Towards Ordinal Suicide Ideation Detection on Social Media

Sawhney, Ravinder Singh; Joshi, Harshit; Gandhi, Saumya; Shah, Rajiv Ratn

doi:10.1145/3437963.3441805

Cited by 31 publications

(25 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Naseem et al (2021b) presented a model that uses domain-specific LM and captures commonsense knowledge into a context-aware bidirectional gated recurrent network. Sawhney et al (2021) presented an ordinal hierarchical attention model for Suicide Risk Assessment where text embeddings obtained by Longformer were fed to BiL-STM with attention and ordinal loss as an objective function. However, there is no PLM trained on health-related text collected from social media that directly benefit the applications related to PHS.…”

Section: Nlp For Public Health Surveillancementioning

confidence: 99%

Benchmarking for Public Health Surveillance tasks on Social Media with a Domain-Specific Pretrained Language Model

Naseem¹,

Lee²,

Khushi³

et al. 2022

Preprint

View full text Add to dashboard Cite

A user-generated text on social media enables health workers to keep track of information, identify possible outbreaks, forecast disease trends, monitor emergency cases, and ascertain disease awareness and response to official health correspondence. This exchange of health information on social media has been regarded as an attempt to enhance public health surveillance (PHS). Despite its potential, the technology is still in its early stages and is not ready for widespread application. Advancements in pretrained language models (PLMs) have facilitated the development of several domain-specific PLMs and a variety of downstream applications. However, there are no PLMs for social media tasks involving PHS. We present and release PHS-BERT, a transformer-based PLM, to identify tasks related to public health surveillance on social media. We compared and benchmarked the performance of PHS-BERT on 25 datasets from different social medial platforms related to 7 different PHS tasks. Compared with existing PLMs that are mainly evaluated on limited tasks, PHS-BERT achieved state-ofthe-art performance on all 25 tested datasets, showing that our PLM is robust and generalizable in the common PHS tasks. By making PHS-BERT available 1 , we aim to facilitate the community to reduce the computational cost and introduce new baselines for future works across various PHS-related tasks.

show abstract

Section: Nlp For Public Health Surveillancementioning

confidence: 99%

Benchmarking for Public Health Surveillance tasks on Social Media with a Domain-Specific Pretrained Language Model

Naseem¹,

Lee²,

Khushi³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Researchers have investigated that user activity data on social media can provide a cue for analyzing individual suicidality (De Choudhury et al, 2016;Shing et al, 2018). Specifically, prior research showed that linguistic characteristics revealed in social media posts (Sawhney et al, 2020(Sawhney et al, , 2021a) could be linked to suicidal ideation. In particular, utilizing suicide dictionaries made by domain experts has been demonstrated as effective (Lv et al, 2015;Cao et al, 2019;Gaur et al, 2019;Lee et al, 2020), and such lexicon-based methods are known to be fast, explainable, and easy to implement (Kotelnikova et al, 2021;Razova et al, 2021).…”

Section: Suicidality Assessment With Suicide Lexiconmentioning

confidence: 99%

“…Like Sawhney et al (2021a), we adopt the ordinal regression loss (Diaz and Marathe, 2019) as an objective function. Instead of using an onehot vector representation of the true labels, they used a soft encoded vector representation by considering the ordinal nature between suicidality levels.…”

Section: Suicidality Detection Decodermentioning

confidence: 99%

“…We compare the proposed model against the following three types of models: (1) Lexicon-based approaches; Rule-based (Gaur et al, 2019), SVM (Lv et al, 2015), and Random Forest (RF) (Amini et al, 2016), (2) Deep learning approaches w/o lexicon; Contextual CNN (Gaur et al, 2019), SISMO (Sawhney et al, 2021a), and BERT (Devlin et al, 2018), and (3) Lexicon + deep learning; Cao et al (2019) and Reformed BERT. Detailed experimental settings for reproducibility are summarized in the Appendix ?…”

