Validation of machine learning models for automated sentiment determination of Russian-language texts DOI Creative Commons
Polina A. Basina,

Darya O. Dunaeva,

Anna Yu. Sarkisova

и другие.

Vestnik Tomskogo gosudarstvennogo universiteta, Год журнала: 2022, Номер 485, С. 206 - 216

Опубликована: Янв. 1, 2022

Sentiment analysis is one of the most demanded natural language processing operations for solving applied problems. One key methods automated sentiment supervised machine learning. In presence a large selection ready-made solutions determining tonality, results models give significant errors due to complexity and contextual conditionality linguistic explication emotions. The article presents validation 6 Russian-language publications using research dataset – expertly marked 300 statements extracted from social network messages on subject quality life corresponding types: positive, negative, neutral. To evaluate performance models, interannotator agreement coefficients were used, in particular, Krippendorff’s alpha, Cohen’s kappa Fleiss’s coefficients. obtained values showed low level reliability between expert labels that assigned by models. Among experiments performed, lowest achieved Blanchefort model trained Rusentiment data, highest same developer medical feedback data. Based obtained, conclusions drawn about common causes disagreements learning Machine correctly identify tone texts if they contain bright lexical markers match general statement. On contrary, problems an emotionally charged message are provoked word with opposite it. use emotive vocabulary does not entire statement, marker words their direct meanings, uppercase, forms complicated communication (including irony, sarcasm) remain risk factors attracting resources: high degree probability automatic classification will be able determine text. main reason “difficulties” determination task focusing utterance as integral unit refusal focus individual formal indicators. minimum communicative speech. Capturing its semantic expressive integrity super analysis. So, it still quite difficult trust such complex categorization It advisable associate prospects directions this area, first all, development high-quality, linguistically sound training datasets.

Язык: Английский

Is the Size of the City Important for the Quality of Urban Life? Comparison of a Small and a Large City DOI Open Access
Lucia Petrikovičová,

Victoria Kurilenko,

Amantius Akimjak

и другие.

Sustainability, Год журнала: 2022, Номер 14(23), С. 15589 - 15589

Опубликована: Ноя. 23, 2022

In the recent past, question of determining optimal city size in relation to quality urban life (QoUL) was raised inhabitants. This article has evaluated correlation QoUL index cities number We also deal with selected variables for which we assume a relationship QoUL. The authors who calculated indices equated its objective dimension considered as place. It turned out that growth inhabitants Slovak did not correlate improving life. Our examined two different countries on scale 0–10 through questionnaires. obtained values are subjective assessment. From global point view, one is small and other big. achieved better results international rankings life, it assumed this fact would be reflected quantification One hypothesis will achieve than large city. paper presents measurement correlation.

Язык: Английский

Процитировано

12

Quality of life issues in rural settlements: Assessment by social media users DOI Creative Commons
Evgeniy V. Shchekotin, Vyacheslav L. Goiko,

Darya O. Dunaeva

и другие.

Data Science and Management, Год журнала: 2024, Номер 7(4), С. 283 - 292

Опубликована: Март 19, 2024

Disparity in the quality of life people living rural and urban areas is among major problems, often leading to greater depression population compared those who have access a high standard living. The goal this study was identify issues that are fundamental from point view settlements typical for various regions Russian Federation with diverse geographical, socioeconomic, demographic characteristics. practical relevance our based on identification scope problems related areas. value significance digital tracks, which can serve as source information. data sources included messages posts discussing aspects published by VKontakte. For study, we used negative implications communities ten regions. These include housing infrastructure utilities, transportation, environment, telecommunications, banking, healthcare, education.

Язык: Английский

Процитировано

1

Monitoring Urban Population’s Quality of Life via Digital Footprint: The Case of Novosibirsk DOI

Elizaveta Babaikina,

Galina Kurcheeva, Maxim Bakaev

и другие.

Springer geography, Год журнала: 2024, Номер unknown, С. 21 - 34

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

0

Digital Footprint: Assessing Student Satisfaction with Education Quality DOI Creative Commons
М. М. Криштал, А. В. Богданова, Mikhail Myagkov

и другие.

Vysshee Obrazovanie v Rossii = Higher Education in Russia, Год журнала: 2024, Номер 33(2), С. 89 - 108

Опубликована: Март 6, 2024

The COVID-19 pandemic has changed the way learning is organized around world. Russian universities have also been faced with need to quickly transfer all teaching an online format. importance of student satisfaction education quality in increasing, since it important condition for motivation. paper shows that based on analysis students’ messages social network, possible observe and analyze overall dynamics trends community / efficiency conduct a comparative identified characteristic data groups their totality. It shown reaction students particular university may significant deviations from totality data, which reflects characteristics events occurring at same time. This indicate internal differences university, form appropriate response external events. To understand transition new implementation format learning. digital traces network VKontakte were analyzed using individual Big Data tools PolyAnalyst software platform. made trace changes mood and, example single identify explain attitude learning, as well verify methodology. methodology developed by authors makes detect problematic issues including moment occurrence, relevance, degree concern students. Such content can be used not only assess but monitor emergence any problems cause strong reactions part community, other communities groups.

Язык: Английский

Процитировано

0

AXIOLOGICAL COMPONENT IN THE STRUCTURE OF VISUAL IMAGES OF THE URBAN ENVIRONMENT DOI Open Access

Сергей Владимирович Пирогов,

Vitaliy V. Kashpur, Daria O. Dunaeva

и другие.

