Blog Posts Research Articles

Effectiveness of the Internet-Based, Job-Related Aftercare GSA-Online Following Inpatient Psychosomatic Rehabilitation: Results of a Randomized Controlled Trial

In order to sustain the treatment results of medical rehabilitation in the long term and to support the transfer of learned strategies into everyday life, outpatient aftercare is often indicated. This is especially true for psychosomatic rehabilitation patients with occupational stress, for whom reintegration into working life is a particular challenge. Since access to care services close to home is limited, internet-based aftercare interventions offer the possibility of timely and flexible aftercare tailored to the patient's needs. In a randomized controlled trial, the internet-based job-related aftercare GSA-Online proved to be effective across indications with regard to job-related and health-related outcome criteria. The aim of the present study is to examine these outcome criteria in a subsample of patients undergoing inpatient psychosomatic rehabilitation. Occupationally stressed rehabilitation inpatients were assigned to the intervention (IG) or control group (CG) by means of cluster randomization after meeting the inclusion criteria and participating an inpatient vocational stress management training. After discharge from rehabilitation, patients in the IG were given access to GSA-Online for twelve weeks. In the intervention, maladaptive social interactions in the workplace were identified with the help of self-written blog posts and addressed with the help of therapeutic comments. The active CG received access to selected online information on health-promoting behaviors. Target measures included subjective prognosis of gainful employment (SPE), depressiveness (PHQ-9), and anxiety (GAD-7). Self-report measures were assessed at the end of aftercare and at follow-up (twelve months after the end of rehabilitation). Missing values were replaced using multiple imputation. Ninety-one percent of the IG (N=89) and 70% of the CG (N=106) logged on to the website at least once. There were no group differences in the subjective prognosis of gainful employment at the end of aftercare but a trend toward more optimistic scores in the IG at follow-up. Significantly lower psychological distress was observed in the IG, especially with regard to anxiety (at both measuring points) but also with regard to depressive symptoms (follow-up) and experience of stress (end of intervention). In the subsample of psychosomatic rehabilitation, the internet-based, job-related aftercare GSA-Online led to a significant reduction in psychological symptoms. Regarding the subjective prognosis of gainful employment, there was at least a trend in favor of IG. A larger sample is needed to more closely examine the results of this exploratory evaluation. Additionally measures to increase adherence in the IG should be explored.

When Immigrant Mothers of Color Become Public School Teachers for English Language Learners: Intersectionality for Transformative Teacher Preparation

AbstractThis article contests the whitewashed ideologies and practices in teacher preparation programs and K–12 schools (Lee‐Johnson, 2019) with a counter‐storytelling analysis (Solorzano & Yosso, 2001, 2002). Based on the analysis of the lived experiences of two immigrant mothers of color (IMC), who had transitioned into certified English Language Learner (ELL) teachers, the researcher proposes a transformative frame, which stems from Crenshaw's (1989) intersectionality and Mezirow's (1990) transformative learning. The data sources include the two immigrant mothers' oral stories told at an ELL parent panel, their written teaching philosophy statements, and blog posts. The findings contribute to the knowledge base of teacher preparation in TESOL and applied linguistics by foregrounding the racially othered experiences of IMC; envisioning intersectionality as an agentive tool for negotiation, emancipation, and advocacy; and proposing a transformative frame for TESOL teacher education. The researcher calls for considerations of using intersectionality for transforming the teaching practices and ideologies in teacher preparation programs and K–12 schools.

Scholarly Publishing, Boundary Processes, and the Problem of Fake Peer Reviews.

