Language Sentiment Research Articles

Language sentiment predicts changes in depressive symptoms

The prevalence of depression is a major societal health concern, and there is an ongoing need to develop tools that predict who will become depressed. Past research suggests that depression changes the language we use, but it is unclear whether language is predictive of worsening symptoms. Here, we test whether the sentiment of brief written linguistic responses predicts changes in depression. Across two studies (N = 467), participants provided responses to neutral open-ended questions, narrating aspects of their lives relevant to depression (e.g., mood, motivation, sleep). Participants also completed the Patient Health Questionnaire (PHQ-9) to assess depressive symptoms and a risky decision-making task with periodic measurements of momentary happiness to quantify mood dynamics. The sentiment of written responses was evaluated by human raters (N = 470), Large Language Models (LLMs; ChatGPT 3.5 and 4.0), and the Linguistic Inquiry and Word Count (LIWC) tool. We found that language sentiment evaluated by human raters and LLMs, but not LIWC, predicted changes in depressive symptoms at a three-week follow-up. Using computational modeling, we found that language sentiment was associated with current mood, but language sentiment predicted symptom changes even after controlling for current mood. In summary, we demonstrate a scalable tool that combines brief written responses with sentiment analysis by AI tools that matches human performance in the prediction of future psychiatric symptoms.

Personalized movie recommendations based on probabilistic linguistic sentiment and integrated DEMATEL-TODIM methods

With the development of society, online reviews are increasingly becoming a crucial factor in decision-making. Especially for entertainment products such as movies, they are preferred for their affordability and high entertainment factor. Therefore, this paper proposes a movie recommendation model that considers user personalization using a probabilistic linguistic approach based on online reviews. Firstly, the method constructs a quantitative sentiment framework that transforms comments into a multi-granular probabilistic sentiment language. Secondly, we build the decision-making trial and evaluation laboratory (DEMATEL) method for probabilistic linguistic environments to explore interrelationships between product attributes, and improve the distance measure and score function to better integrate probabilistic linguistic information into DEMATEL weight calculations. Furthermore, to account for risk preferences, the model employs the extended TODIM (an acronym in Portuguese for interactive and multicriteria decision making) methodology to determine the ranking of alternatives. Finally, we design Douban movie ranking experiments to demonstrate the validity of the model. Compared with other methods, this paper incorporates the emotional tendency of movie attributes and user preference into the decision-making process leading to more reasonable results.

Multi-modal Chinese Text Emotion Metaphor Computation Based on Mutual Information and Information Entropy

Metaphor is to express another thing through one thing, it is not only a rhetorical means, but also embodies a kind of analogical cognition and way of thinking of people. In recent years, metaphors have made more and more progress in Chinese language recognition, so their status has gradually risen. Chinese language sentiment analysis is a relatively difficult problem in language processing. Chinese language semantics is ambiguous and complex, so it is a great challenge to construct Chinese Emotion structure. In this paper, Chinese nouns and Chinese verbs are used as test samples to conduct metaphor test calculation. The test results of mutual information and information entropy show that for the sample metaphor calculation, the correct rate of individual words is less than 70%. The ratio between the macro-average of gerund words and the baseline fluctuates relatively large. The metaphor recognition of nouns is about 20% higher than the original test results and about 10% higher for verbs. It shows that under the action of mutual information and information entropy, the metaphor recognition performance of the basic structure of Chinese Emotion constructed from the perspective of metaphor is improved, and further illustrates the feasibility of mutual information and information entropy algorithm for building the basic structure.

