Computes word correlation networks with community detection and network metrics. Supports multiple feature spaces: unigrams, n-grams, and embeddings. Based on proven implementation for intuitive network visualization.
Usage
semantic_correlation_network(
dfm_object,
doc_var = NULL,
common_term_n = 20,
corr_n = 0.4,
top_node_n = 30,
node_label_size = 14,
pattern = NULL,
showlegend = TRUE,
seed = NULL,
feature_type = "words",
ngram_range = 2,
texts = NULL,
embeddings = NULL
)Arguments
- dfm_object
A quanteda document-feature matrix (dfm).
- doc_var
A document-level metadata variable for categories (default: NULL).
- common_term_n
Minimum term frequency to include (default: 20).
- corr_n
Minimum correlation threshold (default: 0.4).
- top_node_n
Number of top nodes to display (default: 30).
- node_label_size
Font size for node labels (default: 14).
- pattern
Regex pattern to filter specific words (default: NULL).
- showlegend
Whether to show community legend (default: TRUE).
- seed
Random seed for reproducible layout (default: NULL).
- feature_type
Feature space: "words", "ngrams", or "embeddings" (default: "words").
- ngram_range
N-gram size when feature_type = "ngrams" (default: 2).
- texts
Optional character vector of texts for n-gram creation (default: NULL).
- embeddings
Optional embedding matrix for embedding-based networks (default: NULL).