Over the past decade, the phenomenon of "fake" peer reviews has caused growing consternation among scholarly publishers. Yet despite the significant behind-the-scenes impact that anxieties about fakery have had on peer review processes within scholarly journals, the phenomenon itself has been subject to little scholarly analysis. Rather than treating fake reviews as a straightforward descriptive category, in this article, we explore how the discourse on fake reviews emerged and why, and what it tells us about its seeming antithesis, "genuine" peer review. Our primary source of data are two influential adjudicators of scholarly publishing integrity that have been critical to the emergence of the concept of the fake review: Retraction Watch and the Committee on Publication Ethics. Via an analysis of their respective blog posts, Forum cases, presentations, and best practice guidance, we build a genealogy of the fake review discourse and highlight the variety of players involved in staking out the fake. We conclude that constant work is required to maintain clear lines of separation between genuine and fake reviews and highlight how the concept has served to reassert the boundaries between science and society in a context where they have increasingly been questioned.

Improving the Factual Accuracy of Abstractive Clinical Text Summarization using Multi-Objective Optimization.

While there has been recent progress in abstractive summarization as applied to different domains including news articles, scientific articles, and blog posts, the application of these techniques to clinical text summarization has been limited. This is primarily due to the lack of large-scale training data and the messy/unstructured nature of clinical notes as opposed to other domains where massive training data come in structured or semi -structured form. Further, one of the least explored and critical components of clinical text summarization is factual accuracy of clinical summaries. This is specifically crucial in the healthcare domain, cardiology in particular, where an accurate summary generation that preserves the facts in the source notes is critical to the well-being of a patient. In this study, we propose a framework for improving the factual accuracy of abstractive summarization of clinical text using knowledge-guided multi-objective optimization. We propose to jointly optimize three cost functions in our proposed architecture during training: generative loss, entity loss and knowledge loss and evaluate the proposed architecture on 1) clinical notes of patients with heart failure (HF), which we collect for this study; and 2) two benchmark datasets, Indiana University Chest X-ray collection (IU X-Ray), and MIMIC-CXR, that are publicly available. We experiment with three transformer encoder-decoder architectures and demonstrate that optimizing different loss functions leads to improved performance in terms of entity-level factual accuracy.

Prediction of Certain examinations on AI procedures for Text rundown

Making text synopses from huge volume of unstructured text like client surveys, web log posts and online entertainment posts is a significant assignment in text mining applications. These synopses uncover the valuable data which depicts the whole report or surveys. Rundown errand could be performed utilizing two significant methodologies: Extractive and abstractive methodology. Extractive methodology distinguishes the huge piece of the report that uncovered the whole happy and removes it to shape synopsis. Abstractive approach makes synopsis in light of catchphrases and semantics from the archive, which makes it troublesome when contrasted with the other methodology. The extractive synopsis task is perplexing because of overt repetitiveness, enormous volume of text, changeability and semantics of regular language in the text. The wide pertinence as well as trying nature of the errand has enlivened dynamic research in the space by both scholar and industry specialists. This exploration centers around the plan of AI based frameworks for two center regions connected with text mining, specifically, Component based text synopsis and text comparability location. Existing element positioning and outline frameworks utilize an assortment of strategies including inactive semantic ordering; Innocent Bayes' and other semantics based approaches. Because of the intricacy of the undertaking there is a requirement for creating proficient frameworks. In this proposal, an element positioning framework in light of client inclinations have been created. Three different AI approaches have been embraced for highlight based miniature level extractive text synopsis development.

Obituary: Éva S. Balogh (1936–2021)

