Sentiment Score Research Articles

Sentiment analysis of economic text: A lexicon‐based approach

AbstractWe propose an economic lexicon (EL) specifically designed for textual applications in economics. We construct the dictionary with two important characteristics: (1) to have a wide coverage of terms used in documents discussing economic concepts, and (2) to provide a human‐annotated sentiment score in the range [−1, 1]. We illustrate the use of the EL in the context of a simple sentiment measure and consider several applications in economics. The comparison to other lexicons shows that the EL is superior due to its wider coverage of domain relevant terms and its more accurate categorization of the word sentiment.

Understanding public opinion and discussion dynamics of digital humans on social media: an analysis of sentiment, themes and user characteristics

PurposeThe emergence of artificial intelligence-generated content (AIGC) technology has markedly enhanced the capabilities of digital human content generation and natural language processing, thus further advancing the development of digital humans. To enable enterprises and governments to effectively address the challenges and opportunities arising from the rapid development of digital humans, it is imperative to understand the public opinion and discussion dynamics of digital humans.Design/methodology/approachThis study initially analyzed the trends and distribution patterns of public attention to digital humans. By utilizing word cloud technology, we explored the primary focal points of public interest and conducted a topic analysis using latent Dirichlet allocation (LDA) techniques. Subsequently, content analysis was conducted on the popular application domains of digital humans. Finally, this study examined the influence of user characteristics on emotional scores toward digital humans and the presence of differences in focus across user groups.FindingsThe results indicate a sustained increase in public attention toward digital humans, accompanied by notable geographic disparities in the distribution of discussions. Discussions on Weibo are primarily focused on four domains, whereas areas within the digital human application domain that provoke widespread discussion include live streaming, service, cultural entertainment and digital avatars. Significant impacts of user characteristics on sentiment scores were observed, revealing divergent focal points of interest among different user groups toward digital humans.Originality/valueThrough the deep analysis of Weibo data, this study offers new insights into the digital human industry, enabling governments and businesses to understand industry trends and develop targeted digital human customization strategies based on customer characteristics.

A Natural Language Processing Approach for Sentiment Analysis of Hotel Reviews

Businesses rely heavily on review analysis of consumer feedback and reviews when making decisions. This research examines fifteen years' worth of hotel evaluations to learn more about the opinions of visitors and how they relate to review scores in various years and regions. After the dataset was cleansed to concentrate on reviews written in English, TextBlob was used to analyze sentiment and classify the reviews as positive, neutral, or negative. The results validate using sentiment analysis in conjunction with numerical ratings, which show a positive link between more excellent ratings and positive sentiment scores. A significant drop in sentiment scores following the 2008–2009 recession emphasizes how the global financial crisis affected the hospitality sector. Significant differences in guest feelings between cities, states, and nations are shown by geographic analysis, underscoring the necessity of region-specific approaches. Sentiment distribution and trends were revealed via intricate visualizations such as heat maps, box plots, scatter plots, and histograms. To improve guest happiness, the study uses an understanding of geographical variations. These results give hotel management a solid framework for raising customer satisfaction and loyalty, customizing marketing campaigns, and enhancing service quality. For deeper insights, future studies may investigate more sophisticated NLP techniques.

Analyzing User Reviews on Digital Detox Apps: A Text Mining and Sentiment Analysis Approach