Comparison of the performance of naive bayes and support vector machine in sirekap sentiment analysis with the lexicon-based approach

The general public often uses the SiRekap application to see the progress of the election and to provide critical statements. Policies made by the government have good and bad outcomes, and users end up leaving their reviews and ratings on the Google Play Store, where the app can be downloaded. These reviews can be collected and processed into useful information such as sentiment analysis using Naïve Bayes and Support Vector Machine methods. Both methods have differences during training and during evaluation. The difference in results from the various scenarios tested was not much different. When training the Support Vector Machine model is able to process comment data labeled with a lexicon 10% better than the Naïve Bayes model by looking at the results of the accuracy of the two models. During the accuracy evaluation process, the two models produce the same accuracy of 72%. Although both models get the same accuracy during the evaluation process, there are differences in precision, recall, and f1 score. The difference is that the Support Vector Machine model is 5% better for precision, 8% for recall, and 3% for f1-score compared to the Naïve Bayes model. This research is limited to only knowing the performance of two machine learning models, namely the use of naive bayes and svm by using a label lexicon. The results obtained can be improved for the better. Improving the evaluation results can be done by adding data or using text data augmentation and there is creation from experts related to language sentiment.

Taxonomy of an Enslaved Heart

“Taxonomy of an Enslaved Heart” opens up the figuration of heartache, so common to sentimental writings, to consider how it can signify anatomical pain as well. What does it mean to read figuratively—accepting that every instance of a heart broken or throbbing or heavy indexes emotional pain addressing the reader’s sympathy—and, at the same time, to literalize these instances, so that each one refers to a specific episode in the history of a circulatory system? This essay attempts to hold both in tension, even as they resist each other. Attending to texts by Harriet Jacobs, Mary Prince, Sojourner Truth, and James Baldwin, the essay argues for what it calls the story of the heart: a minoritized account of pain that deforms sentimental language to register at once somatically, mentally, and intersubjectively. Because of its insecure legibility, the story of the heart subverts the biopolitical logic of legitimacy that traps many patients who are Black, disabled, or both today. What emerges from holding figuration with literalization subtly shifts the illnesses we know and the conditions by which we know them.

Natural language sentiment as an indicator of depression and anxiety symptoms: a longitudinal mixed methods study1

ABSTRACT The study tested how the use of positive- (e.g. beautiful) and negative-valenced (e.g. horrible) words in natural language and its change in time affects the severity of depression and anxiety symptoms among depressed and non-depressed individuals. This longitudinal mixed methods study (N = 40 participants, n = 1440 narratives) with three measurements within a year showed that at the between-person level the use of negative-valenced words was strongly associated with the increase in anxiety and depression symptoms over time while the use of positive-valenced words was slightly associated with the decrease in anxiety and depression symptom. These effects were not supported for within-person level (i.e. changes in word usage). No significant differences were observed in the effects between depressed and non-depressed groups. Summing up, the overall use of positive- and negative-valenced words (particularly negative-valenced words) had a stronger effect on the severity of psychopathological symptoms than their change over time. The results were discussed in the context of natural language processing and its application in diagnosing depression and anxiety symptoms.

A multimodal approach to cross-lingual sentiment analysis with ensemble of transformer and LLM

Sentiment analysis is an essential task in natural language processing that involves identifying a text’s polarity, whether it expresses positive, negative, or neutral sentiments. With the growth of social media and the Internet, sentiment analysis has become increasingly important in various fields, such as marketing, politics, and customer service. However, sentiment analysis becomes challenging when dealing with foreign languages, particularly without labelled data for training models. In this study, we propose an ensemble model of transformers and a large language model (LLM) that leverages sentiment analysis of foreign languages by translating them into a base language, English. We used four languages, Arabic, Chinese, French, and Italian, and translated them using two neural machine translation models: LibreTranslate and Google Translate. Sentences were then analyzed for sentiment using an ensemble of pre-trained sentiment analysis models: Twitter-Roberta-Base-Sentiment-Latest, bert-base-multilingual-uncased-sentiment, and GPT-3, which is an LLM from OpenAI. Our experimental results showed that the accuracy of sentiment analysis on translated sentences was over 86% using the proposed model, indicating that foreign language sentiment analysis is possible through translation to English, and the proposed ensemble model works better than the independent pre-trained models and LLM.