Born in Pécs, Éva S. Balogh was a twenty-year-old undergraduate student at Eötvös Loránd University at the time of the 1956 revolution. During the uprising, she helped publish a revolutionary student newspaper and fled Hungary in December 1956 to find refuge in Canada following the Soviet suppression of the revolt. Balogh completed a B.A. in history at Carleton University in Ottawa, followed by an M.A. and PhD in history from Yale University. Her scholarly career included serving as a professor of Russian and Eastern European history, as well as working in senior academic administration as the dean of Yale’s Morse College from 1973 to 1978. After this, Balogh’s career shifted focus, and she helped establish Brevis Press, a publishing company.Balogh was an early user of the Internet in March 1994, being an active participant in two nascent discussion groups on Hungary and Hungarian politics. Eventually, she established and moderated her own online discussion group on Hungary, before launching the Hungarian Spectrum blog in June 2007 from her home in Bethany, Connecticut. In a video interview with journalist Benjamin Novak, Balogh once recounted a friend’s initial reaction to her first blog post. “Oh, Éva,” the friend said with a tone of gentle pity, “a blog isn’t meant to be a historical treatise.”1 That comment reflected the challenge of transitioning from academic to journalistic writing.However, it was a challenge that Balogh would meet head on, as she wrote daily blog posts that mostly read like long-form journalistic articles. Her pieces offered analysis of Hungarian current affairs and trends in contemporary society, based on a wide range of Hungarian-language sources, as well as on her own connections to, and discussions with, key people in Hungarian democratic opposition circles. Balogh’s writing combined informed opinion and news reporting—all from a broadly liberal democratic angle. Balogh also produced English translations of Hungarian written sources. Over the course of fourteen years, Balogh published some 5,000 articles in Hungarian Spectrum—rarely missing a day and writing nearly all the published pieces herself. Her process was to spend each morning reading Hungarian news sites—both from the left and the right political spectrum—and then the second half of the day on writing her daily blog posts, most of which exceeded 1,000 words in length.In 2011, Balogh’s commitment to democratic values in Hungary led her to become a founding member of the Canadian Hungarian Democratic Charter. Her diligent daily reporting in Hungarian Spectrum also brought her widespread recognition. In 2019, when George Soros won the Ridenhour Courage Prize, he donated the $10,000 that he had received to Hungarian Spectrum, in order to raise international awareness of Hungary’s slide into authoritarian rule. In addition to her thousands of Hungarian Spectrum articles, Balogh also published guest columns in the Hungarian daily Népszava.When Balogh died suddenly of cardiac arrest on November 30, 2021, Hungarian Spectrum had over 7,000 subscribers. The site became a true vocation for Éva—her commitment and passion never wavered. While her sudden death leaves a void for thousands of loyal readers, Hungarian Spectrum is archived by the Vera and Donald Blinken Open Society Archives at Central European University and by the Library of Congress.2

The Plant Science Blogging Project: A curriculum to develop student science communication skills

Societal Impact StatementThe practice of writing science blogs benefits both the scientist and society alike by providing professional development opportunities and delivering information in a format that is accessible to large and diverse audiences. By designing a project that introduced upper‐level undergraduate students to science blog writing with a focus on plant biology, we piqued students' interest in science writing and the content of a popular plant science blog website. If adopted more widely, this work could broaden the scope of science education and promote the development of effective science communication skills for the next generation of scientists.Summary Successful scientists must communicate their research to broad audiences, including distilling key scientific concepts for the general public. Students pursuing careers in Science, Technology, Engineering, and Mathematics (STEM) fields benefit from developing public communication skills early in their careers, but opportunities are limited in traditional biology curricula. We created the “Plant Science Blogging Project” for a Plant Biology undergraduate course at the University of Pittsburgh in Fall 2018 and 2019. Students wrote blog posts merging personal connections with plants with plant biology concepts for the popular science blogs Plant Love Stories and EvoBites. By weaving biology into their narratives, students learned how to share botanical knowledge with the general public. The project had positive impacts on student learning and public engagement. In post‐assignment surveys, the majority of students reported that they enjoyed the assignment, felt it improved their understanding of plant biology, and piqued their interest in reading and writing science blogs in the future. Approximately one‐third of the student‐authored blogs were published, including two that rose to the top 10 most‐read posts on Plant Love Stories. Some dominant themes in student blogs, including medicine and culture, differed from common story themes published on the web, indicating the potential for students to diversify science blog content. Overall, the Plant Science Blogging Project allows undergraduate students to engage with plant biology topics in a new way, sharpen their scientific communication skills in accordance with today's world of mass information sharing, and contribute to the spread of scientific knowledge for public benefit.

