Automatic Depression Score Estimation with Word Embedding Models (Preprint)

2021 ◽  
Author(s):  
Anxo Pérez ◽  
Javier Parapar ◽  
Alvaro Barreiro
Keyword(s):  
Social Media ◽  
Question Answering ◽  
Rating Scales ◽  
Mixed Model ◽  
Depression Score ◽  
Self Report ◽  
Training Data ◽  
Language Models ◽  
Patients At Risk ◽  
Almost All

BACKGROUND Depression is one of the most common mental health illnesses. Despite existing effective treatments, the biggest obstacle lies in an efficient and early detection of the disorder. Self-report questionnaires are the instruments exploited to elaborate a diagnosis by medical experts. However, questionnaires often encounter certain limitations. Factors such as the lack of awareness and social stigmas negatively affect the success of self-report questionnaires. In this context, social media platforms provide non-direct means of communication capable of being a source of evidence to detect patients at risk. OBJECTIVE This paper aims to describe techniques to automatically estimate the degree of depression from users on social media. We aimed to explore neural language models to exploit various aspects of the subject's writings. Our proposals have focused on automatically completing the Beck Depression Inventory-II (BDI-II). BDI-II is a validated psychometric test consisting of 21 items, each one associated with a different symptom of depression. METHODS We presented three approaches for automatically filling the BDI-II questionnaire based on neural language models. The first proposal captures the overall use of language and communication patterns evidenced by individuals. In the second proposal, we narrow the user's representation by only using limited extracted answers from their posts to the items in the BDI-II. For that, we use state-of-the-art Question Answering models based on bidirectional encoder representations. Finally, we propose a mixed model that selects whether to automatically fill an item using the first or the second model. The rationale behind the mixed model is that, on the one hand, users easily comment the answer to some items in their texts, which made the second method appropriate. On the other hand, on more private or sensitive items, the first method is the best alternative, given that users avoid writing about them explicitly. RESULTS We addressed the task "Measuring the Severity of the Signs of Depression" of eRisk 2020, an initiative in the CLEF Conference. In this task, the participants have to fill in the BDI-II for the collection delivered by the task. We measured our results using the same accuracy metrics proposed by the competition. We compared them with the rest of the 17 methods presented by participants. Our proposals outperformed almost all participants for every official metric. CONCLUSIONS Our results showed that techniques based on neural language models are a feasible alternative for estimating rating scales for depression, even when small amounts of training data are available (20 users). We observe that depending on the symptom, it will be more appropriate to use general language patterns or looking for direct concerns about the particular symptom. In summary, the results of this study have demonstrated the potential of automatic text mining models to serve as a tool helping to diagnose depression disease.

scholarly journals Multiple-Choice Question Answering Models for Automatic Depression Severity Estimation

2021 ◽  
Vol 7 (1) ◽  
pp. 23
Author(s):  
Jorge Gabín ◽  
Anxo Pérez ◽  
Javier Parapar
Keyword(s):  
Social Media ◽  
Question Answering ◽  
Multiple Choice ◽  
Multiple Choice Question ◽  
Self Report ◽  
Training Data ◽  
Language Models ◽  
Depression Severity ◽  
Suitable Alternative ◽  
Severity Estimation

Depression is one of the most prevalent mental health diseases. Although there are effective treatments, the main problem relies on providing early and effective risk detection. Medical experts use self-reporting questionnaires to elaborate their diagnosis, but these questionnaires have some limitations. Social stigmas and the lack of awareness often negatively affect the success of these self-report questionnaires. This article aims to describe techniques to automatically estimate the depression severity from users on social media. We explored the use of pre-trained language models over the subject’s writings. We addressed the task “Measuring the Severity of the Signs of Depression” of eRisk 2020, an initiative in the CLEF Conference. In this task, participants have to fill the Beck Depression Questionnaire (BDI-II). Our proposal explores the application of pre-trained Multiple-Choice Question Answering (MCQA) models to predict user’s answers to the BDI-II questionnaire using their posts on social media. These MCQA models are built over the BERT (Bidirectional Encoder Representations from Transformers) architecture. Our results showed that multiple-choice question answering models could be a suitable alternative for estimating the depression degree, even when small amounts of training data are available (20 users).

