What you can do

topica is a general topic-modeling toolkit. This section tours what it does. If your goal is a publishable analysis, pair it with Publishing in a journal.

Choose an approach

Find your goal in the left column. Every model is validated before it ships, against a maintained reference implementation where one exists and otherwise by planted recovery on a synthetic corpus with a known answer (see validation), so the choice is about fit to your research design, not quality. A specialized model is often the right first choice when your data calls for it. The full roster lists every model for each goal.

Common openings

If your goal is…	Start with	Also consider	First calls	Note
Explore themes with no prior structure	LDA	NMF	search_k(), topic_table()	The default first pass. NMF is a fast, deterministic alternative.
Relate topics to metadata (author, date, party)	STM	DMR	estimate_effect(), one_hot(), spline()	STM gives covariate effects with uncertainty; DMR is a lighter Gibbs prior.
Measure concepts you can name in advance	KeyATM	SeededLDA	KeyATM(keywords=…), .keyword_rate	Anchor named topics with a few seed words each.
Very short documents: tweets, headlines, survey answers	GSDMM	PT	fit()	One topic per document; standard LDA over-fragments short text.
Cluster by meaning using embeddings	BERTopic	ETM	fit(docs, doc_embeddings=…)	Clustering, not a posterior: topic-proportion uncertainty and effect estimation behave differently than the models above.

Specialized approaches. Start here when your design calls for one.

If your data or goal is…	Start with	Also consider	First calls	Note
Topics shift over time slices	DTM	DETM	fit(docs, times=…)	Prevalence and content evolve across periods; DETM adds embeddings.
Documents linked in a network (citations, replies)	RTM	—	fit(docs, links=…)	Models the text and the link graph jointly.
Documents in more than one language	PolylingualLDA	—	fit(doc_tuples)	Aligned topics across languages from translation-linked tuples.
Place authors or actors on an ideological scale	Wordfish	TBIP	fit(docs)	Scaling from word usage; TBIP adds a text-based ideal-point prior.
How tone or sentiment varies with metadata	STS	—	estimate_effect()	Sentiment-discourse decomposition; reach for it when tone is the question.

In code, topica.list_models(common_start=True) returns the common openings and list_models(group=…, brings=…) filters the rest.

Tour

• The models

  More than 40 models, from LDA through STM and its STS sentiment-discourse extension, HDP, dynamic and supervised topics, to short-text and embedding-based models, all with one consistent API.

• Preprocessing

  Tokenize, build a Corpus, prune the vocabulary, detect phrases, and split long documents while preserving metadata.

• Covariates & STM

  Relate topics to document metadata: prevalence and content covariates, effect estimation, clustered SEs, GLM links.

• Diagnostics & validation

  Coherence, exclusivity, intrusion tests, stability, alignment, ensemble consensus across runs, FREX labels, and pyLDAvis, all model-agnostic.

• Distinguishing words

  Fighting Words: which words separate two corpora, with significance.

• Short text

  Models built for tweets, headlines, and survey answers (PT, GSDMM).

• Held-out inference

  transform new documents onto a fitted model across every model family.

Everything returns NumPy arrays, fits are deterministic for a fixed seed, and the variational models parallelize across cores automatically.