배달 플랫폼 참여자의 경험주제 분석

Due to the spread of COVID-19, the number of delivery application users has exploded. As a result, the number of people participating as crowd workers on the delivery platform increased. The participation of individuals such as office workers, housewives, and students has increased. This can be understood as the expansion of the gig economy through the platform business. As such, the number of participants is increasing, but related academic discussions have not progressed properly. Therefore, this study analyzed the experiences of individuals who participated in the delivery platform as a partner. The data collection period is one year after COVID-19. In this study, online blog posts were collected through text mining. As a result of the analysis, a total of 10 topics were extracted. In this study, extracted keywords and collected original texts were checked to interpret the meaning of topics. The results of this study have the following implications. The limitations of previous studies are that there is a lack of platform participant-centered research, and the majority of studies have used qualitative methodologies. This study has academic significance in that it conducted a participant- centered study that applied the text mining technique. In addition, the results of this study are expected to contribute to enhancing the understanding of participants, who are key members of the platform business.

Ascriptions of migration: Racism, migratism and Brexit

This article offers an analysis of scholarly attempts to make sense of the nexus of race and migration in Brexit-era UK discourse. To illustrate my arguments that intend to challenge and extend existing scholarship, I discuss exemplary snapshots from news articles, blog posts and social media sources. Building on critical race and postcolonial studies as theoretical background, I trace the phenomenon of naming the discrimination against East Europeans – which is undeniably one of the driving forces of the Brexit discussion – ‘racism’. Sometimes, and this shows the pattern of the overgeneralization of the term, ‘racism’ gets extended to name the post-Brexit exclusion of any EU nationals. This use of ‘racism’, however, is based on methodological nationalism and conceptual whiteness as my analyses show. To make sense of the overlapping racist, anti-immigration and anti-EU rhetoric that marks the pre- and post-Brexit moment in the United Kingdom, this article introduces the concept of ‘migratism’ – a name for the power relation that ascribes migration to certain people, constructing them as migrants and discriminating against them. The terms ‘migratism’ and ‘migratization’ function grammatically in an analogue way to ‘racism’ and ‘racialization’. If racism is the power relation that racializes (=ascribes race to) people, migratism if the power relation that migratizes (=ascribes migration to) people. The terms are not symmetrical but have a complicated interdependent relationship. Racism and migratism are bound to each other and play a crucial role in organizing the Western nation state. The suggested concepts foreground a postcolonial understanding of race and racism and make it possible to analyse both migration-based discrimination and discrimination based on perceived migration in violence and hate crimes connected to the Brexit referendum.

Video-blogs and linguistic simplification for students with intellectual disability.

BackgroundThe Internet provides individuals with intellectual disability with access to information and participation in a broader society, but it also presents risks when content is difficult to comprehend. This study aimed to test whether students with intellectual disability enhanced their comprehension of online blogs as a function of the format (text vs. video) and linguistic simplification.MethodYoung students with intellectual disability read or watched text and video‐blog posts about environment, health, society, and technology in their original version or a linguistically simplified/easy‐to‐read version. Then, they completed tests that assessed their reading processing (e.g., calibration) and comprehension of blog content.ResultsParticipants predicted that they would correctly answer 80% of the comprehension questions, but their average comprehension of the blog posts was 55%, regardless of the condition. Previous levels of students' reading comprehension skills predicted their blog comprehension scores and interacted with format and simplification. Those participants with higher reading comprehension skills learned more from non‐simplified than from simplified blog posts while those with lower levels of reading comprehension skills did not benefit from linguistic simplification nor video‐blog format although the difference with regard to higher comprehenders was lower in the video and simplified conditions.ConclusionsImproving reading comprehension abilities of students with intellectual disability is essential to prevent the digital divide while linguistic simplification is not a useful accommodation and even counterproductive for higher reading comprehenders.

Security, Suspicion, and Surveillance? There’s an App for That

Despite the recent rise in popularity of mobile safety applications, social scientists have yet to examine these applications in any considerable depth. In this paper we undertake the case studies of bSafe, Citizen, and Nextdoor – analyzing promotional materials and blog posts – in order to further theorize digital security consumption and the potential concomitant social harms. We find these app companies frame crime and risk in ways that obscure the structural elements that precede crime and encourage social divisions. Drawing from over 30,000 user reviews, we speculate about the ways these apps might shape understandings, feelings, and experiences of risk, crime, and victimization. A closer examination of these apps is particularly urgent given these digital technologies have been mobilized in similar ways to respond to the COVID-19 pandemic.