scholarly journals QASC: A Dataset for Question Answering via Sentence Composition

2020 ◽  
Vol 34 (05) ◽  
pp. 8082-8090
Author(s):  
Tushar Khot ◽  
Peter Clark ◽  
Michal Guerquin ◽  
Peter Jansen ◽  
Ashish Sabharwal
Keyword(s):  
Common Sense ◽  
Human Performance ◽  
Question Answering ◽  
State Of The Art ◽  
Multiple Choice ◽  
Training Data ◽  
Language Models ◽  
Current State ◽  
New Concepts ◽  
Large Corpus

Composing knowledge from multiple pieces of texts is a key challenge in multi-hop question answering. We present a multi-hop reasoning dataset, Question Answering via Sentence Composition (QASC), that requires retrieving facts from a large corpus and composing them to answer a multiple-choice question. QASC is the first dataset to offer two desirable properties: (a) the facts to be composed are annotated in a large corpus, and (b) the decomposition into these facts is not evident from the question itself. The latter makes retrieval challenging as the system must introduce new concepts or relations in order to discover potential decompositions. Further, the reasoning model must then learn to identify valid compositions of these retrieved facts using common-sense reasoning. To help address these challenges, we provide annotation for supporting facts as well as their composition. Guided by these annotations, we present a two-step approach to mitigate the retrieval challenges. We use other multiple-choice datasets as additional training data to strengthen the reasoning model. Our proposed approach improves over current state-of-the-art language models by 11% (absolute). The reasoning and retrieval problems, however, remain unsolved as this model still lags by 20% behind human performance.

scholarly journals Feature Enhancement in Attention for Visual Question Answering

2018 ◽  
Author(s):  
Yuetan Lin ◽  
Zhangyang Pang ◽  
Donghui Wang ◽  
Yueting Zhuang
Keyword(s):  
Question Answering ◽  
Semantic Representation ◽  
Attention Mechanism ◽  
Training Data ◽  
Feature Enhancement ◽  
Visual Question Answering ◽  
Textual Features ◽  
Almost All ◽  
Region Representation ◽  
Different Sources

Attention mechanism has been an indispensable part of Visual Question Answering (VQA) models, due to the importance of its selective ability on image regions and/or question words. However, attention mechanism in almost all the VQA models takes as input the image visual and question textual features, which stem from different sources and between which there exists essential semantic gap. In order to further improve the accuracy of correlation between region and question in attention, we focus on region representation and propose the idea of feature enhancement, which includes three aspects. (1) We propose to leverage region semantic representation which is more consistent with the question representation. (2) We enrich the region representation using features from multiple hierarchies and (3) we refine the semantic representation for richer information. With these three incremental feature enhancement mechanisms, we improve the region representation and achieve better attentive effect and VQA performance. We conduct extensive experiments on the largest VQA v2.0 benchmark dataset and achieve competitive results without additional training data, and prove the effectiveness of our proposed feature-enhanced attention by visual demonstrations.

scholarly journals SEED: Symptom Extraction from English Social Media Posts using Deep Learning and Transfer Learning

2021 ◽  
Author(s):  
Arjun Magge ◽  
Karen O’Connor ◽  
Matthew Scotch ◽  
Gonzalez-Hernandez Graciela
Keyword(s):  
Social Media ◽  
Deep Learning ◽  
Language Processing ◽  
Viral Infections ◽  
Drug Effects ◽  
Self Report ◽  
Training Data ◽  
Social Media Mining ◽  
Study Results ◽  
Media Mining

AbstractThe increase of social media usage across the globe has fueled efforts in public health research for mining valuable information such as medication use, adverse drug effects and reports of viral infections that directly and indirectly affect human health. Despite its significance, such information can be incredibly rare on social media. Mining such non-traditional sources for disease monitoring requires natural language processing techniques for extracting symptom mentions and normalizing them to standard terminologies for interpretability. In this work, we present the first version of a social media mining tool called SEED that detects symptom and disease mentions from social media posts such as Twitter and DailyStrength and further normalizes them into the UMLS terminology. Using multi-corpus training and deep learning models, the tool achieves an overall F1 score of 0.85 for extracting mentions of symptoms on a health forum dataset and an F1 score of 0.72 on a balanced Twitter dataset significantly improving over previously systems on the datasets. We apply the tool on recently collected Twitter posts that self-report COVID19 symptoms to observe if the SEED system can extract novel diseases and symptoms that were absent in the training data. By doing so, we describe the advantages and shortcomings of the tool and suggest techniques to overcome the limitations. The study results also draw attention to the potential of multi-corpus training for performance improvements and the need for continual training on newly obtained data for consistent performance amidst the ever-changing nature of the social media vocabulary.

