Public culture is a powerful source of cognitive socialization; for example, media language full meanings about body weight. Yet it remains unclear how individuals process in public culture. We suggest that schema learning core mechanism by which becomes personal propose burgeoning approach computational text analysis - neural word embeddings can be interpreted as formal model cultural learning. Embeddings allow us to empirically and activation from natural data. illustrate our extracting...

Abstract In this article, we apply computational word embeddings to a 200-million-word corpus of American print media (1930–2009) examine how education-relevant gender stereotypes changed as women’s educational attainment caught up with and eventually surpassed men’s. This case presents rare opportunity observe cultural components the system transform alongside reversal an important pattern stratification. We track six that prior work linked academic outcomes. Our results suggest most...

Why are some diseases more stigmatized than others? And, has disease stigma declined over time? Answers to these questions have been hampered by a lack of comparable, longitudinal data. Using word embedding methods, we analyze 4.7 million news articles create new measures for 106 health conditions from 1980 2018. mixed-effects regressions, find that behavioral and preventable attract the strongest connotations immorality negative personality traits, infectious most marked disgust. These...

There is an escalating need for methods to identify latent patterns in text data from many domains. We introduce a new method topics corpus and represent documents as topic sequences. Discourse Atom Topic Modeling draws on advances theoretical machine learning integrate modeling word embedding, capitalizing the distinct capabilities of each. first set vectors ("discourse atoms") that provide sparse representation embedding space. can be interpreted topics: Through generative model, atoms map...

Despite extensive efforts to standardize definitions of obesity, clinical practices diagnosing obesity vary widely. This study examined (1) discrepancies between biometric body mass index (BMI) measures and documented diagnoses in patient electronic health records (EHRs) (2) how these by gender race ethnicity from an intersectional lens. Observational 383,380 participants the National Institutes Health All Us Research Program dataset. Over half (60 %) with a BMI indicating had no diagnosis...

Word embeddings are language models that represent words as positions in an abstract many-dimensional meaning space. Despite a growing range of applications demonstrating their utility for sociology, there is little conceptual clarity regarding what exactly measure and whether this matches we need them to measure. Here, fill theoretical gap by clarifying how cultural can be understood spatial terms. We argue operationalize context spaces, where words’ reflect any regularity usage. then...

Measuring meaning is a central problem in cultural sociology and word embeddings may offer powerful new tools to do so. But like any tool, they build on exert theoretical assumptions. In this paper, I theorize the ways which model three core premises of structural linguistic theory meaning: that coherent, relational, be analyzed as static system. certain ways, are vulnerable enduring critiques these premises. other novel solutions critiques. More broadly, formalizing study with offers...

We develop a novel application of machine learning and apply it to the interview transcripts from American Voices Project (N = 1,396), using discourse atom topic modeling explore social class variation in centrality family adults' lives. take two-phase approach, first analyzing at person level then line level. Our findings suggest that family, as represented by talk, is more central lives those without college degree than among educated. However, institutional overlap between other key...

An abundance of methodological work aims to detect hateful and racist language in text. However, these tools are hampered by problems like low annotator agreement remain largely disconnected from theoretical on race racism the social sciences. Using annotations 5188 tweets 291 annotators, we investigate how perceptions vary racial identity two text features tweets: relevant keywords latent topics identified through structural topic modeling. We provide a descriptive summary our data estimate...

This preprint is an earlier draft the published article: Arseniev-Koehler, Alina, and Jacob G. Foster. "Machine learning as a model for cultural learning: Teaching algorithm what it means to be fat." Sociological Methods & Research 51.4 (2022): 1484-1539.

Computational models to detect mental illnesses from text and speech could enhance our understanding of health while offering opportunities for early detection intervention. However, these are often disconnected the lived experience depression larger diagnostic debates in health. This article investigates disconnects, primarily focusing on labels used diagnose depression, how computationally represented, performance metrics evaluate computational models. We also consider medical instruments...

Objectives. To investigate racial/ethnic differences in legal intervention‒related deaths using state-of-the-art topic modeling of law enforcement and coroner text summaries drawn from the 2003–2017 US National Violent Death Reporting System (NVDRS). Methods. Employing advanced modeling, we identified 8 topics consistent with dangerousness death incidents NVDRS narratives written by public health workers (PHWs). Using logistic regression, then evaluated PHW-coded variables narrative among...

This is a preprint for the published article: Arseniev-Koehler, A., Cochran, S. D., Mays, V. M., Chang, K. W., & Foster, J. G. (2022). Integrating topic modeling and word embedding to characterize violent deaths. Proceedings of National Academy Sciences (PNAS), 119(10), e2108801119.

In the process of retelling information, individuals often inadvertently transform it to be more consistent with their cultural schemas. We explore long‐term change inherent in this process, focusing on utterances about tastes as our case study (e.g., music, food, and outdoor hobbies). use a word embedding model simulate “telephone game” where each actor partially hears an utterance, uses schemas guess missing word, tells result next actor. While laboratory games” short transmission chains...

Gender stereotypes have important consequences for boys’ and girls’ academic outcomes. In this article, we apply computational word embeddings to a 200-million-word corpus of American print media (1930-2009) examine how these changed as women’s educational attainment caught up with eventually surpassed men’s. This transformation presents rare opportunity observe change alongside the reversal an pattern stratification. We track six that prior work has linked Our results suggest...

Has disease stigma declined? Our ability to answer this question has been hampered by the lack of comparable data across diseases and over time. Using word embeddings, we analyze 4.5 million news articles create new measures for 107 health conditions. We find that in 1980s, most were marked strong connotations disgust, danger, impurity, negative personality traits. Since then, declined dramatically physical illnesses; cancers, neurological conditions, genetic diseases, many other conditions...

This preprint is an earlier version of the published article: Arseniev-Koehler, Alina. "Theoretical foundations and limits word embeddings: what types meaning can they capture?." Sociological Methods & Research (2021): 00491241221140142.

This preprint is an earlier version of the published article: Larimore, Savannah, Ian Kennedy, Breon Haskett, and Alina Arseniev-Koehler. "Reconsidering annotator disagreement about racist language: Noise or signal?." In Proceedings Ninth International Workshop on Natural Language Processing for Social Media, pp. 81-90. 2021.

This is a preprint for book chapter published in Dehghani, Morteza, and Ryan L. Boyd, eds. Handbook of Language Analysis Psychology. Guilford Publications, 2022. In this chapter, we describe the burgeoning empirical evidence showing how meanings captured word embeddings correspond to human meanings. We then review theoretical illustrating — diverge from cognitive strategies represent process meaning. turn, these divergences illustrate challenges future directions research with embeddings.

Ankith Uppunda, Susan Cochran, Jacob Foster, Alina Arseniev-Koehler, Vickie Mays, Kai-Wei Chang. Proceedings of the 2021 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2021.