Modeling Latent Biographic Attributes in Conversational Genres

By Nikesh Garera, David Yarowsky. 2009

98 citations recorded by the source corpus.

This paper presents and evaluates several original techniques for the latent classification of biographic attributes such as gender, age and native language, in diverse genres (conversation transcripts, email) and languages (Arabic, English). First, we present a novel partner-sensitive model for extracting biographic attributes in conversations, given the differences in lexical usage and discourse style such as observed between same-gender and mixed-gender conversations. Then, we explore a rich variety of novel sociolinguistic and discourse-based features, including mean utterance length, passive/active usage, percentage domination of the conversation, speaking rate and filler word usage. Cumulatively up to 20% error reduction is achieved relative to the standard Boulis and Ostendorf (2005) algorithm for classifying individual conversations on Switchboard, and accuracy for gender detection on the Switchboard corpus (aggregate) and Gulf Arabic corpus exceeds 95%.

Subfields: Artificial Intelligence, Human-Computer Interaction

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 Human-Computer Interaction

Citations in the corpus

"How Old Do You Think I Am?" A Study of Language and Age in Twitter (2013, background): “…automatic age prediction has mostly been approached as a two-class or three-class classiﬁcation problem based on age spans with for example boundaries at 30 or 40 years (e.g. (Rao et al. 2010; Garera and Yarowsky 2009; Goswami, Sarkar, and Rustagi 2009)), thus corresponding to an etic approach.”
Exploring Demographic Language Variations to Improve Multilingual Sentiment Analysis in Social Media (2013, background): “…found that the majority of emotion-bearing features, e.g., emoticons, repeated letters, exasperation, are used more by female than male users, which is consistent with results reported in other recent work (Garera and Yarowsky, 2009; Burger et al., 2011; Goswami et al., 2009; Argamon et al., 2007).”
Computational Sociolinguistics: A Survey (2016, background): “Early computational studies on social variables and language use were based on formal texts, such as the British National Corpus [6, 145], or datasets collected from controlled settings, such as recorded conversations [236] and telephone conversations [27, 81, 258] where protocols were used to coord”
Classifying latent user attributes in twitter (2010, background): “Previous work by Garera and Yarowsky [7] showed the importance of sociolinguistic models in classifying latent speaker attributes in conversational speech transcripts.”
Author Age Prediction from Text using Linear Regression (2011, background): “, 2009), telephone conversations (Garera and Yarowsky, 2009), and recently Twitter (Rao et al.”
User Demographics and Language in an Implicit Social Network (2012, background): “An exception to this is Garera & Yarowsky (2009) who showed that, for gender prediction in a dialogue, it helps to know the interlocutor’s gender.”
A Machine Learning Approach to Twitter User Classification (2011, background): “Researchers investigated the detection of gender from well-written, traditional text (Herring and Paolillo 2010; Singh 2001), blogs (Burger and Henderson 2010) reviews (Otterbacher 2010), e-mail (Garera and Yarovsky 2007), user search queries (Jones et al.”
Democrats, republicans and starbucks afficionados: user classification in twitter (2011, background): “For example, researchers investigated the detection of gender from wellwritten, traditional text [10], blogs [4] reviews [14], e-mail [9], user search queries [12, 24] and, for the first time, Twitter [19].”
Confounds and Consequences in Geotagged Twitter Data (2015, background): “Binning is often employed in work on text-based age prediction (Garera and Yarowsky, 2009; Rao et al., 2010; Rosenthal and McKeown, 2011); it enables word and name counts to be shared over multiple ages, and avoids the complexity inherent in regressing a high-dimensional textual predictors against a”
Arabic Dialect Identification (2014, background): “In the same vein, being able to recognize dialectal content in user-generated text could aid in characterizing communicants and their biographic attributes (Garera and Yarowsky 2009).”
Why Gender and Age Prediction from Tweets is Hard: Lessons from a Crowdsourcing Experiment (2014, background): “, 2005), telephone conversations (Garera and Yarowsky, 2009), YouTube (Filippova, 2012) and chats in social networks (Peersman et al.”
(Male, Bachelor) and (Female, Ph.D) have different connotations: Parallelly Annotated Stylistic Language Dataset with Multiple Personas (2019, background): “…applications us-ing stylistic features have been studied such as poetry generation (Ghazvininejad et al., 2017), stylometry with demographic information (Ver-hoeven and Daelemans, 2014), modeling style bias (Vogel and Jurafsky, 2012) and modeling bio-graphic attributes (Garera and Yarowsky, 2009).”
Gender-Based Vocation Identification in Swedish 19th Century Prose Fiction using Linguistic Patterns, NER and CRF Learning (2015, background): “The research questions themselves are not new, but in fact central to the field of gender studies and to a certain extent, economic history.”