Johns Hopkins or johnny-hopkins: Classifying Individuals versus
            Organizations on Twitter

Wood-Doughty, Zach; Mahajan, Praateek; Dredze, Mark

doi:10.18653/v1/w18-1108

Cited by 18 publications

(20 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We first gathered the unique Twitter user IDs of all the Twitter users in our data set. Following the approach outlined in [ 25 ], we used a naïve Bayes machine learning model to classify the tweeters into individuals versus organizations. We used a published data set that contained 8945 Twitter users and their profile descriptions, which human coders used to annotate users as individual or institutional [ 26 ].…”

Section: Methodsmentioning

confidence: 99%

Topics, Trends, and Sentiments of Tweets About the COVID-19 Pandemic: Temporal Infoveillance Study

Ranganathan¹,

Mehta²,

Valkunde³

et al. 2020

J Med Internet Res

200

187

View full text Add to dashboard Cite

Background With restrictions on movement and stay-at-home orders in place due to the COVID-19 pandemic, social media platforms such as Twitter have become an outlet for users to express their concerns, opinions, and feelings about the pandemic. Individuals, health agencies, and governments are using Twitter to communicate about COVID-19. Objective The aims of this study were to examine key themes and topics of English-language COVID-19–related tweets posted by individuals and to explore the trends and variations in how the COVID-19–related tweets, key topics, and associated sentiments changed over a period of time from before to after the disease was declared a pandemic. Methods Building on the emergent stream of studies examining COVID-19–related tweets in English, we performed a temporal assessment covering the time period from January 1 to May 9, 2020, and examined variations in tweet topics and sentiment scores to uncover key trends. Combining data from two publicly available COVID-19 tweet data sets with those obtained in our own search, we compiled a data set of 13.9 million English-language COVID-19–related tweets posted by individuals. We use guided latent Dirichlet allocation (LDA) to infer themes and topics underlying the tweets, and we used VADER (Valence Aware Dictionary and sEntiment Reasoner) sentiment analysis to compute sentiment scores and examine weekly trends for 17 weeks. Results Topic modeling yielded 26 topics, which were grouped into 10 broader themes underlying the COVID-19–related tweets. Of the 13,937,906 examined tweets, 2,858,316 (20.51%) were about the impact of COVID-19 on the economy and markets, followed by spread and growth in cases (2,154,065, 15.45%), treatment and recovery (1,831,339, 13.14%), impact on the health care sector (1,588,499, 11.40%), and governments response (1,559,591, 11.19%). Average compound sentiment scores were found to be negative throughout the examined time period for the topics of spread and growth of cases, symptoms, racism, source of the outbreak, and political impact of COVID-19. In contrast, we saw a reversal of sentiments from negative to positive for prevention, impact on the economy and markets, government response, impact on the health care industry, and treatment and recovery. Conclusions Identification of dominant themes, topics, sentiments, and changing trends in tweets about the COVID-19 pandemic can help governments, health care agencies, and policy makers frame appropriate responses to prevent and control the spread of the pandemic.

show abstract

Section: Methodsmentioning

confidence: 99%

Topics, Trends, and Sentiments of Tweets About the COVID-19 Pandemic: Temporal Infoveillance Study

Ranganathan¹,

Mehta²,

Valkunde³

et al. 2020

J Med Internet Res

200

187

View full text Add to dashboard Cite

show abstract

“…No multilingual organization recognition system exists; so, we limit our evaluation to the only publicly-available dataset of personvs-organization which was scored by the Humanizr [54] and Demographer [79] systems. This data consists of a uniform sample across Twitter accounts, which were approximately 10% organizations.…”

Section: Comparison Systems Andmentioning

confidence: 99%

Demographic Inference and Representative Population Estimates from Multilingual Social Media Data

Wang

Hale

Adelani³

et al. 2019

The World Wide Web Conference

175

180

View full text Add to dashboard Cite

Social media provide access to behavioural data at an unprecedented scale and granularity. However, using these data to understand phenomena in a broader population is difficult due to their nonrepresentativeness and the bias of statistical inference tools towards dominant languages and groups. While demographic attribute inference could be used to mitigate such bias, current techniques are almost entirely monolingual and fail to work in a global environment. We address these challenges by combining multilingual demographic inference with post-stratification to create a more representative population sample. To learn demographic attributes, we create a new multimodal deep neural architecture for joint classification of age, gender, and organization-status of social media users that operates in 32 languages. This method substantially outperforms current state of the art while also reducing algorithmic bias. To correct for sampling biases, we propose fully interpretable multilevel regression methods that estimate inclusion probabilities from inferred joint population counts and ground-truth population counts. In a large experiment over multilingual heterogeneous European regions, we show that our demographic inference and bias correction together allow for more accurate estimates of populations and make a significant step towards representative social sensing in downstream applications with multilingual social media.

show abstract

“…Further, we observe in other data sets that usernames are some of the most frequent tokens classified as entities (Ritter et al, 2011;Derczynski et al, 2016). For our experiments, we consider all usernames as non-entities, as otherwise, identifying these using character features would be trivial, and typing entities would be similar to the task of Twitter handle classification (McCorriston et al, 2015;Wood-Doughty et al, 2018), which is outside the scope of the current paper.…”

Section: Annotationmentioning

confidence: 99%

Temporally-Informed Analysis of Named Entity Recognition

Rijhwani¹,

Preoţiuc-Pietro²

2020

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

Natural language processing models often have to make predictions on text data that evolves over time as a result of changes in language use or the information described in the text. However, evaluation results on existing data sets are seldom reported by taking the timestamp of the document into account. We analyze and propose methods that make better use of temporally-diverse training data, with a focus on the task of named entity recognition. To support these experiments, we introduce a novel data set of English tweets annotated with named entities. 1 We empirically demonstrate the effect of temporal drift on performance, and how the temporal information of documents can be used to obtain better models compared to those that disregard temporal information. Our analysis gives insights into why this information is useful, in the hope of informing potential avenues of improvement for named entity recognition as well as other NLP tasks under similar experimental setups.

show abstract

Johns Hopkins or johnny-hopkins: Classifying Individuals versus Organizations on Twitter

Cited by 18 publications

References 22 publications

Topics, Trends, and Sentiments of Tweets About the COVID-19 Pandemic: Temporal Infoveillance Study

Topics, Trends, and Sentiments of Tweets About the COVID-19 Pandemic: Temporal Infoveillance Study

Demographic Inference and Representative Population Estimates from Multilingual Social Media Data

Temporally-Informed Analysis of Named Entity Recognition

Contact Info

Product

Resources

About