Figure 2.

(A) Average paragraph-level valence for paragraphs containing gay, homosexual, any LGBTQ term, and American, grouped into 5-years intervals. Paragraph-level scores are calculated as the average valence over all words that appear in the NRC VAD Valence Lexicon, which range from 0 (most negative) to 1 (most positive) (Mohammad, 2018). Paragraphs containing LGBTQ labels become more positive over time. Paragraphs containing homosexual are significantly more negative than those containing other LGBTQ labels. (B) Average connotation frame perspective scores over 5-years intervals. Scores are calculated for each subject-verb-object triple containing these group labels as the writer's perspective based on the head verb's entry in the Connotation Frames lexicon (Rashkin et al., 2016). (C) Average valence of 500 nearest words to vector representations of gay, homosexual, all LGBTQ terms, and American, averaged over 10 word2vec models trained on New York Times data from each year. The solid lines are Lowess curves for visualization purposes. Words' valence scores are from the NRC VAD Valence Lexicon. For all plots, the shaded bands represent 95% confidence intervals.