Section: Baselines and Experiments Settingsmentioning

confidence: 99%

See 1 more Smart Citation

Detecting Suicidality with a Contextual Graph Neural Network

Lee¹,

Kang²,

Kim³

et al. 2022

Proceedings of the Eighth Workshop on Computational Linguistics and Clinical Psychology

View full text Add to dashboard Cite

Discovering individuals' suicidality on social media has become increasingly important. Many researchers have studied to detect suicidality by using a suicide dictionary. However, while prior work focused on matching a word in a post with a suicide dictionary without considering contexts, little attention has been paid to how the word can be associated with the suicide-related context. To address this problem, we propose a suicidality detection model based on a graph neural network to grasp the dynamic semantic information of the suicide vocabulary by learning the relations between a given post and words. The extensive evaluation demonstrates that the proposed model achieves higher performance than the state-of-the-art methods. We believe the proposed model has great utility in identifying the suicidality of individuals and hence preventing individuals from potential suicide risks at an early stage.

show abstract

“…how likely it is that a passenger aircraft will be shot along its flight path. Social media has already been used in many applications to assess risk during a public health crisis (Gui et al , 2017), suicide risk estimation via online postings (Sawhney et al , 2021), assessing the risk crisis via communications (Veil et al , 2011), disaster early warning (Wu and Cui, 2018), assess the risk of disasters (Kryvasheyeu, 2016), etc.…”

Section: Introductionmentioning

confidence: 99%

Using social media for flight path safety assessment

Abin

Nabavi

Moghaddam

2021

AEAT

View full text Add to dashboard Cite

Purpose Artificial intelligence (AI)-based systems can save the lives of many people by assessing the safety of flight paths. Unfortunately, the world witnessed a horrible event in January 2020 with the case of flight 752 of Ukrainian International Airlines from Tehran to Kiev and it has prompted us to ask how AI can prevent such events by warning to flight path planners. This paper aims to propose a framework for assessing the safety of flight paths from a shooting of an airplane by air defense systems installed on the path. Unlike the existing studies, this study takes a new look at pre-flight risk assessment by using textual information in social and news networks. To this end, the authors use existing information retrieval techniques to identify high flight risk areas by examining the news articles, comments, posts, tweets, etc., in social media and then estimate the probability of targeting a passenger aircraft by the air defense systems probably installed on high-risk areas with the help of a statistical model. This estimation can then be used by fight planners to avoid such events. Design/methodology/approach To design a framework for estimating the probability of a fatal shooting of an airplane by air defense systems installed on its flight path, the authors have used the idea of information retrieval in conjunction with statistical methods. The authors have extracted some significant variables in the shooting of flights and proposed an AI-based framework to estimate the probability of a fatal shooting of an airplane during its flight and sketched a case study for using machine learning approaches to assist with flight path planning. As a case study, the authors covered flight 752 to explain the usefulness of the proposed framework in this context. Findings Unlike the existing methods, this study investigates flight path safety assessment from the social media and crowdsourcing perspective. In this study, the authors proposed an AI-based framework to avoid aviation hazards by estimating the probability of a shooting of an airplane by air defense systems installed on its flight path. Moreover, this study was designed to show how estimating the safety of flight paths by using AI-based methods can help flight planners to avoid such events and gain further insights into the use of AI-based systems in pre-flight risk assessment. Originality/value The idea behind the proposed method is original and as the authors’ best knowledge, there is no similar framework using social media for flight path safety assessment.

show abstract

Towards Ordinal Suicide Ideation Detection on Social Media

Cited by 31 publications

References 22 publications

Benchmarking for Public Health Surveillance tasks on Social Media with a Domain-Specific Pretrained Language Model

Benchmarking for Public Health Surveillance tasks on Social Media with a Domain-Specific Pretrained Language Model

Detecting Suicidality with a Contextual Graph Neural Network

Using social media for flight path safety assessment

Contact Info

Product

Resources

About