Sentiment analysis methods for understanding large-scale texts: a case for using continuum-scored words and word shift graphs

Reagan, Andrew J.; Danforth, Christopher M.; Tivnan, Brian F.; Williams, Jake Ryland; Dodds, Peter Sheridan

doi:10.1140/epjds/s13688-017-0121-9

Cited by 67 publications

(61 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The labMT happiness measure proved to be the most important predictor in our model, and was considerably stronger than ANEW or LIWC happiness indicators. This is in line with a series of previous findings, which have found labMT measures to be a superior for tracking happiness in Twitter data (39), and suggests that future research in this field should incorporate this instrument for more accurate measurements. That average tweet word count was the second most important predictor is intriguing, especially as increases in word count were positively associated with depression and PTSD.…”

Section: Discussionsupporting

confidence: 90%

Forecasting the onset and course of mental illness with Twitter data

Reece

Reagan

Lix

et al. 2017

Sci Rep

Self Cite

302

226

View full text Add to dashboard Cite

We developed computational models to predict the emergence of depression and Post-Traumatic Stress Disorder in Twitter users. Twitter data and details of depression history were collected from 204 individuals (105 depressed, 99 healthy). We extracted predictive features measuring affect, linguistic style, and context from participant tweets (N = 279,951) and built models using these features with supervised learning algorithms. Resulting models successfully discriminated between depressed and healthy content, and compared favorably to general practitioners’ average success rates in diagnosing depression, albeit in a separate population. Results held even when the analysis was restricted to content posted before first depression diagnosis. State-space temporal analysis suggests that onset of depression may be detectable from Twitter data several months prior to diagnosis. Predictive results were replicated with a separate sample of individuals diagnosed with PTSD (Nusers = 174, Ntweets = 243,775). A state-space time series model revealed indicators of PTSD almost immediately post-trauma, often many months prior to clinical diagnosis. These methods suggest a data-driven, predictive approach for early screening and detection of mental illness.

show abstract

Section: Discussionsupporting

confidence: 90%

Forecasting the onset and course of mental illness with Twitter data

Reece

Reagan

Lix

et al. 2017

Sci Rep

Self Cite

302

226

View full text Add to dashboard Cite

show abstract

“…It has been suggested that, at a molar level, the emotional tone of communication is affected mainly by the most strongly valenced words (e.g., Dodds & Danforth, 2009). For example, when evaluating large samples of printed text, Reagan et al (2015) focused mainly on words that were rated below 4 and above 6 on a nine-point "happy" scale. Applying the same heuristic to our lists of terms shows that strongly positive words outnumbered strongly negative words in the general clinical (12:9), general science (9:4), and behavioral assessment (7:5) categories.…”

Section: Resultsmentioning

confidence: 99%

Normative Emotional Responses to Behavior Analysis Jargon or How Not to Use Words to Win Friends and Influence People

Critchfield

Doepke

Epting

et al. 2017

Behav Analysis Practice

View full text Add to dashboard Cite

It has been suggested that non-experts regard the jargon of behavior analysis as abrasive, harsh, and unpleasant. If this is true, excessive reliance on jargon could interfere with the dissemination of effective services. To address this often discussed but rarely studied issue, we consulted a large, public domain list of English words that have been rated by members of the general public for the emotional reactions they evoke. Selected words that behavior analysts use as technical terms were compared to selected words that are commonly used to discuss general science, general clinical work, and behavioral assessment. There was a tendency for behavior analysis terms to register as more unpleasant than other kinds of professional terms and also as more unpleasant than English words generally. We suggest possible reasons for this finding, discuss its relevance to the challenge of deciding how to communicate with consumers who do not yet understand or value behavior analysis, and advocate for systematic research to guide the marketing of behavior analysis.

show abstract

“…Sentiment analysis of written text is a widely studied problem in natural language processing [39,40,41]. In this study, we have considered sentiment in a simple formthe presence of positive or negative affectand applied SentiStrength [42], a widely used open-source Java library designed for sentiment analysis of tweets.…”

Section: Sentiment Measuresmentioning

confidence: 99%

Modeling Spatiotemporal Factors Associated With Sentiment on Twitter: Synthesis and Suggestions for Improving the Identification of Localized Deviations

Shah¹,

Martin²,

Coiera³

et al. 2019

J Med Internet Res

View full text Add to dashboard Cite

Background Studies examining how sentiment on social media varies depending on timing and location appear to produce inconsistent results, making it hard to design systems that use sentiment to detect localized events for public health applications. Objective The aim of this study was to measure how common timing and location confounders explain variation in sentiment on Twitter. Methods Using a dataset of 16.54 million English-language tweets from 100 cities posted between July 13 and November 30, 2017, we estimated the positive and negative sentiment for each of the cities using a dictionary-based sentiment analysis and constructed models to explain the differences in sentiment using time of day, day of week, weather, city, and interaction type (conversations or broadcasting) as factors and found that all factors were independently associated with sentiment. Results In the full multivariable model of positive (Pearson r in test data 0.236; 95% CI 0.231-0.241) and negative (Pearson r in test data 0.306; 95% CI 0.301-0.310) sentiment, the city and time of day explained more of the variance than weather and day of week. Models that account for these confounders produce a different distribution and ranking of important events compared with models that do not account for these confounders. Conclusions In public health applications that aim to detect localized events by aggregating sentiment across populations of Twitter users, it is worthwhile accounting for baseline differences before looking for unexpected changes.

show abstract

Sentiment analysis methods for understanding large-scale texts: a case for using continuum-scored words and word shift graphs

Cited by 67 publications

References 33 publications

Forecasting the onset and course of mental illness with Twitter data

Forecasting the onset and course of mental illness with Twitter data

Normative Emotional Responses to Behavior Analysis Jargon or How Not to Use Words to Win Friends and Influence People

Modeling Spatiotemporal Factors Associated With Sentiment on Twitter: Synthesis and Suggestions for Improving the Identification of Localized Deviations

Contact Info

Product

Resources

About