scholarly journals Optical character recognition with neural networks and post-correction with finite state methods

2020 ◽  
Vol 23 (4) ◽  
pp. 279-295
Author(s):  
Senka Drobac ◽  
Krister Lindén
Keyword(s):  
Neural Networks ◽  
Character Recognition ◽  
Optical Character Recognition ◽  
Mixed Model ◽  
Training Data ◽  
Language Models ◽  
High Quality ◽  
Link Type ◽  
Optical Character ◽  
Mixed Language

Abstract The optical character recognition (OCR) quality of the historical part of the Finnish newspaper and journal corpus is rather low for reliable search and scientific research on the OCRed data. The estimated character error rate (CER) of the corpus, achieved with commercial software, is between 8 and 13%. There have been earlier attempts to train high-quality OCR models with open-source software, like Ocropy (https://github.com/tmbdev/ocropy) and Tesseract (https://github.com/tesseract-ocr/tesseract), but so far, none of the methods have managed to successfully train a mixed model that recognizes all of the data in the corpus, which would be essential for an efficient re-OCRing of the corpus. The difficulty lies in the fact that the corpus is printed in the two main languages of Finland (Finnish and Swedish) and in two font families (Blackletter and Antiqua). In this paper, we explore the training of a variety of OCR models with deep neural networks (DNN). First, we find an optimal DNN for our data and, with additional training data, successfully train high-quality mixed-language models. Furthermore, we revisit the effect of confidence voting on the OCR results with different model combinations. Finally, we perform post-correction on the new OCR results and perform error analysis. The results show a significant boost in accuracy, resulting in 1.7% CER on the Finnish and 2.7% CER on the Swedish test set. The greatest accomplishment of the study is the successful training of one mixed language model for the entire corpus and finding a voting setup that further improves the results.

Articles Published in EJPA 2009–2010

2011 ◽  
Vol 27 (4) ◽  
pp. 290-298 ◽  
Author(s):  
Tuulia M. Ortner ◽  
Isabella Vormittag
Keyword(s):  
Psychological Assessment ◽  
Rating Scales ◽  
Self Report ◽  
Related Field ◽  
Broad Scope ◽  
Future Challenges ◽  
Health Related ◽  
Report Data ◽  
Self Report Data

With reference to EJPA’s unique and broad scope, the current study analyzed the characteristics of the authors as well as the topics and research aims of the 69 empirical articles published in the years 2009–2010. Results revealed that more than one third of the articles were written by authors affiliated with more than one country. With reference to their research aims, an almost comparable number of articles (1) presented a new measure, (2) dealt with adaptations of measures, or (3) dealt with further research on existing measures. Analyses also revealed that most articles did not address any particular field of application. The second largest group was comprised of articles related to the clinical field, followed by the health-related field of application. The majority of all articles put their focus on investigating questionnaires or rating scales, and only a small number of articles investigated procedures classified as tests or properties of interviews. As to further characteristics of the method(s) used, a majority of EJPA contributions addressed self-report data. Results are discussed with reference to publication demands as well as the current and future challenges and demands of psychological assessment.

