Topic Modeling

Latent topics (LDA / NMF / LSA): top terms, coherence, document-topic weights.

Topic Modeling

The Topic Modeling widget.

Overview

Discovers latent topics in a text field (Abstract, Title, or their combination) using LDA, NMF or LSA. Reports the top terms per topic with a per-topic coherence score, and the per-document topic-weight matrix. A static, whole-corpus complement to Dynamic Topic Models (over time) and to Document Clustering (hard clusters).

Inputs

  • Data (Table) — data with a text field.

Outputs

  • Topic Terms (Table) — Topic / Term / Weight.

  • Document Topics (Table) — per-document topic weights.

  • Topic Summary (Table) — one row per topic (top terms + coherence).

Controls

  • Text — the text source (Abstract / Title / Title+Abstract / Author Keywords).

  • Model — LDA, NMF or LSA.

  • Topics (0 = auto) — fix the number of topics, or 0 to choose automatically up to the cap.

  • Max topics (auto) — the upper bound when auto-selecting.

  • Top terms shown — number of terms listed/plotted per topic.

  • Topic (View) — which topic’s top-term bar chart to display.

Actions: Run.