A Machine Learning Approach to Twitter User Classification

By Marco Pennacchiotti, Ana-Maria Popescu. 2011

640 citations recorded by the source corpus.

This paper addresses the task of user classification in social media, with an application to Twitter. We automatically infer the values of user attributes such as political orientation or ethnicity by leveraging observable information such as the user behavior, network structure and the linguistic content of the user’s Twitter feed. We employ a machine learning approach which relies on a comprehensive set of features derived from such user information. We report encouraging experimental results on 3 tasks with different characteristics: political affiliation detection, ethnicity identification and detecting affinity for a particular business. Finally, our analysis shows that rich linguistic features prove consistently valuable across the 3 tasks and show great promise for additional user classification needs.

Subfields: Artificial Intelligence, Information Systems

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

Citations in the corpus

Developing Age and Gender Predictive Lexica over Social Media (2014, background): “Use of social media has enabled the study of psychological and social questions at an unprecedented scale (Lazer et al., 2009).”
Beyond Binary Labels: Political Ideology Prediction of Twitter Users (2017, background): “These include users who explicitly state their political orientation on user lists of party supporters (Zamal et al., 2012; Pennacchiotti and Popescu, 2011), supporting par-”
Quantifying Political Leaning from Tweets and Retweets (2013, background): “Machine learning techniques have been proposed to classify Twitter users using e.g., linguistic content, mention/retweet behavior and social network structure (Boutet, Kim, and Yoneki 2012; Al Zamal, Liu, and Ruths 2012; Pennacchiotti and Popescu 2011).”
TrGraph: Cross-Network Transfer Learning via Common Signature Subgraphs (2015, methodology): “By combining the original node content features and the new structure features, we develop an iterative classification algorithm, TrGraph, that utilizes label dependency to jointly classify nodes in the target network.”
Gender Inference of Twitter Users in Non-English Contexts (2013, methodology): “In prior work, e.g. (Rao et al., 2010; Pennacchiotti and Popescu, 2011; Za-mal et al., 2012), the dominant way of obtaining datasets consisting of Twitter users with high-conﬁdence gender-labels is to use gender-name associations.”
User-Level Race and Ethnicity Predictors from Twitter Text (2018, positive): “We presented a detailed study of user-level race/ethnicity prediction along the lines of previous work on predicting user traits from text (Burger et al., 2011; Rao et al., 2010; Pennacchiotti and Popescu, 2011; Schwartz et al., 2013; Sap et al., 2014; Volkova et al., 2014; Preoţiuc-Pietro et al., 2”
Cybercrime detection in online communications: The experimental case of cyberbullying detection in the Twitter network (2016, background): “Activity features measure the online communication activity of a user (Pennacchiotti & Popescu, 2011).”
"On the internet, nobody knows you're a dog": a twitter case study of anonymity in social networks (2014, background): “However, it has been shown that Twitter profile pictures are often misleading, making it hard to even deduce ethnicity or gender, and are often virtual characters (such as cartoons) or belong to celebrities [37].”
Who Does What on the Web: A Large-Scale Study of Browsing Behavior (2012, background): “Recent work (Pennacchiotti and Popescu, 2011) inferred user demographics on Twitter, leveraging profile, content, and network information to predict political affiliation and race with reasonable accuracy.”
Homophily and Latent Attribute Inference: Inferring Latent Attributes of Twitter Users from Neighbors (2012, methodology): “In (Pennacchiotti and Popescu 2011), commonly friended and mentioned accounts were identified among users with a given attribute (e.”
Online and Social Media Data As an Imperfect Continuous Panel Survey (2016, background): “(2011) document the demographic skew on Twitter along with other studies that focus on inferring unknown demographics and attributes of social media users (Chang et al., 2010; Pennacchiotti and Popescu, 2011; Zamal et al., 2012; Kosinski et al., 2013), but do little in joining this information with ”
Computational Sociolinguistics: A Survey (2016, background): “These three variables have received the most attention in computational linguistics, compared to other variables such as ethnicity [5, 201, 221] and social class.”
Classifying Political Orientation on Twitter: It's Not Easy! (2013, methodology): “Here we briefly survey how support vector machines (SVMs), boosted decision trees (BDTs), and latent dirichlet allocation-based (LDAs) methods and strategies that have been used in the past.”
Semantic Web in data mining and knowledge discovery: A comprehensive survey (2016, background): “(iii) How can we extract reliable information from these noisy, dynamic content streams? (iv) How can we model users’ digital identity and social media activities? For example, discovering user demographics [68], deriving user interests [69] and capturing user behavior [70].”
Obfuscating Gender in Social Media Writing (2016, background): “This results in 432 , 983 users in the Yelp corpus and 945 , 951 users in the Twitter data.”
Write Like I Write: Herding in the Language of Online Reviews (2014, background): “The trend is clear even for the remaining features where this complete compliance is not observed, and we expect that given more data and the ability to examine sufficiently larger contexts (i.e., values of L) we would still get a complete compliance.”
A Meta-Analysis of State-of-the-Art Electoral Prediction From Twitter Data (2012, background): “Needless to say, geographical and demographical information about the users should be inferred if not available (e.g., Mahmud, Nichols, & Drews, 2012; Pennacchiotti & Popescu, 2011).”