Skip to contents

Match Text Against SADCAT and/or SOCATS Dictionaries

match_dictionaries.Rd

Match text responses against SADCAT stereotype dictionaries and/or SOCATS social category dictionaries. Tokenizes response text, matches against dictionaries, and computes binary indicators, percentages, direction scores, and per-dimension valence/direction columns.

Usage

match_dictionaries(
  data,
  text_col = "tv3",
  response_col = NULL,
  valence_col = "ValenceYesNA",
  valence_nona_col = "ValenceNoNA",
  sadcat_dict = NULL,
  socats_dict = NULL,
  socats = FALSE,
  keep_intermediate = FALSE
)

Arguments

data

A data.frame with preprocessed text and valence scores

text_col

Column with singularized text to match (default "tv3")

response_col

Column used for NA-gating (default: same as text_col). Only needed if your NA-indicator column differs from text_col.

valence_col

Name of NA-gated combined valence column (default "ValenceYesNA")

valence_nona_col

Name of zero-imputed combined valence column (default "ValenceNoNA")

sadcat_dict

Pre-computed SADCAT quanteda dictionary. If NULL and sadcat = TRUE (default), calls prepare_sadcat_dictionaries(). Set to FALSE to skip SADCAT matching entirely.

socats_dict

Pre-computed SOCATS quanteda dictionary. If NULL and socats = TRUE, calls prepare_socats_dictionaries().

socats

Logical. Match against SOCATS dictionaries? (default FALSE)

keep_intermediate

Logical. If FALSE (default), return compact SADCAT outputs with informative names: {Dim}_prevalence, {Dim}_valence, {Dim}_direction, {Dim}_valenceNoNA, {Dim}_directionNoNA, and NoMatch. If TRUE, keep legacy SADCAT intermediate columns as well.

Value

The input data with dictionary-derived columns appended. By default, SADCAT outputs are compact. For each SADCAT dimension {Dim}, three valence columns are produced:

  • {Dim}_Valence (default for downstream means): NA when the dimension is not tagged in the response; otherwise the global ValenceNoNA (0 if no sentiment words matched, else the signed negation-aware mean). Use mean(., na.rm = TRUE) to get the average valence among tagged responses, where sentiment-less tagged responses contribute 0.

  • {Dim}_valenceStrictNA: NA whenever either the dimension is not tagged OR the global ValenceYesNA is itself NA. Strictly NA-gated on both axes.

  • {Dim}_valenceNoNA: 0 whenever either the dimension is not tagged OR the global sentiment is NA. Strictly zero-imputed on both axes; interpretable as a prevalence-weighted valence over the full response set.