Lexicon‐based fine‐tuning of multilingual language models for low‐resource language sentiment analysis

AbstractPre‐trained multilingual language models (PMLMs) such as mBERT and XLM‐R have shown good cross‐lingual transferability. However, they are not specifically trained to capture cross‐lingual signals concerning sentiment words. This poses a disadvantage for low‐resource languages (LRLs) that are under‐represented in these models. To better fine‐tune these models for sentiment classification in LRLs, a novel intermediate task fine‐tuning (ITFT) technique based on a sentiment lexicon of a high‐resource language (HRL) is introduced. The authors experiment with LRLs Sinhala, Tamil and Bengali for a 3‐class sentiment classification task and show that this method outperforms vanilla fine‐tuning of the PMLM. It also outperforms or is on‐par with basic ITFT that relies on an HRL sentiment classification dataset.

Manager’s Perception, Accounting Training and Chinese Tone

The purpose of this paper is to measure if the undergraduates majored in accounting could more clear identification about management perception embedded in the Chinese financial reports. After comprehensive research designs and detailed questionnaires, this study finds that the more solid the accounting professional training, the more clearly the students interpret the financial statements, and the less likely they are to be disturbed by the language sentiment (tone) embedded in financial disclosures. Our findings contribute the importance of the socialized accounting education, which could minimize the level of misunderstanding by uninformed investors, and so do managers hold opportunism.

Multinomial Naive Bayes Algorithm for Indonesian language Sentiment Classification Related to Jakarta International Stadium (JIS)

The research focuses on analysing public evaluations, particularly those on Google Maps, about the Jakarta International Stadium (JIS). The study aims to employ the multinomial Naive Bayes algorithm to ascertain the sentiment expressed in these reviews. The objective of this study was to employ the multinomial Naive Bayes method to analyse the reviews on Google Maps pertaining to the Jakarta International Stadium (JIS). The utilised data consists of 2971 public reviews on Google Maps specifically pertaining to Jakarta International Stadium (JIS). These reviews were acquired through web scraping using a data miner. The acquired data is next processed in the text preparation phase to generate a prepared dataset suitable for analysis. This preprocessing stage includes operations such as casefolding, stopword removal, tokenizing, and stemming. The study yielded an accuracy of 0.83, or 83%, when tested on 733 data points. Out of these, 292 positive data points were correctly anticipated, while 59 positive data points were incorrectly forecast. Additionally, 317 negative data points were correctly predicted, while 65 negative data points were incorrectly predicted. The conducted modelling is subsequently categorised using a novel dataset of 161 review data points, with the objective of discerning the sentiment expressed within the dataset. The analysis of the new dataset yielded 101 reviews with positive sentiment and 50 reviews with negative sentiment.

Are Emotions Conveyed Across Machine Translations? Establishing an Analytical Process for the Effectiveness of Multilingual Sentiment Analysis with Italian Text


 
 
 Abstract Natural language processing (NLP) is being widely used globally for a variety of value-creation tasks ranging from chat-bots and machine translations to sentiment and topic analysis and multilingual large language models (LLMs). However, most of the advances are initially implemented within the English language framework, and it takes time and resources to develop comparable resources in other languages. The advances in machine translations have enabled the rapid and effective conversion of content in global languages into English and vice-versa. This creates potential opportunities to apply English language NLP methods and tools to other languages via machine translations. However, although this idea is powerful, it needs to be validated and processes and best practices need to be developed and kept updated. The present research is an effort to contribute to the development of best practices and an evaluation framework. We present a systematic and repeatable state-of-the-art process to evaluate the viability of applying English language sentiment analysis tools to Italian text by using multiple English language machine translation mechanisms such that it can be easily extended to other languages.
 
