"What's ur type?" Contextualized Classification of User Types in Marijuana-related Communications using Compositional Multiview Embedding

Kursuncu, Ugur; Lokala, Usha; Illendula, Anurag; Thirunarayan, Krishnaprasad; Daniulaityte, Raminta; Sheth, Amit P.; Arpinar, I. Budak

doi:10.48550/arxiv.1806.06813

Cited by 1 publication

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There is also a line of work that integrates information from multiple user views to learn more robust user embeddings (Li et al, 2017;Tao and Yang, 2017;Kursuncu et al, 2018;Hazarika et al, 2018). The models we present are all trying to extract value out of features that are only distantly related to a task of interest.…”

Section: Expanding What Constitutes As Model Supervisionmentioning

confidence: 99%

“…Li et al (2017) takes a multitask approach to learning user embeddings and evaluates the embeddings according to how well they predict which text and other users they are likely to agree with. Although considering a supervised objective to learn user embeddings, Kursuncu et al (2018) also collapses features from each view into a joint user embedding. Multiview user embeddings have even been shown to be predictive of sarcasm in author tweets (Hazarika et al, 2018).…”

Section: Learning Social Media User Embeddingsmentioning

confidence: 99%

See 1 more Smart Citation

Learning Representations of Social Media Users

Benton¹

2018

Preprint

View full text Add to dashboard Cite

Social media users routinely interact by posting text updates, sharing images and videos, and establishing connections with other users through friending. User representations are routinely used in recommendation systems by platform developers, targeted advertisements by marketers, and by public policy researchers to gauge public opinion across demographic groups. Computer scientists consider the problem of 5.5 Precision-recall curves for predicting each mental health condition. xxii 6.1 Boxplots of mean cross-fold F1-score as a function of different hyperparameters: RNN bi-directionality (upper left), number of layers (upper right), hidden layer width (lower left), embedding width (lower center), and dropout rate (lower right) for an Author Text-pretrained RNN on the "Feminism Movement" stance classification task. . . . 176 xxiii

show abstract

Section: Expanding What Constitutes As Model Supervisionmentioning

confidence: 99%

Section: Learning Social Media User Embeddingsmentioning

confidence: 99%