ABSTRACTDue to the growing concerns around problematic smartphone use and its negative impact, there is a rising interest in digital detox. While many digital detox apps have been developed in recent years, there is still limited understanding of the long‐term effectiveness of digital detox applications and the attitude of people towards these apps. This study fills this gap by identifying the topics that people post in their reviews on the Google Play Store about digital detox apps and the emotion‐based sentiment of those reviews. A total of 3500 reviews of 25 digital detox apps were collected from the Google Play Store using a scraping tool called “Parsehub.” Data was analyzed using R studio. Sentiment analysis results suggest that positive sentiments dominated the data frame. “Trust” and “anticipation” were the two most expressed emotions in the reviews. Regression analysis confirmed that sentiment scores could explain the ratings of the apps. Through LDA topic modeling four major topics of the reviews were identified and are discussed in detail in the later section of the research paper. The findings of this study may help app developers and marketers improve digital detox apps so that people can learn and practice mindful smartphone use with the help of these apps. This study fills a gap in digital detox research by adopting a new methodological approach and procedure since it combines text mining, sentiment analysis (NRC Lexicon using Syuzhet package), regression analysis, and LDA topic modeling. To the best of our knowledge, this is the first study which uses this research approach in the context of digital detox apps.

Supervised Sentiment Analysis of COVID-19 Tweets: Insights into Public Reactions and Trends

In the modern digital era, social media platforms have experienced exponential growth, serving as a medium for individuals to share their emotions and opinions. Twitter, a prominent social media platform, hosts a substantial number of users who regularly express their thoughts through tweets. This research paper focuses on conducting sentiment analysis on tweets pertaining to the COVID-19 situation in India. By utilizing a lexicon-based approach using TextBlob, we determine sentiment scores for each tweet. These scores are then used to categorize the tweets into Positive, Negative, or Neutral sentiment categories. To further enhance the accuracy of sentiment classification, we employ the Naïve Bayes machine learning model, achieving an impressive accuracy rate of 87.65%. The sentiment analysis results obtained through this study provide valuable insights for governmental bodies and other organizations, enabling them to formulate effective policies in order to address similar issues in the future.

Chinese Public Attitudes and Opinions on Health Policies During Public Health Emergencies: Sentiment and Topic Analysis.

By the end of 2021, the new wave of COVID-19 sparked by the Omicron variant spread rapidly due to its highly contagious nature, affecting more than 170 countries worldwide. Nucleic acid testing became the gold standard for diagnosing novel coronavirus infections. As of July 2022, numerous cities and regions in China have implemented regular nucleic acid testing policies, which have had a significant impact on socioeconomics and people's lives. This policy has garnered widespread attention on social media platforms. This study took the newly issued regular nucleic acid testing policy during the COVID-19 pandemic as an example to explore the sentiment responses and fluctuations of netizens toward new policies during public health emergencies. It aimed to propose strategies for managing public opinion on the internet and provide recommendations for policy making and public opinion control. We collected blog posts related to nucleic acid testing on Weibo from April 1, 2022, to July 31, 2022. We used the topic modeling technique latent Dirichlet allocation (LDA) to identify the most common topics posted by users. We used Bidirectional Encoder Representations from Transformers (BERT) to calculate the sentiment score of each post. We used an autoregressive integrated moving average (ARIMA) model to examine the relationship between sentiment scores and changes over time. We compared the differences in sentiment scores across various topics, as well as the changes in sentiment before and after the announcement of the nucleic acid price reduction policy (May 22) and the lifting of the lockdown policy in Shanghai (June 1). We collected a total of 463,566 Weibo posts, with an average of 3799.72 (SD 1296.06) posts published daily. The LDA topic extraction identified 8 topics, with the most numerous being the Shanghai outbreak, nucleic acid testing price, and transportation. The average sentiment score of the posts was 0.64 (SD 0.31), indicating a predominance of positive sentiment. For all topics, posts with positive sentiment consistently outnumbered those with negative sentiment (χ27=24,844.4, P<.001). The sentiment scores of posts related to "nucleic acid testing price" decreased after May 22 compared with before (t120=3.882, P<.001). Similarly, the sentiment scores of posts related to the "Shanghai outbreak" decreased after June 1 compared with before (t120=11.943, P<.001). During public health emergencies, the topics of public concern were diverse. Public sentiment toward the regular nucleic acid testing policy was generally positive, but fluctuations occurred following the announcement of key policies. To understand the primary concerns of the public, the government needs to monitor social media posts by citizens. By promptly sharing information on media platforms and engaging in effective communication, the government can bridge the information gap between the public and government agencies, fostering a positive public opinion environment.