Navigating grief and pregnancy loss through online story telling

Miscarriage, a common experience, is uncommonly discussed in personal and professional support networks. The normal reaction of grief from a miscarriage in addition to other external factors can lead to complicated grief, a condition that has negative mental health implications on the bereaved mother. Through the lens of Worden’s tasks of meaning, a content analysis was conducted on six public blog posts in which women openly shared their grief experiences. The authors discuss the need for more training in social work practice regarding reproductive health and the intersection of reproductive health and mental health.

Die Einführung von Gamification im DaF Unterricht an der Obafemi Awolowo Universität, Ile-Ife, Nigeria

Most Nigerian students of German are between 16 and 30 years old. This age group tends to be more interested in playing computer and video games than doing their home- and class work with traditional materials such as textbooks and workbooks. Younger students need interesting activities that are fun for them and motivate them to learn and, as a result, make progress and achieve rapid success in their learning. It is therefore surprising that gamification has not yet been introduced into the curricula of Nigerian higher institutions. As a result, there is hardly any research in the field of gamification in foreign language learning in general or its use in the teaching of German in particular. This lacuna in research on gamification in German studies in Nigeria led to the present study, which examines the introduction of gamification into German lessons at Obafemi Awolowo University. The study aims at initiating a targeted discussion on gamification in German studies and stimulating further scholarly conversations on the topic. Through this, the possibility of creating a didactic design for the use of gamification in German studies in Nigeria can be achieved. This study uses secondary data sources including articles, books and reviews about gamification in education or classrooms. Other sources include websites and blog posts. These sources provide useful insights from various studies on gamification and its use in education, especially in foreign language teaching, which help in drawing up suggestions for a successful introduction of gamification into the curriculum of Nigerian universities especially that of German at the Obafemi Awolowo University in Nigeria. The analysis of the data is based on a descriptive and content analysis. 

Extracting Multiple Worries From Breast Cancer Patient Blogs Using Multilabel Classification With the Natural Language Processing Model Bidirectional Encoder Representations From Transformers: Infodemiology Study of Blogs.

BackgroundPatients with breast cancer have a variety of worries and need multifaceted information support. Their accumulated posts on social media contain rich descriptions of their daily worries concerning issues such as treatment, family, and finances. It is important to identify these issues to help patients with breast cancer to resolve their worries and obtain reliable information.ObjectiveThis study aimed to extract and classify multiple worries from text generated by patients with breast cancer using Bidirectional Encoder Representations From Transformers (BERT), a context-aware natural language processing model.MethodsA total of 2272 blog posts by patients with breast cancer in Japan were collected. Five worry labels, “treatment,” “physical,” “psychological,” “work/financial,” and “family/friends,” were defined and assigned to each post. Multiple labels were allowed. To assess the label criteria, 50 blog posts were randomly selected and annotated by two researchers with medical knowledge. After the interannotator agreement had been assessed by means of Cohen kappa, one researcher annotated all the blogs. A multilabel classifier that simultaneously predicts five worries in a text was developed using BERT. This classifier was fine-tuned by using the posts as input and adding a classification layer to the pretrained BERT. The performance was evaluated for precision using the average of 5-fold cross-validation results.ResultsAmong the blog posts, 477 included “treatment,” 1138 included “physical,” 673 included “psychological,” 312 included “work/financial,” and 283 included “family/friends.” The interannotator agreement values were 0.67 for “treatment,” 0.76 for “physical,” 0.56 for “psychological,” 0.73 for “work/financial,” and 0.73 for “family/friends,” indicating a high degree of agreement. Among all blog posts, 544 contained no label, 892 contained one label, and 836 contained multiple labels. It was found that the worries varied from user to user, and the worries posted by the same user changed over time. The model performed well, though prediction performance differed for each label. The values of precision were 0.59 for “treatment,” 0.82 for “physical,” 0.64 for “psychological,” 0.67 for “work/financial,” and 0.58 for “family/friends.” The higher the interannotator agreement and the greater the number of posts, the higher the precision tended to be.ConclusionsThis study showed that the BERT model can extract multiple worries from text generated from patients with breast cancer. This is the first application of a multilabel classifier using the BERT model to extract multiple worries from patient-generated text. The results will be helpful to identify breast cancer patients’ worries and give them timely social support.

