Skip to contents

Contrastive Learning Topic Modeling

Source: R/topic_modeling.R
run_contrastive_topics_internal.Rd

Implements contrastive learning approaches for topic modeling to improve topic separation and discriminability.

Usage

run_contrastive_topics_internal(
  texts,
  n_topics = 10,
  temperature = 0.1,
  negative_sampling_rate = 5,
  embedding_model = "all-MiniLM-L6-v2",
  seed = 123
)

Arguments

texts

Character vector of documents

n_topics

Number of topics to discover

temperature

Temperature parameter for contrastive learning

negative_sampling_rate

Rate of negative sampling

embedding_model

Transformer model for embeddings

seed

Random seed for reproducibility

Value

List containing contrastive topic model and metrics

See also

Other topic-modeling: analyze_semantic_evolution(), assess_hybrid_stability(), calculate_assignment_consistency(), calculate_eval_metrics_internal(), calculate_keyword_stability(), calculate_semantic_drift(), calculate_topic_probability(), calculate_topic_stability(), find_optimal_k(), find_topic_matches(), fit_embedding_topics(), fit_hybrid_model(), fit_temporal_model(), generate_topic_labels(), get_topic_prevalence(), get_topic_terms(), get_topic_texts(), identify_topic_trends(), plot_model_comparison(), plot_quality_metrics(), run_neural_topics_internal(), run_temporal_topics_internal(), validate_semantic_coherence()