The impact of multimodal information features of short sales videos on consumer engagement behavior: A multi-method approach

In the emerging business model of short video commerce, consumers can engage in short sales videos by liking, commenting, and sharing. Compared with traditional shopping contexts, consumer engagement behavior in short video commerce settings is affected quickly and simultaneously by text, audio, and visual information, but the extant literature lacks sufficient exploration of the effects of such multimodal information. Based on the multimodal theory and signaling theory, this study aims to investigate the role of multimodal information including textual sentiment score, audio spectrum and visual effects of short sales videos on consumer engagement behavior. We collect 4292 veridical short sales videos from short video platform Douyin, while based on a multi-method approach of multiple regression analysis (MRA) and fuzzy set qualitative comparative analysis (fsQCA), it is found that the multimodal information features of short sales videos contribute to the promotion of consumer engagement behavior, and the cross-modal configurational solutions to enhance engagement is not unique. This study bridges the gap in the existing literature by capturing the impact of multimodal information features on consumer engagement behaviors and their asymmetric effects, which also provides practical implications for sellers and marketers on attracting consumers to engage in short video commerce.

Innovative Sentiment Analysis and Prediction of Stock Price Using FinBERT, GPT-4 and Logistic Regression: A Data-Driven Approach

This study explores the comparative performance of cutting-edge AI models, i.e., Finaance Bidirectional Encoder representations from Transsformers (FinBERT), Generatice Pre-trained Transformer GPT-4, and Logistic Regression, for sentiment analysis and stock index prediction using financial news and the NGX All-Share Index data label. By leveraging advanced natural language processing models like GPT-4 and FinBERT, alongside a traditional machine learning model, Logistic Regression, we aim to classify market sentiment, generate sentiment scores, and predict market price movements. This research highlights global AI advancements in stock markets, showcasing how state-of-the-art language models can contribute to understanding complex financial data. The models were assessed using metrics such as accuracy, precision, recall, F1 score, and ROC AUC. Results indicate that Logistic Regression outperformed the more computationally intensive FinBERT and predefined approach of versatile GPT-4, with an accuracy of 81.83% and a ROC AUC of 89.76%. The GPT-4 predefined approach exhibited a lower accuracy of 54.19% but demonstrated strong potential in handling complex data. FinBERT, while offering more sophisticated analysis, was resource-demanding and yielded a moderate performance. Hyperparameter optimization using Optuna and cross-validation techniques ensured the robustness of the models. This study highlights the strengths and limitations of the practical applications of AI approaches in stock market prediction and presents Logistic Regression as the most efficient model for this task, with FinBERT and GPT-4 representing emerging tools with potential for future exploration and innovation in AI-driven financial analytics.

Bridging sentiment and finance: An exploration of BOD statements and bank ratios in Jordan

Purpose: This study explores the intricate interplay between sentiment extracted from Board of Directors (BOD) statements and the traditional financial metrics of banks. It investigates the relationship between a bank’s financial performance, measured by various financial ratios, and the sentiment expressed in BOD statements.Design/methodology/approach: Drawing from a dataset encompassing BOD statements from 15 Jordanian banks spanning 2017 to 2021, sentiment analysis techniques are employed to derive sentiment scores. These scores are then juxtaposed with principal components distilled from financial metrics to examine their predictive relationship.Findings: Preliminary findings reveal significant non-linear patterns between Hu and Liu’s (2004) sentimental analysis (Lui_sen), the Data Science Lab’s (n.d.) multilingual sentiment tool (multi_sen), and the principal components. A subsequent random-effects panel regression further elucidates these relationships, highlighting the significant influence of specific principal components on sentiment scores. This study provides a compelling case for integrating sentiment analysis with traditional financial indicators, offering a more comprehensive toolkit for evaluating bank performance.Research limitations/implications: Future research should explore broader datasets across multiple sectors, to validate the generalisability of the findings beyond the banking sector in Jordan.Practical implications: This research holds implications for investors, analysts, and stakeholders by combining sentiment analysis with traditional financial metrics. It offers deeper insights into the health of banks and the intentions of management. This integrated approach allows stakeholders to make informed decisions and mitigate risks based on a more comprehensive view of banks' financial situations. Additionally, the possibility of real-time sentiment analysis enables the dynamic monitoring of banking institutions' financial health as updated data becomes available.Social implications: The study’s results are valuable to users of financial information, aiding the Association of Banks and the Central Bank of Jordan in formulating policies for the banking sector. Auditors, consultants, and regulatory bodies in Jordan can also benefit from these insights.Originality/value: This is the first study exploring the relationship between sentiment extracted from BOD statements and traditional financial metrics in Jordan. It provides novel evidence of this relationship, contributing to the literature on the integration of textual data analysis and financial performance.

