Automated Dictionary Creation for Analyzing Text: An Illustration from Stereotype Content

Abstract

Recent advances in natural language processing provide new approaches to analyze psychological open-ended data. However, many of these methods require translating to the needs of psychologists working with text. Here, we introduce automated methods to create and validate extensive dictionaries of psychological constructs using Wordnet and word embeddings. Specifically, we first expand an initial list of seed words by using Wordnet to obtain synonyms, antonyms, and other semantically related terms. Next, we evaluate dictionary reliability by using word embeddings trained on independent sources. Finally, we evaluate the dictionaries’ convergent validity against traditional scale ratings and human judgments. We illustrate these innovations by creating stereotype content dictionaries, a construct in social psychology that lacks specialized and validated dictionaries for coding open-ended data. These dictionaries achieved over 80% coverage of new responses, compared to 20% coverage by the seed-word-only approach. Cosine similarity with word embeddings confirmed that the dictionaries are more similar within the same concept than across concepts. Moreover, open-ended responses predicted both traditional scale ratings and human judgments about the dictionary topic. The R package Automated Dictionary Creation for Analyzing Text (ADCAT; https://github.com/gandalfnicolas/ADCAT) allows anyone to create novel dictionaries for constructs of interest and to access the stereotype content dictionaries.

Navigating the social world: Toward an integrated framework for evaluating self, individuals, and groups.

“…Groups array in Warmth-by-Competence space, as participants report societal beliefs (see Table 1). Findings generalize to about 50 countries (Bai, Ramos, & Fiske, 2020; Cuddy et al, 2009; Durante et al, 2017, 2013), retroactively to the 1930s (Bergsieker, Leslie, Constantine, & Fiske, 2012; Durante, Volpato, et al, 2010), for both groups and subgroups (see Fiske, 2018), for nonhuman intent-having entities (animals, Sevillano & Fiske, 2016; corporations, Kervyn, Fiske, et al, 2012), and in spontaneous, open-ended, as well as structured, scaled reports (Nicolas, Bai, & Fiske, 2020a, 2020b). The model’s key predictions follow.…”

“…The first SCM studies of open-ended description, using both the speed and order of response (Nicolas, Bai, & Fiske, 2020a, 2020b), fit this idea: One Vertical (Competence) facet, Ability, tends to be more immediately mentioned but recedes over time. Given the target’s importance, the Horizontal (Warmth) facet Sociability (although less immediately mentioned) prevails eventually, which fits the subjective weight criterion described earlier.…”

“…First, regarding methods, a move from researchers presenting participants with dimensions to less constrained research that lets participants themselves select dimensions might increase the number of dimensions required to fully capture social evaluation. Reverse engineering similarity judgments (Imhoff et al, 2018; Koch et al, 2016) or analyses of natural language (Nicolas, Bai, & Fiske, 2020a, 2020b; see also Abele & Bruckmüller, 2011, 2013) provide paradigms for such research on additional dimensions.…”

Stereotypes as Historical Accidents: Images of Social Class in Postcommunist Versus Capitalist Societies

“…Competing models agree on two dimensions: one vertical (status/competence/agency/dominance) and one horizontal, variously interpreted as warmth, cooperation-competition, communion, trustworthiness, and me–them differences (see adversarial synthesis by Abele, Ellemers, Fiske, Koch, & Yzerbyt, under review). Spontaneous descriptions of 87 societal groups prioritize terms machine-coded as warmth (morality, sociability) and competence (capability, assertiveness; Nicolas, Bai, & Fiske, under review), so these continue to be viable dimensions. Evidence from both correlational (Cuddy et al, 2009; Eckes, 2002) and experimental (Caprariello, Cuddy, & Fiske, 2009; Durante, Tablante, & Fiske, 2017) studies supports the SCM’s assumptions: groups’ status translates into perceptions of competence (capability, effectiveness), and groups’ cooperativeness translates into perceptions of warmth (trustworthiness, sociability).…”

Scared into Action: How Partisanship and Fear are Associated with Reactions to Public Health Directives

“…We initially compared the word vectors of 35 synonyms of fear to the word vector for fear (Nicolas et al, 2019), calculated the cosine similarity, created two distinct clusters of words, and created a dictionary from the tightest cluster to use in the previously described distributed dictionary representation (DDR) method (Garten et al, 2018). The final validated dictionary consists of the 26 words: abhor, alarm, anxiety, apprehension, aversion, concern, consternation, cowardice, doubt, dread, fear, fearful, foreboding, fright, horror, nerves, nervous, nightmare, panic, perturb, scare, terror, trepidation, unease, unrest, worry.…”