Search results

Add top-down (divisive) strategy for hierarchical_topics()

Status: Open.

#2508 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

Support pre-tokenized documents to avoid redundant tokenization across pipeline stages

Status: Open.

#2507 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

Add contrastive and embedding-aware feature importance as representation aspects

Status: Open.

#2506 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

Lightweight soft topic assignment via temperature-scaled embedding distances

Status: Open.

#2505 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

Support reusing pre-computed UMAP embeddings across runs (grid search, caching)

Status: Open.

#2504 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

hierarchical_topics() parent nodes show raw c-TF-IDF keywords instead of representation model labels

Status: Open.

#2503 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

approximate_distribution: add target-based min_similarity ("at most N unassigned documents")

Status: Open.

#2502 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

reduce_outliers: add target-based outlier reduction ("reduce to X% outliers")

Status: Open.

#2501 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

Allow recalculating representative docs with a configurable count (currently hardcoded to 3)

Status: Open.

#2500 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

_extract_representative_docs recomputes from scratch on every call — no caching

Status: Open.

#2499 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

MaximalMarginalRelevance makes 2N embedding calls instead of 1

Status: Open.

#2498 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026

Extract duplicated document-aggregation and topic-naming patterns into shared helpers

Status: Open.

#2497 In MaartenGr/BERTopic;

· pidefrem opened on Jul 8, 2026