Semantic Severity Method

Source code to evaluate the semantic severity (vertical expansion) of concepts.

Note: Collocate files use three centre terms ($word = "trauma", "anxiety", "depression") but the list can be expanded to include more terms.

Data

Input folder

[1] "warriner_rat.csv": csv file containing valence and arousal ratings to be matched with lemmas (collocates for the target/centre term). The file contains normed ratings of 13,915 words (lemmas specifically) on a 9-point likert scale: "mean", "sd", "gender" responses for US residents grouped by age. Only these columns are relevant to the method:

• word = English lemmas
• V.Mean.Sum = Summed average ratings of lemmas on Valence (i.e., the pleasantness of the emotions invoked by a word) from happy (9) to unhappy (1)
• A.Mean.Sum = Summed average ratings of lemmas on Arousal (i.e., the intensity of emotion provoked by a stimulus) from aroused (9) to calm/unaroused (1)

[2i--n] "$word_year_coll_repet.csv": csv file containing corpus statistics for collocate's occurence near the target/centre term per year.

• year = the year in which the lemma appeared in the corpus (1970-2017)
• coll = collocate appearing within the target term's context window (+/- 5 words); also a lemma (the root form of the English word)
• repet = the number of times that the collocates occur near (+/- 5-word context-window) the target/centre term in corpus sentences (abstracts) per year

Output folder

[3] "$word_$corpus_severity_indices.csv: csv file containing time series of severity indices

• year = the year in which the lemma appeared in the corpus (1970-2017)
• sev_index = index for the severity (repetition-weighted valence+arousal ratings) of $word-related lemmas
• aro_index = index for the (repetition-weighted) arousal ratings of $word-related lemmas
• val_index = index for the (repetition-weighted) valence ratings of $word-related lemmas

Method

Corpus Preprocessing

See "preprocess_data_spacy.py" script (Python). Dependencies: ``spacy'' library and a pre-trained model (e.g., "en_core_web_lg").

• Steps: (i) Remove punctuation, (ii) Lemmatize words, (iii) De-capitalize words, (iv) Remove stop-words

Collocations

See "retrieve_coll_year_repet_corpora.sh" script (bash) for how to extract collocations (word) and their repetitions (repet) for 3 terms in 2 corpora.

• Steps: (i) Restrict corpus to only lines (abstracts/articles) containing the target/centre term, (ii) identify collocates (within 5-word context window), (iii) compute statistics for the number of times collocate occurs near the target/centre term and in what corpus year.

Severity Index

See "warrmatched_coll_year_stats_indices.R" script (R Programming). "severity_indices.R" is an example of how to run the method on one term.

• Steps: (i) Link the Warriner ratings with the collocate data and reverse score V.Mean.Sum so scores range from happy (1) to unhappy (9), (ii) compute repetition-weighted average semantic severity index by (a) summing V.Mean.Sum and A.Mean.Sum, (b) weighting it by the repetition of each colocate by year and (c) dividing this numerator by the sum of collocate repetitions grouped by year (denominator).

References

[1] Warriner, A. B., Kuperman, V., & Brysbaert, M. (2013). Norms of valence, arousal, and dominance for 13,915 English lemmas. Behavior research methods, 45(4), 1191-1207. https://doi.org/10.3758/s13428-012-0314-x

This method has been used in the following manuscripts:

[2] Baes, N., Vylomova, E., Zyphur, M., & Haslam, N. (2023). The semantic inflation of “trauma” in psychology. Psychology of Language and Communication, 27(1), 23-45. DOI: https://doi.org/10.58734/plc-2023-0002

[3] XX (2023). Have the concepts of ‘anxiety’ and ‘depression’ been normalized or pathologized? A corpus study of historical semantic change