 

Critical analysis of the construction of Turkish national identity

This paper investigates the relationship between argumentation and political discourse. More specifically, this study focuses on the discursive and argumentative strategies employed by one of the most prominent Turkish leaders and current ruling president, Recep Tayyip Erdoğan. The aim is to shed light on linguistic elements that fuel the sense of national identity after the coup attempt of 2016 through employing the Discourse-Historical Approach (DHA) for unveiling the use of specific argumentative strategies in the translated version of Erdogan’s 2016 Republic Day speech. English's growing presence in politics highly influences translation practices (House 2013). Not all politicians are proficient in English, thus, their speeches undergo translation from their native language. These translated speeches are intended for a global audience, which underscores the pivotal role of English analysis in both political discourse and the field of translation studies. Although the original language of the speech is Turkish, the examination of its English translation serves to highlight the significance of analysing translated political speeches and making these analyses accessible to a wider audience. The findings emphasise the importance of argumentative topics (topoi), such as the topos of History, Definition, Comparison, Threat and Favoruable Time in the advocation of national sentiment at various linguistic and structural levels within the translated political speech of Erdoğan. The implications of this research extend to the understanding of the intricate interplay between argumentative language and concepts of national sentiment within political discourse. Future research can further explore the application ofDHA in the analysis of political discourse, allowing for deeper insights into the communicative strategies employed by political leaders.

Quantum computing and machine learning for Arabic language sentiment classification in social media

With the increasing amount of digital data generated by Arabic speakers, the need for effective and efficient document classification techniques is more important than ever. In recent years, both quantum computing and machine learning have shown great promise in the field of document classification. However, there is a lack of research investigating the performance of these techniques on the Arabic language. This paper presents a comparative study of quantum computing and machine learning for two datasets of Arabic language document classification. In the first dataset of 213,465 Arabic tweets, both classic machine learning (ML) and quantum computing approaches achieve high accuracy in sentiment analysis, with quantum computing slightly outperforming classic ML. Quantum computing completes the task in approximately 59 min, slightly faster than classic ML, which takes around 1 h. The precision, recall, and F1 score metrics indicate the effectiveness of both approaches in predicting sentiment in Arabic tweets. Classic ML achieves precision, recall, and F1 score values of 0.8215, 0.8175, and 0.8121, respectively, while quantum computing achieves values of 0.8239, 0.8199, and 0.8147, respectively. In the second dataset of 44,000 tweets, both classic ML (using the Random Forest algorithm) and quantum computing demonstrate significantly reduced processing times compared to the first dataset, with no substantial difference between them. Classic ML completes the analysis in approximately 2 min, while quantum computing takes approximately 1 min and 53 s. The accuracy of classic ML is higher at 0.9241 compared to 0.9205 for quantum computing. However, both approaches achieve high precision, recall, and F1 scores, indicating their effectiveness in accurately predicting sentiment in the dataset. Classic ML achieves precision, recall, and F1 score values of 0.9286, 0.9241, and 0.9249, respectively, while quantum computing achieves values of 0.92456, 0.9205, and 0.9214, respectively. The analysis of the metrics indicates that quantum computing approaches are effective in identifying positive instances and capturing relevant sentiment information in large datasets. On the other hand, traditional machine learning techniques exhibit faster processing times when dealing with smaller dataset sizes. This study provides valuable insights into the strengths and limitations of quantum computing and machine learning for Arabic document classification, emphasizing the potential of quantum computing in achieving high accuracy, particularly in scenarios where traditional machine learning techniques may encounter difficulties. These findings contribute to the development of more accurate and efficient document classification systems for Arabic data.

