Multilingual Sentiment Analysis and Toxicity Detection for Text Messages in Russian

Bogoradnikova, Darya; Makhnytkina, Olesia; Matveev, Anton; Zakharova, Anastasia; Akulov, Artem

doi:10.23919/fruct52173.2021.9435584

Cited by 9 publications

(5 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We adopted the LDA TM technique, which assumes that texts are generated from a mixture of topics [ 44 ]. LDA is efficient and can generate topics of better quality [ 45 ]. From the data set created, we generated 2 probability distribution outputs: the probability distribution of topics over documents and the probability distribution of terms over topics [ 41 , 43 ].…”

Section: Methodsmentioning

confidence: 99%

Public Trust in Artificial Intelligence Applications in Mental Health Care: Topic Modeling Analysis

Shan¹,

Ji²,

Xie³

et al. 2022

JMIR Hum Factors

View full text Add to dashboard Cite

Background Mental disorders (MDs) impose heavy burdens on health care (HC) systems and affect a growing number of people worldwide. The use of mobile health (mHealth) apps empowered by artificial intelligence (AI) is increasingly being resorted to as a possible solution. Objective This study adopted a topic modeling (TM) approach to investigate the public trust in AI apps in mental health care (MHC) by identifying the dominant topics and themes in user reviews of the 8 most relevant mental health (MH) apps with the largest numbers of reviewers. Methods We searched Google Play for the top MH apps with the largest numbers of reviewers, from which we selected the most relevant apps. Subsequently, we extracted data from user reviews posted from January 1, 2020, to April 2, 2022. After cleaning the extracted data using the Python text processing tool spaCy, we ascertained the optimal number of topics, drawing on the coherence scores and used latent Dirichlet allocation (LDA) TM to generate the most salient topics and related terms. We then classified the ascertained topics into different theme categories by plotting them onto a 2D plane via multidimensional scaling using the pyLDAvis visualization tool. Finally, we analyzed these topics and themes qualitatively to better understand the status of public trust in AI apps in MHC. Results From the top 20 MH apps with the largest numbers of reviewers retrieved, we chose the 8 (40%) most relevant apps: (1) Wysa: Anxiety Therapy Chatbot; (2) Youper Therapy; (3) MindDoc: Your Companion; (4) TalkLife for Anxiety, Depression & Stress; (5) 7 Cups: Online Therapy for Mental Health & Anxiety; (6) BetterHelp-Therapy; (7) Sanvello; and (8) InnerHour. These apps provided 14.2% (n=559), 11.0% (n=431), 13.7% (n=538), 8.8% (n=356), 14.1% (n=554), 11.9% (n=468), 9.2% (n=362), and 16.9% (n=663) of the collected 3931 reviews, respectively. The 4 dominant topics were topic 4 (cheering people up; n=1069, 27%), topic 3 (calming people down; n=1029, 26%), topic 2 (helping figure out the inner world; n=963, 25%), and topic 1 (being an alternative or complement to a therapist; n=870, 22%). Based on topic coherence and intertopic distance, topics 3 and 4 were combined into theme 3 (dispelling negative emotions), while topics 2 and 1 remained 2 separate themes: theme 2 (helping figure out the inner world) and theme 1 (being an alternative or complement to a therapist), respectively. These themes and topics, though involving some dissenting voices, reflected an overall high status of trust in AI apps. Conclusions This is the first study to investigate the public trust in AI apps in MHC from the perspective of user reviews using the TM technique. The automatic text analysis and complementary manual interpretation of the collected data allowed us to discover the dominant topics hidden in a data set and categorize these topics into different themes to reveal an overall high degree of public trust. The dissenting voices from users, though only a few, can serve as indicators for health providers and app developers to jointly improve these apps, which will ultimately facilitate the treatment of prevalent MDs and alleviate the overburdened HC systems worldwide.

show abstract

Section: Methodsmentioning

confidence: 99%

Public Trust in Artificial Intelligence Applications in Mental Health Care: Topic Modeling Analysis

Shan¹,

Ji²,

Xie³

et al. 2022

JMIR Hum Factors

View full text Add to dashboard Cite

show abstract

“…These algorithms enable users to organize and summarize numerous documents which is unimaginable through manual annotation [18], therefore ascertaining topics hidden in documents [21]. Assuming that texts are generated from a mixture of topics [22], the LDA topic modeling technique can e ciently generate topics of better quality [23]. Two probability distribution outputs were generated from the data set created: the probability distribution of topics over documents and the probability distribution of terms over topics [18,21].…”

Section: Topic Modeling With Ldamentioning

confidence: 99%

Exploring People’s Anxiety-Related Concerns Raised on a Computer-Mediated Health Platform: Topic Modeling Analysis