ΠΡΑΞΗMΑ Journal of Visual Semiotics, Год журнала: 2022, Номер 1(31), С. 75 - 89

Опубликована: Фев. 10, 2022

Статья посвящена возможностям изучения ценностного отношения к городской среде на материале пользовательского контента в социальных сетях. Визуальный образ города контексте цифровых следов рассматривается как когнитивная модель среды, которой отражаются интерпретативные и проективные конструкции – субъективные представления о локусах которые становятся освоенными «местами», выражают определённые ценностные ожидания субъектов восприятия. Показано, что контент сетей реализует прежде всего коммуникативную функцию взаимные между пользователями, а также пользователей с самим городом. Утверждается, межличностная коммуникация городом являются условием построения личностной идентичности, так существования изменения самого города, который существует том числе городская идентичность, освоенные места городского пространства. Выявленная тенденция апроприации среды вполне отчётливо выражена визуальных материалах городских сообществ, является новым значимым явлением. Анализ сетях показывает, фотографии слабо репрезентируют витальные ценности горожан, связанные вопросами их бытовой жизнедеятельности, однако них выражаются экзистенциальные потребности жителей города. Доминирующей ценностно-смысловой тональностью оценки возможность её личностного освоения присвоения. The article is dedicated to the possibilities of studying axiological attitude towards urban environment on basis user-generated content in social networks. visual image city context digital footprints considered be a cognitive model environment, which reflects interpretative and projective constructions subjective ideas about loci that become adopted “places” express certain attitudes expectations individuals. shows networks implements, first all, communicative function between users, them itself. Interpersonal communication with are conditions for constructing personal identity existence change. exists, particular, as an identity, forms places space. described tendency appropriation quite clearly expressed materials communities, appear new significant phenomenon. analysis photographs poorly represent dwellers’ vital values everyday life, but they their existential needs. dominant value-semantic meaning its ability developed appropriated by person

Язык: Русский

Процитировано

2

Analysis of environmental problems based on social media data (on the example of atmospheric air quality) DOI Creative Commons
Evgeniy V. Shchekotin

E3S Web of Conferences, Год журнала: 2023, Номер 458, С. 08010 - 08010

Опубликована: Янв. 1, 2023

The article discusses the state and prospects of two new methods to study environmental issues: Internet ecology (iEcology) conservation culturomics. Both approaches are very similar; both them based on big data analysis, which is not directly meant solve issues (publications in social networks, search, photos videos posted platforms, etc.). authors offer methodology (as exemplified by quality atmospheric air) from VK network machine learning algorithms. For content analysis we used PolyAnalyst software. results publications air Magnitogorsk city for 2020-2022 presented. We identified 433 messages characterizing condition Magnitogorsk. Our research demonstrates that ecological culturomics can contribute situation. let us issue important residents be as an additional source information subjective assessment quality.words.

Язык: Английский

Процитировано

0

Social issues of rural areas: social media open data analysis DOI Creative Commons
Evgeniy V. Shchekotin

E3S Web of Conferences, Год журнала: 2023, Номер 462, С. 03008 - 03008

Опубликована: Янв. 1, 2023

This article offers a new approach to study social issues in Russian rural settlements. Our is based on the analysis of digital footprints left by population largest network VKontakte (VK). media data and two methods – big mining. For this we used negative-sounding messages posts published some communities settlements from 10 (out 85) regions. We PolyAnalyst software for automatic such identify groups deteriorating quality life those These include housing utilities infrastructure, capital improvement transportation infrastructure; environmental issues, with availability poor telecommunication banking healthcare education issues; safety various economic related homeless neglected animals (dogs); alcohol abuse inefficiency authorities

Язык: Английский

Процитировано

0

Validation of machine learning models for automated sentiment determination of Russian-language texts DOI Creative Commons
Polina A. Basina,

Darya O. Dunaeva,

Anna Yu. Sarkisova

и другие.

Vestnik Tomskogo gosudarstvennogo universiteta, Год журнала: 2022, Номер 485, С. 206 - 216

Опубликована: Янв. 1, 2022

Sentiment analysis is one of the most demanded natural language processing operations for solving applied problems. One key methods automated sentiment supervised machine learning. In presence a large selection ready-made solutions determining tonality, results models give significant errors due to complexity and contextual conditionality linguistic explication emotions. The article presents validation 6 Russian-language publications using research dataset – expertly marked 300 statements extracted from social network messages on subject quality life corresponding types: positive, negative, neutral. To evaluate performance models, interannotator agreement coefficients were used, in particular, Krippendorff’s alpha, Cohen’s kappa Fleiss’s coefficients. obtained values showed low level reliability between expert labels that assigned by models. Among experiments performed, lowest achieved Blanchefort model trained Rusentiment data, highest same developer medical feedback data. Based obtained, conclusions drawn about common causes disagreements learning Machine correctly identify tone texts if they contain bright lexical markers match general statement. On contrary, problems an emotionally charged message are provoked word with opposite it. use emotive vocabulary does not entire statement, marker words their direct meanings, uppercase, forms complicated communication (including irony, sarcasm) remain risk factors attracting resources: high degree probability automatic classification will be able determine text. main reason “difficulties” determination task focusing utterance as integral unit refusal focus individual formal indicators. minimum communicative speech. Capturing its semantic expressive integrity super analysis. So, it still quite difficult trust such complex categorization It advisable associate prospects directions this area, first all, development high-quality, linguistically sound training datasets.

Язык: Английский

Процитировано

0