Efficient dynamic feature adaptation for cross language sentiment analysis with biased adversarial training

Fine-tuning a large multi-lingual pretrained language model demonstrates impressive results in cross-language understanding. However, it still suffers when the training and test data have different distributions owing to various languages and domains. On one hand, annotating target data for different languages or domains is time-consuming or infeasible. On the other hand, fine-tuning a large language model often incurs high computational costs. In this paper, we aim to develop an efficient and effective adaptation framework for cross-language sentiment analysis based on a fixed pretrained multi-lingual model. Specifically, we propose a Dynamic Feature Adaptation (DFA) module to fully leverage the features from different layers of the pretrained model such that its large backbone is not involved during adaptation training. Furthermore, we observe that traditional adversarial domain adaptation training could compromise the discriminative information of the model by pushing source and target features towards each other. The source features obtained with supervised training preserved the discriminability of the model, which should be less affected. Therefore, we propose a novel Biased Adversarial Training (BAT) method, that encourages only the target features towards source features. Extensive experimental results on various cross-lingual and cross-lingual-and-domain sentiment analysis tasks demonstrate the superiority of the proposed framework. Additionally, several ablation studies are conducted to validate the effectiveness of each proposed module.

Review on Sentiment Analysis of Marathi Language of Maharashtra

Abstract: Sentiment analysis on numerous Regional languages is performed, and classification algorithms based on Lexicon, Dictionary, and Machine Learning are employed. Because of the widespread usage of social media platforms, people are rapidly turning to the internet to find and discuss information, thoughts, opinions, feelings, perspectives, facts, and suggestions, resulting in a plethora of user-generated emotion enormous amounts of text data available for analysis. A large number of individuals in India express themselves in multiple languages, resulting in a massive amount of Natural Language Processing text data for (NLP) researchers. Sentiment Analysis (SA) of code-mixed text provides valuable information in politics, education, services marketing, business, health, sports, and other sectors. Work on Indian Language Sentiment Analysis Textual Data, particularly in Hindi, has gained steam in the previous decade in comparison to code-mixed Indian language text. However, due to a lack of language and vocabulary (linguistic and lexical) tools and annotated resources, the process of Sentiment Analysis of Regional Languages becomes very difficult. The goal of this research was to present a complete summary of the Sentiment Analysis of Regional languages, with a focus on code-mixed Regional languages.

Categorization of Arabic posts using Artificial Neural Network and hash features

Sentiment analysis is an important study topic with diverse application domains including social network monitoring and automatic analysis of the body of natural language communication. Existing research on sentiment analysis has already utilised substantial domain knowledge available online comprising users’ opinion in various areas such as business, education, and social media. There is however limited literature available on Arabic language sentiment analysis. Furthermore, datasets used in majority of these studies have poor classification. In the present study, we utilised a primary dataset comprising 2122 sentences and 15,331 words compiled from 206 publicly available online posts to perform sentiment classification by using advanced machine learning technique based on Artificial Neural Networks. Unlike lexicon-based techniques that suffer from low accuracy due to their computational nature and parameter configuration, Artificial Neural Networks were used to classify people opinion posts into three categories including conservative, reform and revolution, accompanied by multiple hasher vector size to benchmark the performance of the proposed model. Extensive simulation results indicated an accuracy of 93.33%, 100%, and 100% for the classification of conservation, reform, and revolutionary classes, respectively.

나홀로 여행객과 동반여행객의 온라인 리뷰 차이 분석: 토픽모델링 기법을 중심으로

Online reviews based on experiences serve as important sources of information for potential tourists and provide valuable insights for tourism marketers to understand the preferences of tourists. After COVID-19, external activities have been restricted, leading to an increase in the online market. Additionally, the trend of solo travel has increased due to the rise in single-person households. To understand the difference between solo travelers and accompanied travelers, we analyzed the sentiment of reviews written by TripAdvisor users from 2015 to 2020 based on the presence or absence of companions. As a result of the analysis, we found that solo travelers expressed positive sentiment towards topics such as ‘relaxing’, ‘commentary’, ‘knowledge’, ‘sightseeing’, ‘palace tour’, ‘scenery’, and ‘walking’, with no negative sentiment topics identified. On the other hand, we found that companion travelers expressed positive sentiment towards the topics of ‘entrance fee’, ‘memory’, and ‘palace tour’, while expressing negative sentiment towards the topics of ‘scenery’, ‘park’, and ‘family’. Based on these results, it was confirmed that reviews written by Seoul visitors differ in terms of topics and emotional expressions depending on the presence or absence of companions, with the exception of ‘palace tours’. Therefore, by distinguishing preferred travel behaviors and sentiment language, it is expected that an emotional marketing strategy can be established using customized target marketing and online reviews.