Shan,

Liu,

Sun

et al. 2024

Preprint

View full text Add to dashboard Cite

Background About one in four Chinese people might suffer or have already suffered from anxiety conditions, with a lifetime prevalence rate of 4.8%. However, many of those who could have benefited from psychological or pharmacological treatments fail to be recognized or treated due to the lack of timely recognition and accurate diagnosis.Objective This study used a topic modeling approach to explore people’s anxiety-related concerns raised on a computer-mediated Chinese health platform, YOU WEN BI DA (questioning and answering), to provide implications for accurate diagnosis, targeted education, tailored intervention, and informed policy-making in the course of addressing this condition of public concern.Methods First, we extracted data from YOU WEN BI DA between May 2022 and February 2023. After cleaning the extracted data both using the Python text processing tool spaCy and manually, we ascertained the optimal number of topics by drawing on the coherence scores and used latent Dirichlet allocation (LDA) topic modeling to generate the most salient topics and related terms. We then categorized the topics ascertained into different classes of themes by plotting them onto a 2D plane via multidimensional scaling using the pyLDAvis visualization tool. Finally, we analyzed these topics and themes qualitatively to better understand people’s anxiety-related concerns.Results 5 topics with different overall prevalence were ascertained through data analysis. Topic 2 (tinnitus phobia-incurred concerns, n = 639) is the most popular dominant topic, occurring in 25.1% of the 2545 collected concerns, closely followed by Topics 1 (neurosis-incurred concerns, n = 512;) and 3 (sleep, dyskinesia, bipolar, cognitive, and somatic disorders-incurred concerns, n = 619), which appeared in 20.1% and 24.3% of the 2545 concerns respectively. Topic 5 (social phobia-incurred concerns, n = 428) ranks as the fourth most popular dominant topic, showing up in 16.8% of the 2545 concerns. Topic 4 (autonomic nerve dysfunction-incurred concerns, n = 347) accounts for 13.6% of the 2545 concerns. The t-distributed Stochastic Neighbor Embedding analysis reveals partial similarities between Topics 2 and 5 as well as between Topics 4 and 5 because many concerns involved in Topics 2 and 5 pertain to people’s psychological status of fear and anxiety and the relief and dispelling of such symptoms through medication, and many concerns involved in Topics 4 and 5 relate to people’s worries about the negative impact on their nerves and the adjustment and conditioning of such effects through medication.Conclusions This was the first study that investigated Chinese people’s anxiety-related concerns raised on YOU WEN BI DA using the topic modeling technique. The automatic text analysis and complementary manual interpretation of the collected data allowed for the discovery of the dominant topics hidden in the data and the categorization of these topics into different themes to reveal the overall status of people’s anxiety-related concerns. The research findings can provide some practice implications for health and medical educators, practitioners, and policy-makers to make joint efforts to address this common public concern effectively and efficiently.

show abstract

“…Perspective API detects incivility in short text written in several languages, including Russian [18]. The service was used to detect toxicity in short comments in Russian-language online communities and services (Bogoradnikova, et al, 2021) as a baseline for comparison, and outperformed many other methods checked by the authors.…”

Section: Analysis Of Comment Toxicitymentioning

confidence: 99%

Navalny’s direct-casting: Affective attunement and polarization in the online community of the most vocal Russian opposition politician

Zinnatullin

2023

View full text Add to dashboard Cite

Social media has a significant impact on the process of political polarization. Despite a large body of research on polarization and social media in democracies, studying this relationship in autocracies remains a niche field. This paper describes the content, composition, and behavioral patterns of discussions that take place on YouTube in the community of Russia’s most vocal opposition politician, Alexei Navalny. Based on a corpus of more than eight million comments, this study provides empirical evidence on the relatively short-term nature of affective attunement induced by a leader promising social changes within an authoritarian context. This discovery is manifested in the observation that periods of high public interest in Navalny’s activities are marked by a significant influx of new audience members into his community. However, the retention rate of this cohort of users was lower than that of the cohort of commenters, who started discussions during periods of lower public interest in Navalny’s activities. This conclusion applies not only to the entire set of commenters, but also to pro-government and anti-government users. According to the exploratory text analysis, the most common topics in discussions were praising Navalny’s activities, criticizing the government, and enticing people to share videos to change the minds of apolitical citizens or pro-government supporters. Finally, one of the affective polarization parameters, the degree of toxicity of discussions, is higher on Navalny’s community than on an apolitical celebrity’s YouTube channel, which establishes a baseline for the level of incivility.

show abstract

Multilingual Sentiment Analysis and Toxicity Detection for Text Messages in Russian

Cited by 9 publications

References 30 publications

Public Trust in Artificial Intelligence Applications in Mental Health Care: Topic Modeling Analysis

Public Trust in Artificial Intelligence Applications in Mental Health Care: Topic Modeling Analysis

Exploring People’s Anxiety-Related Concerns Raised on a Computer-Mediated Health Platform: Topic Modeling Analysis

Navalny’s direct-casting: Affective attunement and polarization in the online community of the most vocal Russian opposition politician

Contact Info

Product

Resources

About