Differences in Reward and Persistency characterise Behavior in Chronic Pain for the Iowa Gambling Task in a Web-based Lab-in-the-field study (Preprint)

2020 ◽  
Author(s):  
Lili Zhang ◽  
Himanshu Vashisht ◽  
Alekhya Nethra ◽  
Brian Slattery ◽  
Tomas Ward
Keyword(s):  
Decision Making ◽  
Social Media ◽  
Chronic Pain ◽  
Iowa Gambling Task ◽  
Learning Curves ◽  
Self Report ◽  
Healthy Controls ◽  
Gambling Task ◽  
Convenience Sample ◽  
Pain Experiences

BACKGROUND Chronic pain is a significant world-wide health problem. It has been reported that people with chronic pain experience decision-making impairments, but these findings have been based on conventional lab experiments to date. In such experiments researchers have extensive control of conditions and can more precisely eliminate potential confounds. In contrast, there is much less known regarding how chronic pain impacts decision-making captured via lab-in-the-field experiments. Although such settings can introduce more experimental uncertainty, it is believed that collecting data in more ecologically valid contexts can better characterize the real-world impact of chronic pain. OBJECTIVE We aim to quantify decision-making differences between chronic pain individuals and healthy controls in a lab-in-the-field environment through taking advantage of internet technologies and social media. METHODS A cross-sectional design with independent groups was employed. A convenience sample of 45 participants were recruited through social media - 20 participants who self-reported living with chronic pain, and 25 people with no pain or who were living with pain for less than 6 months acting as controls. All participants completed a self-report questionnaire assessing their pain experiences and a neuropsychological task measuring their decision-making, i.e. the Iowa Gambling Task (IGT) in their web browser at a time and location of their choice without supervision. RESULTS Standard behavioral analysis revealed no differences in learning strategies between the two groups although qualitative differences could be observed in learning curves. However, computational modelling revealed that individuals with chronic pain were quicker to update their behavior relative to healthy controls, which reflected their increased learning rate (95% HDI from 0.66 to 0.99) when fitted with the VPP model. This result was further validated and extended on the ORL model because higher differences (95% HDI from 0.16 to 0.47) between the reward and punishment learning rates were observed when fitted on this model, indicating that chronic pain individuals were more sensitive to rewards. It was also found that they were less persistent in their choices during the IGT compared to controls, a fact reflected by their decreased outcome perseverance (95% HDI from -4.38 to -0.21) when fitted using the ORL model. Moreover, correlation analysis revealed that the estimated parameters had predictive value for the self-reported pain experiences, suggesting that the altered cognitive parameters could be potential candidates for inclusion in chronic pain assessments. CONCLUSIONS We found that individuals with chronic pain were more driven by rewards and less consistent when making decisions in our lab-in-the-field experiment. In this case study, it was demonstrated that compared to standard statistical summaries of behavioral performance, computational approaches offered superior ability to resolve, understand and explain the differences in decision- making behavior in the context of chronic pain outside the lab.

A comparison of the Remote Food Photography Method and the Automated Self-Administered 24-hour Dietary Assessment Tool for measuring full day dietary intake among school-aged children

2021 ◽  
pp. 1-26
Author(s):  
Traci A. Bekelman ◽  
Corby K. Martin ◽  
Susan L. Johnson ◽  
Deborah H. Glueck ◽  
Katherine A. Sauder ◽  
...  
Keyword(s):  
Dietary Intake ◽  
Energy Intake ◽  
Mixed Model ◽  
Linear Mixed Model ◽  
Assessment Tool ◽  
Dietary Assessment ◽  
Ease Of Use ◽  
Self Report ◽  
Food Photography ◽  
Photography Method

Abstract The limitations of self-report measures of dietary intake are well known. Novel, technology-based measures of dietary intake may provide a more accurate, less burdensome alternative to existing tools. The first objective of this study was to compare participant burden for two technology-based measures of dietary intake among school-age children: the Automated-Self Administered 24-hour Dietary Assessment Tool-2018 (ASA24-2018) and the Remote Food Photography Method (RFPM). The second objective was to compare reported energy intake for each method to the Estimated Energy Requirement for each child, as a benchmark for actual intake. Forty parent-child dyads participated in 2, 3-day dietary assessments: a parent proxy-reported version of the ASA24 and the RFPM. A parent survey was subsequently administered to compare satisfaction, ease of use and burden with each method. A linear mixed model examined differences in total daily energy intake (TDEI) between assessments, and between each assessment method and the EER. Reported energy intake was 379 kcal higher with the ASA24 than the RFPM (p=0.0002). Reported energy intake with the ASA24 was 231 kcal higher than the EER (p = 0.008). Reported energy intake with the RFPM did not differ significantly from the EER (difference in predicted means = −148 kcal, p = 0.09). Median satisfaction and ease of use scores were 5 out of 6 for both methods. A higher proportion of parents reported that the ASA24 was more time consuming than the RFPM (74.4% vs. 25.6%, p = 0.002). Utilization of both methods is warranted given their high satisfaction among parents.