Representations of Latinos in a Democratic Party Campaign in the United States: Identity Ownership, Narratives, and Values

AbstractIn the 2016 United States presidential election, candidates Trump and Clinton embraced the demands of certain social groups and in this way, politically and symbolically, chose to “own” the social identities of these groups. Trump decided to attack the Latino community, while Clinton positioned herself as an advocate for this community. This article presents the results of a social narrative analysis of the values that Clinton and her team used to reach out to Latino communities during the 2016 election. The Spanish-language messages produced by the Democratic campaign compose the sample, which includes blog posts, Facebook posts, tweets, and television ads. Clinton’s campaign produced narratives about who the “good Latinos” are and, consequently, the “good immigrants” while at the same time promoting values such as globalism, cosmopolitanism, and multiculturalism. Paradoxically, these narratives and values failed to portray Latinos’ diversity because they left out this community’s historical, social, and cultural complexity.

After the Book, the Book? The Digital Writing Experiments of François Bon

While most commentators believe that the traditional print book will not be replaced by digital media, there has been considerable conjecture as to the possible effects on the forms of the book of so pervasive a revolution as has been ushered in by the digital revolution. One of the most sustained and searching investigation of these effects has been conducted by the French writer, François Bon. In his theoretical writings on the subject gathered in the book, Après le livre, Bon makes a strong case for the dependency of literary forms on the “material conditions of their enunciation.” Extrapolating from historical examples, he suggests that the book will accordingly undergo major, if yet unforeseen, transformations, while also arguing that it is urgent for writers to experiment with the possibilities generated by the material conditions brought about by the digital revolution, lest the field of possibilities be constrained by the commercial interests of the large technology companies. True to his sense of urgency, Bon himself has engaged in a number of experiments with the form of the book in the digital environment, most notably in his “novel” Tumulte, which consists of daily blog posts mixing fiction, memoir, criticism and other genres, and in a series of digital remediations of his early novel, Limite. While Bon’s experiments raise profound questions about the concept of the book, it is less certain whether they represent viable avenues for the book’s development, since their main appeal would arguably seem to be for scholars and theoreticians. With his concern with forms of organization and delivery, and ontological complexities arising from the multiple versions of the same or nearly the same text in a digital environment, Bon’s experiments seem necessarily to scant the power of concentrated prose that has been the principle means whereby books have conveyed thought and sustained intellectual culture.

Blogpost Website

As the name implies "Blog Post Website" is an educational website where there will be many blogs from a variety of topics. The main purpose of our website is to create a powerful platform where users can read various blogs and at the same time, users can publish their content or blogs on the website by creating an account, other users can log in / sign in. They can engage by doing likes, commenting on blogs. We will be adding this feature to increase engagement in our platform. Our Blogpost website is a website that is used for the chronological listing of the blog posts. Contains the most recent content first followed by previously updated content. In this there are many sections and posts written by a large number of authors. People can just create an account and start publishing their blogs. Blogs are a new and exciting way to communicate and to learn. Personal talk has evolved into a new and useful tool. Blogs have become an important form of emotional release and information for a growing number of people. With this website we are trying to enhance the learning perspective through blogs.

