Culture and Moral Expressions in Language Corpus

Extracting psychological constructs from language data is key to modeling human behavior at scale, and doing so requires building theory-driven annotated datasets. Here, we introduce the Culture and Moral Expressions in Language (CAMEL) corpus, a collection of over 56,000 texts that has been curated from different sources, labeled for 25 cultural and moral-psychological constructs. The annotators (N = 30) underwent an intensive training program involving reading the relevant literature, an in-depth review of our in-house annotation manual, and recurrent group discussions to resolve ambiguities. The overarching goal of developing the CAMEL corpus is to build a methodologically rigorous, theory-driven, and topically diverse dataset, laying the groundwork for scalable psychological research in natural language processing. We also develop preliminary classifiers using both word-counting and language-embedding approaches. The CAMEL corpus offers a resource for researchers seeking to extract cultural- and moral-psychological content from text without requiring substantial investment of time, money, and human labor.