scholarly journals ‘Ruined’ lives: Mediated white male victimhood

2021 ◽  
Vol 24 (1) ◽  
pp. 60-80
Author(s):  
Sarah Banet-Weiser
Keyword(s):  
Social Media ◽  
Sexual Harassment ◽  
Sexual Violence ◽  
Public Perception ◽  
White Male ◽  
The Public ◽  
Multi Media ◽  
Patriarchal Power ◽  
Almost All ◽  
False Accusations

When the hashtag #metoo began to circulate in digital and social media, it challenged a familiar interpretation of those who are raped or sexually harassed as victims, positioning women as embodied agents. Yet, almost exactly a year after the #metoo movement shot to visible prominence, a different, though eerily similar, story began to circulate on the same multi-media platforms as #metoo: a story about white male victimhood. Powerful men in positions of privilege (almost always white) began to take up the mantle of victimhood as their own, often claiming to be victims of false accusations of sexual harassment and assault by women. Through the analysis of five public statements by highly visible, powerful men who have been accused of sexual violence, I argue that the discourse of victimhood is appropriated not by those who have historically suffered but by those in positions of patriarchal power. Almost all of the statements contain some sentiment about how the accusation (occasionally acknowledging the actual violence) ‘ruined their life’, and all of the statements analyzed here center the author, the accused white man, as the key subject in peril and the authors position themselves as truth-tellers about the incidents. These statements underscore certain shifts in the public perception of sexual violence; the very success of the #metoo movement in shifting the narrative has meant that men have had to defend themselves more explicitly in public. In order to wrestle back a hegemonic gender stability, these men take on the mantle of victimhood themselves.

Neurodevelopmental disorders among Spanish school-age children: prevalence and sociodemographic correlates

2021 ◽  
pp. 1-11
Author(s):  
Rosa Bosch ◽  
Mireia Pagerols ◽  
Cristina Rivas ◽  
Laura Sixto ◽  
Laura Bricollé ◽  
...  
Keyword(s):  
Neurodevelopmental Disorders ◽  
Rating Scales ◽  
Communication Disorders ◽  
Autism Spectrum ◽  
Self Report ◽  
Motor Disorder ◽  
School Type ◽  
Final Population ◽  
Prevalence Estimates ◽  
Sociodemographic Correlates

Abstract Background Prevalence estimates of neurodevelopmental disorders (ND) are essential for treatment planning. However, epidemiological research has yielded highly variable rates across countries, including Spain. This study examined the prevalence and sociodemographic correlates of ND in a school sample of Spanish children and adolescents. Methods The Child Behaviour Checklist/Teacher's Report Form/Youth Self-Report and the Conners' Rating Scales were administered for screening purposes. Additionally, teachers provided information on reading and writing difficulties. Subjects who screened positive were interviewed for diagnostic confirmation according to the Diagnostic and Statistical Manual of Mental Disorders criteria. The final population comprised 6834 students aged 5–17. Multivariate analyses were performed to determine the influence of gender, age, educational stage, school type, socioeconomic status (SES), and ethnicity on the prevalence estimates. Results A total of 1249 (18.3%) subjects met criteria for at least one ND, although only 423 had already received a diagnosis. Specifically, the following prevalence rates were found: intellectual disabilities (ID), 0.63%; communication disorders, 1.05%; autism spectrum disorder (ASD), 0.70%; attention-deficit/hyperactivity disorder (ADHD), 9.92%; specific learning disorder (SLD), 10.0%; and motor disorders, 0.76%. Students of foreign origin and from low SES evidenced higher odds of having ID. Boys were more likely to display ASD or a motor disorder. Age, SES, and ethnicity were significant predictors for SLD, while communication disorders and ADHD were also associated with gender. Conclusions The prevalence of ND among Spanish students is consistent with international studies. However, a substantial proportion had never been previously diagnosed, which emphasise the need for early detection and intervention programmes.

Sign in / Sign up

Export Citation Format

Share Document