Perceived Responses of International Tourists to Transportation and Tourism Services During Typhoons Faxai and Hagibis in Japan

There is a limited understanding on the information-seeking behavior of international tourists during disaster response scenarios due to the lack of empirical studies on crisis communication in Japan. This study clarifies the topics generated from both international tourists and official Twitter accounts by applying the embedding Bidirectional Encoder Representations from Transformers (BERT) topic model and examines the temporal sentiment changes toward transportation and tourism using the sentiment scores obtained from topic-based Valence Aware Dictionary and sEntiment Reasoner (VADER) sentiment analysis during disasters. A case study was conducted using Twitter data on Typhoons Faxai and Hagibis, which struck Japan in 2019. This study found differences in the topics generated among international tourists and officials in response and a continuous negative sentiment toward specific transportation services. The managerial implications of these findings regarding the use of social media in crisis communication in tourism are also discussed.

Do all complain the same? Examining the role of luxury hotels reviewer attributes using text mining

Purpose Although extant studies found the determinants of customer complaints in luxury hotels, comparison of complaint severity using reviewer characteristics (nationality, reviewer type and expertise) is limited. This study aims to identify the variation in customers’ complaint attributes among different customer categories. Design/methodology/approach This study tried to address this gap by using structural topic modelling and text mining techniques to find the sentiment score, word count, complaints attributes and severity score. The research sample covers luxury hotels in India, and online review text and their associated data has been collected and analysed. Findings The authors found that leisure customers wrote lengthier reviews and complained about the facility-related aspects (valet parking, entertainment and amenities) and food-related aspects (food variety, service and menu) while the business customers’ complaint severity score is more on salient aspects (room size, bathroom and staff related services). The international customers wrote lengthier reviews and complained more about the facility-related aspects (amenities, pick-up and entertainment) and core service offerings (stay experience, bathroom, dirtiness and complaint resolution). Whereas domestic customers complained more about stay-related aspects (food variety, quality, services, staff behaviour and stay experience). Also, this study used correlation analysis to find the dependencies between topics and uncovered underlying patterns, trends and associations, aiding in a deeper understanding and interpretation of the complaints. This study can help service providers tailor their services based on customer attributes to ensure maximum customer satisfaction. Originality/value The findings of this study can enable hotel service providers to customise their services based on customer attributes to reduce dissatisfaction among customers and promote customer loyalty. The findings of this study can also be used to devise differential promotional strategies for different types of customers.

The role of news sentiment in salmon price prediction using deep learning

This paper employs deep learning models and sentiment analysis to predict salmon spot prices. Our data includes historical price data and sentiment scores from 2018 to 2022. We extract sentiment scores from salmon-related news headlines by using FinBERT and TextBlob. We begin with price prediction using only historical price data and then introduce sentiment scores to improve the prediction accuracy of deep learning models. We find that the prediction performance of deep learning models outperforms traditional prediction methods in the salmon market. Our primary hybrid CNN-LSTM model outperforms other deep learning models and traditional models. Additionally, deep learning models incorporating sentiment scores exhibit reduced prediction errors. Our findings confirm the value of sentiment information in improving forecasting performance. These findings highlight the effectiveness and robustness of our CNN-LSTM model combined with sentiment analysis for price prediction in the salmon market.

