Analysis of Twitter to Identify Topics Related to Eating Disorder Symptoms

Zhou, Sicheng; Zhao, Yunpeng; Rizvi, Rubina F.; Bian, Jiang; Haynos, Ann F.; Zhang, Rui

doi:10.1109/ichi.2019.8904863

Cited by 11 publications

(26 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared with our previous study [ 17 ], several new topics were identified, including Questions or Concerns , Reflection or Planning , Comorbidity , Ambivalence , Insults , and Diagnostic criteria. Textbox 3 shows these topic themes and example tweets.…”

Section: Resultsmentioning

confidence: 99%

“…The CorEx model identifies the topics that are maximally informative about a collection of documents [ 25 ]. The BTM and the CorEx model were tested in our preliminary study [ 17 ], and the CorEx model was adopted because it produced more meaningful topics using our collected tweets.…”

Section: Methodsmentioning

confidence: 99%

“…Although quantitative metrics are used to infer a reasonable number of topics (eg, perplexity and coherence), they sometimes cannot identify the optimal number of topics. On the basis of our experience in a previous study [ 17 ], we manually reviewed the topics produced by the different experiments to determine the optimal number of topics for further topic evaluation.…”

Section: Methodsmentioning

confidence: 99%

“…In 2019, Musaev et al [ 10 ] applied a latent Dirichlet allocation (LDA) topic model to tweets to examine public discussions about cardiovascular disease and found that state health departments play an important role in communicating with the public about cardiovascular health. We have also conducted a series of studies using Twitter data on various health-related topics, from detecting adverse events to studying laypeople’s discussion of human papillomavirus vaccination [ 11 - 17 ]. We have also conducted initial work on using Twitter data to identify discussion topics relevant to EDs.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Exploring Eating Disorder Topics on Twitter: Machine Learning Approach

Zhou¹,

Zhao²,

Bian³

et al. 2020

JMIR Med Inform

Self Cite

View full text Add to dashboard Cite

Background Eating disorders (EDs) are a group of mental illnesses that have an adverse effect on both mental and physical health. As social media platforms (eg, Twitter) have become an important data source for public health research, some studies have qualitatively explored the ways in which EDs are discussed on these platforms. Initial results suggest that such research offers a promising method for further understanding this group of diseases. Nevertheless, an efficient computational method is needed to further identify and analyze tweets relevant to EDs on a larger scale. Objective This study aims to develop and validate a machine learning–based classifier to identify tweets related to EDs and to explore factors (ie, topics) related to EDs using a topic modeling method. Methods We collected potential ED-relevant tweets using keywords from previous studies and annotated these tweets into different groups (ie, ED relevant vs irrelevant and then promotional information vs laypeople discussion). Several supervised machine learning methods, such as convolutional neural network (CNN), long short-term memory (LSTM), support vector machine, and naïve Bayes, were developed and evaluated using annotated data. We used the classifier with the best performance to identify ED-relevant tweets and applied a topic modeling method—Correlation Explanation (CorEx)—to analyze the content of the identified tweets. To validate these machine learning results, we also collected a cohort of ED-relevant tweets on the basis of manually curated rules. Results A total of 123,977 tweets were collected during the set period. We randomly annotated 2219 tweets for developing the machine learning classifiers. We developed a CNN-LSTM classifier to identify ED-relevant tweets published by laypeople in 2 steps: first relevant versus irrelevant (F1 score=0.89) and then promotional versus published by laypeople (F1 score=0.90). A total of 40,790 ED-relevant tweets were identified using the CNN-LSTM classifier. We also identified another set of tweets (ie, 17,632 ED-relevant and 83,557 ED-irrelevant tweets) posted by laypeople using manually specified rules. Using CorEx on all ED-relevant tweets, the topic model identified 162 topics. Overall, the coherence rate for topic modeling was 77.07% (1264/1640), indicating a high quality of the produced topics. The topics were further reviewed and analyzed by a domain expert. Conclusions A developed CNN-LSTM classifier could improve the efficiency of identifying ED-relevant tweets compared with the traditional manual-based method. The CorEx topic model was applied on the tweets identified by the machine learning–based classifier and the traditional manual approach separately. Highly overlapping topics were observed between the 2 cohorts of tweets. The produced topics were further reviewed by a domain expert. Some of the topics identified by the potential ED tweets may provide new avenues for understanding this serious set of disorders.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Exploring Eating Disorder Topics on Twitter: Machine Learning Approach

Zhou¹,

Zhao²,

Bian³

et al. 2020

JMIR Med Inform

Self Cite

View full text Add to dashboard Cite

show abstract

“…Such work demonstrated the capability of this novel technique to extract topics able to separate multiple facets within each analyzed discussion. Along this line, authors in [50] applied CorEx to tweets concerning eating disorders. They extracted novel topics and provided insights into factors that may foster the perpetuation of such behaviors, thus contributing to better understand -and possibly deal -with this serious issue.…”

Section: Invite Link Network Enrichment: Characterizing Online Crmentioning

confidence: 99%

Charting the Landscape of Online Cryptocurrency Manipulation

et al. 2020

View full text Add to dashboard Cite

Cryptocurrencies represent one of the most attractive markets for financial speculation. As a consequence, they have attracted unprecedented attention on social media. Besides genuine discussions and legitimate investment initiatives, several deceptive activities have flourished. In this work, we chart the online cryptocurrency landscape across multiple platforms. To reach our goal, we collected a large dataset, composed of more than 50M messages published by almost 7M users on Twitter, Telegram and Discord, over three months. We performed bot detection on Twitter accounts sharing invite links to Telegram and Discord channels, and we discovered that more than 56% of them were bots or suspended accounts. Then, we applied topic modeling techniques to Telegram and Discord messages, unveiling two different deception schemes-"pump-and-dump" and "Ponzi"-and identifying the channels involved in these frauds. Whereas on Discord we found a negligible level of deception, on Telegram we retrieved 296 channels involved in pump-anddump and 432 involved in Ponzi schemes, accounting for a striking 20% of the total. Moreover, we observed that 93% of the invite links shared by Twitter bots point to Telegram pump-and-dump channels, shedding light on a little-known social bot activity. Charting the landscape of online cryptocurrency manipulation can inform actionable policies to fight such abuse.

show abstract

Identifying hidden semantic structures in Instagram data: a topic modelling comparison

Egger

2021

View full text Add to dashboard Cite

Purpose Intrigued by the methodological challenges emerging from text complexity, the purpose of this study is to evaluate the effectiveness of different topic modelling algorithms based on Instagram textual data. Design/methodology/approach By taking Instagram posts captioned with #darktourism as the study context, this research applies latent Dirichlet allocation (LDA), correlation explanation (CorEx), and non-negative matrix factorisation (NMF) to uncover tourist experiences. Findings CorEx outperforms LDA and NMF by classifying emerging dark sites and activities into 17 distinct topics. The results of LDA appear homogeneous and overlapping, whereas the extracted topics of NMF are not specific enough to gain deep insights. Originality/value This study assesses different topic modelling algorithms for knowledge extraction in the highly heterogeneous tourism industry. The findings unfold the complexity of analysing short-text social media data and strengthen the use of CorEx in analysing Instagram content.

show abstract

Analysis of Twitter to Identify Topics Related to Eating Disorder Symptoms

Cited by 11 publications

References 13 publications

Exploring Eating Disorder Topics on Twitter: Machine Learning Approach

Exploring Eating Disorder Topics on Twitter: Machine Learning Approach

Charting the Landscape of Online Cryptocurrency Manipulation

Identifying hidden semantic structures in Instagram data: a topic modelling comparison

Contact Info

Product

Resources

About