Investigating Social Media to Evaluate Emergency Medicine Physicians’ Emotional Well-being During COVID-19

Emergency medicine (EM) physicians experience tremendous emotional health strain, which has been exacerbated during COVID-19, and many have taken to social media to express themselves. To analyze social media content from academic EM physicians and resident physicians to investigate changes in content and language as indicators of their emotional well-being. This cross-sectional study used machine learning and natural language processing of Twitter posts from self-described academic EM physicians and resident physicians between March 2018 and March 2022. Participants included academic EM physicians and resident physicians with publicly accessible posts (at least 300 total words across the posts) from the US counties with the top 10 COVID-19 case burdens. Data analysis was performed from June to September 2022. Being an EM physician or resident physician who posted on Twitter. Social media content themes during the prepandemic period, during the pandemic, and across the phases of the pandemic were analyzed. Psychological constructs evaluated included anxiety, anger, depression, and loneliness. Positive and negative language sentiment within posts was measured. This study identified 471 physicians with a total of 198 867 posts (mean [SD], 11 403 [18 998] words across posts; median [IQR], 3445 [1100-11 591] words across posts). The top 5 prepandemic themes included free open-access medical education (Cohen d, 0.44; 95% CI, 0.38-0.50), residency education (Cohen d, 0.43; 95% CI, 0.37-0.49), gun violence (Cohen d, 0.37; 95% CI, 0.32-0.44), quality improvement in health care (Cohen d, 0.33; 95% CI, 0.27-0.39), and professional resident associations (Cohen d, 0.33; 95% CI, 0.27-0.39). During the pandemic, themes were significantly related to healthy behaviors during COVID-19 (Cohen d, 0.83; 95% CI, 0.77-0.90), pandemic response (Cohen d, 0.71; 95% CI, 0.65-0.77), vaccines and vaccination (Cohen d, 0.60; 95% CI, 0.53-0.66), unstable housing and homelessness (Cohen d, 0.40; 95% CI, 0.34-0.47), and emotional support for others (Cohen d, 0.40; 95% CI, 0.34-0.46). Across the phases of the pandemic, thematic content within social media posts changed significantly. Compared with the prepandemic period, there was significantly less positive, and concordantly more negative, language used during COVID-19. Estimates of loneliness, anxiety, anger, and depression also increased significantly during COVID-19. In this cross-sectional study, key thematic shifts and increases in language related to anxiety, anger, depression, and loneliness were identified in the content posted on social media by academic EM physicians and resident physicians during the pandemic. Social media may provide a real-time and evolving landscape to evaluate thematic content and linguistics related to emotions and sentiment for health care workers.

Protesting With Feeling in Puerto Rico: Twitter and El Verano Del 19

Communication research has extensively addressed the influence of social media on protest. We seek to add to this body of research by examining how sentiment contained in Twitter communication about protest can condition the public reach of this communication. Specifically, we are interested in whether Twitter communication couched in negative sentiments like anger, disgust, and fear will drive this reach more than communication couched in positive sentiment such as trust, joy, and surprise. We rely on sentiment analysis to examine roughly 30,000 tweets surrounding the summer of 2019 Protests in Puerto Rico. These protests centered on a scandal involving former Governor Ricardo Rossello that ultimately led to his resignation. The analysis required adapting an English language sentiment dictionary to Spanish. Our results suggest that protesters frequently tweeted with both positive and negative sentiment when calling for the governor’s resignation, but ultimately, those tweets couched in negative sentiment, when compared to those with positive sentiment, had the most reach. That said, those tweets including either positive or negative sentiment had more reach than those absent sentiment.

Swoon: A Poetics of Passing Out

Swoon: A Poetics of Passing Out