A novel method for a technology enhanced learning recommender system considering changing user interest based on neural collaborative filtering

This study introduces an advanced recommender system for technology enhanced learning (TEL) that synergizes neural collaborative filtering, sentiment analysis, and an adaptive learning rate to address the limitations of traditional TEL systems. Recognizing the critical gap in existing approaches—primarily their neglect of user emotional feedback and static learning paths—our model innovatively incorporates sentiment analysis to capture and respond to nuanced emotional feedback from users. Utilizing Bidirectional Encoder Representations from Transformers for sentiment analysis, our system not only understands but also respects user privacy by processing feedback without revealing sensitive information. The adaptive learning rate, inspired by AdaGrad, allows our model to adjust its learning trajectory based on the sentiment scores associated with user feedback, ensuring a dynamic response to both positive and negative sentiments. This dual approach enhances the system’s adaptability to changing user preferences and improves its contentment understanding. Our methodology involves a comprehensive analysis of both the content of learning materials and the behaviors and preferences of learners, facilitating a more personalized learning experience. By dynamically adjusting recommendations based on real-time user data and behavioral analysis, our system leverages the collective insights of similar users and relevant content. We validated our approach against three datasets—MovieLens, Amazon, and a proprietary TEL dataset—and saw significant improvements in recommendation precision, F-score, and mean absolute error. The results indicate the potential of integrating sentiment analysis and adaptive learning rates into TEL recommender systems, marking a step forward in developing more responsive and user-centric educational technologies. This study paves the way for future advancements in TEL systems, emphasizing the importance of emotional intelligence and adaptability in enhancing the learning experience.

An Optimal Multi-Modal Approach for Stock Market Price Forecasting with Fused Sentiment Analysis for Real Time Data

Abstract: This study presents an innovative fusion-based methodology that integrates real-time stock market technical indicators with news sentiment analysis from financial news feeds to enhance stock selection decisions. The proposed framework employs a Bidirectional Long Short-Term Memory (Bi-LSTM) model for forecasting stock prices and a Deep Neural Network (DNN) used in conjunction with transformer-based model for sentiment classification, both optimized through the incorporation of real-time datasets. To further refine feature selection, Artificial Bee Colony (ABC) and Firefly algorithms are utilized, significantly improving the accuracy of price trend predictions and sentiment analysis. Technical indicators from stock market data are processed using the Bi-LSTM model to predict future stock prices. Concurrently, sentiment data from the Economic Times is pre-processed through Term Frequency-Inverse Document Frequency (TF-IDF) vectorization and pre-trained transformer models to extract key sentiment scores. The ABC algorithm enhances textual feature selection, while the Firefly algorithm reduces skewness in the error distribution of technical indicators, thereby improving the alignment between forecasted and actual stock prices. The fusion mechanism integrates predicted price movements with sentiment analysis to provide actionable insights for stock selection. The decision-making framework classifies stocks based on the consistency between price movements and sentiment: stocks predicted to rise with positive sentiment are labelled as "Sure Selection"; rising prices with negative sentiment are categorized as "Dicey Selection"; declining prices with positive sentiment result in a "Risky Selection"; and negative sentiment with falling prices trigger an "Avoid Selection". The proposed approach achieved a prediction accuracy of 67.00% for the Bi-LSTM model, with a precision of 0.714, recall of 0.695, and an F1 score of 0.677, demonstrating the model's robustness. The mean absolute error (MAE) for the Bi-LSTM model was 4.30, indicating strong predictive performance. The combined use of ABC and Firefly algorithms optimizes both technical and sentiment features, significantly contributing to the overall performance of the prediction model. By fusing multiple data sources and employing advanced optimization techniques, this methodology offers a unique and effective solution for stock market prediction and decision-making