Intensive mothering and the perpetuation of gender inequality: Evidence from a mixed methods research

Intensive mothering is a cultural model of appropriate childrearing according to which mothers should unselfishly make a tremendous investment in their child. Using a mixed methodology, we examined the relevance of this ideology to understand the persistence of gender inequality. A content analysis of the most popular French mommy blogs indicates that this ideology remains commonplace, and has even incorporated contemporary concerns regarding sustainable development. Besides the expected themes of the sacredness of the child, the primary responsibility of the mother, and the use of intensive methods for all aspects of childrearing, the analysis of blog posts highlights new themes, including the sacredness of home, need for balance, and the praise of fathers. Furthermore, mommy blogs, as public online diaries involving everyday experience, prompt mothers to confess their failure to comply with intensive mothering demands and, at the same time, to reaffirm their commitment to its principles. Social influence is evidenced by the comments in response to the posts, which demonstrate polarization toward intensive mothering among the readers. A survey study further demonstrates that this ideology is positively related to a series of gender hierarchy-enhancing beliefs and attitudes. As a whole, the present research indicates that intensive mothering should be considered a system justifying ideology, while mommy blogs provide a platform for its diffusion and strengthening.

A mental state Knowledge–aware and Contrastive Network for early stress and depression detection on social media

Stress and depression detection on social media aim at the analysis of stress and identification of depression tendency from social media posts, which provide assistance for the early detection of mental health conditions. Existing methods mainly model the mental states of the post speaker implicitly. They also lack the ability to mentalise for complex mental state reasoning. Besides, they are not designed to explicitly capture class-specific features. To resolve the above issues, we propose a mental state Knowledge–aware and Contrastive Network (KC-Net). In detail, we first extract mental state knowledge from a commonsense knowledge base COMET, and infuse the knowledge using Gated Recurrent Units (GRUs) to explicitly model the mental states of the speaker. Then we propose a knowledge–aware mentalisation module based on dot-product attention to accordingly attend to the most relevant knowledge aspects. A supervised contrastive learning module is also utilised to fully leverage label information for capturing class-specific features. We test the proposed methods on a depression detection dataset Depression_Mixed with 3165 Reddit and blog posts, a stress detection dataset Dreaddit with 3553 Reddit posts, and a stress factors recognition dataset SAD with 6850 SMS-like messages. The experimental results show that our method achieves new state-of-the-art results on all datasets: 95.4% of F1 scores on Depression_Mixed, 83.5% on Dreaddit and 77.8% on SAD, with 2.07% average improvement. Factor-specific analysis and ablation study prove the effectiveness of all proposed modules, while UMAP analysis and case study visualise their mechanisms. We believe our work facilitates detection and analysis of depression and stress on social media data, and shows potential for applications on other mental health conditions.

Identification of hand-foot syndrome from cancer patients' blog posts: BERT-based deep-learning approach to detect potential adverse drug reaction symptoms.

Early detection and management of adverse drug reactions (ADRs) is crucial for improving patients’ quality of life. Hand-foot syndrome (HFS) is one of the most problematic ADRs for cancer patients. Recently, an increasing number of patients post their daily experiences to internet community, for example in blogs, where potential ADR signals not captured through routine clinic visits can be described. Therefore, this study aimed to identify patients with potential ADRs, focusing on HFS, from internet blogs by using natural language processing (NLP) deep-learning methods. From 10,646 blog posts, written in Japanese by cancer patients, 149 HFS-positive sentences were extracted after pre-processing, annotation and scrutiny by a certified oncology pharmacist. The HFS-positive sentences described not only HFS typical expressions like “pain" or “spoon nail”, but also patient-derived unique expressions like onomatopoeic ones. The dataset was divided at a 4 to 1 ratio and used to train and evaluate three NLP deep-learning models: long short-term memory (LSTM), bidirectional LSTM and bidirectional encoder representations from transformers (BERT). The BERT model gave the best performance with precision 0.63, recall 0.82 and f1 score 0.71 in the HFS user identification task. Our results demonstrate that this NLP deep-learning model can successfully identify patients with potential HFS from blog posts, where patients’ real wordings on symptoms or impacts on their daily lives are described. Thus, it should be feasible to utilize patient-generated text data to improve ADR management for individual patients.