis for mental health research