Social media in healthcare emergency management: insights from Spanish hospitals during the COVID-19 pandemic

PurposeThis study aims to explore the Facebook communication strategies of Spanish hospitals during health emergencies, focusing on their role in crisis management and public information dissemination.Design/methodology/approachAutomatic topic modelling and deep learning sentiment analysis were applied to analyse 151,738 posts from 274 hospital Facebook pages (March 2020–Feb 2022). Regression analyses were used to explore the relationships between topics, sentiment scores and hospital characteristics.FindingsThe analysis revealed nine main topics, with the three most prevalent related to COVID-19: vaccine information, security measures and situational updates. This indicates that Spanish hospitals significantly relied on Facebook to manage the emergency. The communication strategies dynamically adapted to the intensity of the pandemic and varied across hospital types. Sentiment analysis showed a negative tone for posts about security measures and situational updates. These findings align with the Agenda-Setting Theory, suggesting that hospitals influenced public discourse. Vaccine information posts were more positive, resonating with the Uses and Gratifications Theory by fulfilling the audience’s need for reassurance and guidance.Originality/valueUsing replicable machine learning techniques, this study elucidates the communication strategies employed by Spanish hospitals to manage healthcare emergencies, such as the COVID-19 pandemic. It highlights factors that potentially influence these strategies and provides theoretical justifications for them. The variation in communication strategies on Facebook among different hospital categories underscores the imperative for stricter guidelines and regulations to guarantee consistent and reliable communication during emergencies. This research provides valuable insights for practitioners and policymakers aimed at developing effective health communication strategies on social media.

Media Sentiment and Its Role in Shaping IPO Performance

The media performs an essential role in providing information regarding initial public offerings (IPOs) to investors. Often, public information about IPO companies is scarce and investors find it difficult to comprehend complex documents, making media a cost-effective and authentic source of information. This study explores how media sentiment affects the underpricing and initial aftermarket performance of IPOs using 3,124 media articles published from April 2019 to March 2022. In doing so, the authors employed textual sentiment analysis using the Loughran and McDonald dictionary to deduce the media net sentiment score. Subsequently, robust regression is conducted by incorporating the media sentiment, the volume of media articles, and other control variables. The result mentions that the way media present information has the potential to influence the decision of retail investors. The net sentiment score shows that the positive information in the media article makes investors optimistic about the prospects of the business. It boosts their confidence and creates more demand for shares, thereby increasing the IPO’s first-day market-adjusted abnormal return. Further, it has been discovered that media sentiment significantly influences initial aftermarket performance primarily due to the limited trading history available for the IPO company during the initial stage.

Navigating the spectrum of service in hospitality: hotel employees’ experiences toward guests with disabilities

ABSTRACT The primary goal of the present study was to investigate the experiences and challenges hotel employees face when interacting with guests with disabilities. The findings of this study, derived from data collected using Reddit’s Application Programming Interface, are not just significant but crucial and timely. Using Structural Topic Modelling and quadrant analysis, 12 topics emerged, each stratified based on topic frequency and sentiment scores. These topics delve into the daily interactions of hotel employees, the emotional and practical challenges they face, and the critical issues within the hotel industry, including disability accommodation, the role of service animals, and the complexities of reservation systems in accommodating special needs. This study also sheds light on the emotional toll accommodating guests with special needs entails on employees, emphasising the urgent need for better training and clear policies to support employees and guests with disabilities. The findings suggest that service interactions can be significantly improved by enhancing disability awareness and accommodation practices.

Aspect based sentiment analysis of consumer reviews using unsupervised attention neural framework

Consumer reviews posted on e-commerce platforms offer an assessment of their opinion and sentiments about various aspects of a product or service. The aspects may belong to a specific category at subsystem and component level. The irrealis often associated with an opinion term, if not handled carefully, may distort the sentiment scores. This paper introduces a clustering and negative-sampling backed feed-forward Unsupervised Attention Deep Neural Framework for Opinion-Aspect-Category-Irrealis-Sentiment (UAN-OACIS) Quintuple extraction. This framework splits the five subtasks into two stages. In the first stage, aspects and opinions are extracted using word embedding-based multi-layer Attention Deep Neural Network (DNN) models; while the second stage projects corresponding categories, irrealis, and sentiment scores by creating a hash table and a couple of algorithms. We also propose a POS-Tag-based algorithm to make an irrealis dictionary to suggest sentiment modification percentage scores; and another algorithm to modify existing sentiment-lexicon based on domain-specific words. We compare UAN-OACIS with four unsupervised and two semi-supervised approaches at subtask level, and two supervised approaches at quadruples level excluding irrealis, and have got comparable results. Since no method is available for overall comparison at quintuple level, we additionally study the impact of irrealis on sentiment scores and observe substantial improvement compared to state-of-the-art lexicons methods.

Beyond Traditional Banking: Exploring the Influence of Technology on Depositors' Decision-Making in Bank Selection

Purpose: The banking sector is crucial for the economy, and with growing competition, banks must pinpoint factors influencing depositors' choices. This study aims to uncover how technology adoption by commercial banks plays a role in depositors' decisions. Methodology: A hybrid approached is attempted in this study to explore the phenomenon of interest. In the qualitative phase, focus group discussion was conducted on two heterogeneous groups, each with ten participants. The recorded audio was further transcribed using speech-to-text modules. A natural language processing tool has been used to identify words or phrases that are semantically related or have similar meanings to depositors’ sentiment analysis. The quantitative phase involved descriptive statistics, principal component analysis and regression analysis to explore insight numerically. Findings: Respondents expressed a generally positive sentiment toward technology in banking, influencing their selection decisions, in the qualitative phase. In the quantitative phase, significant technological aspects affecting depositors' post-selection evaluations are identified. Findings from both approaches converge on the same line. Practical implications: Findings of this study may aid decision-makers in banking industry to articulate distinct technological aspect implied in banking service delivery and better design service offerings to attract new depositors and gain competitive advantages. Originality/Value: This is the first study to focus solely on technology to explore how technology influences depositors' bank choices in the context of Bangladesh. Research Limitations: Significant parts of several conversations were sorted out for sentiment analysis. If it was possible to consider all relevant statements of the focus group discussion for sentiment analysis, it might convey a more pertinent sentiment score.

Gender Differences in Letters of Recommendations and Personal Statements for Neurotology Fellowship over 10 Years: A Deep Learning Linguistic Analysis.

Personal statements (PSs) and letters of recommendation (LORs) are critical components of the neurotology fellowship application process but can be subject to implicit biases. This study evaluated general and deep learning linguistic differences between the applicant genders over a 10-year span. Retrospective cohort. Two institutions. PSs and LORs were collected from 2014 to 2023 from two institutions. The Valence Aware Dictionary and Sentiment Reasoner (VADER) natural language processing (NLP) package was used to compare the positive or negative sentiment in LORs and PSs. Next, the deep learning tool, Empath, categorized the text into scores, and Wilcoxon rank sum tests were performed for comparisons between applicant gender. Among 177 applicants over 10 years, 120 were males and 57 were females. There were no differences in word count or VADER sentiment scores between genders for both LORs and PSs. However, among Empath sentiment categories, male applicants had more words of trust ( p = 0.03) and leadership ( p = 0.002) in LORs. Temporally, the trends show a consistently higher VADER sentiment and Empath "trust" and "leader" in male LORs from 2014 to 2019, after which there was no statistical significance in sentiment scores between genders, and females even have higher scores of trust and leadership in 2023. Linguistic content overall favored male applicants because they were more frequently described as trustworthy and leaders. However, the temporal analysis of linguistic differences between male and female applicants found an encouraging trend suggesting a reduction of gender bias in recent years, mirroring an increased composition of women in neurotology over time.