A balanced sample of 1,679 floor speeches from the U.S. House and Senate,
Congresses 100-111 (1987-2011). Speeches are sampled evenly across
party x chamber x congress so covariate effects are estimable, then lowercased
and pruned of stop words and rare terms. Metadata: party
(Democrat/Republican), chamber (House/Senate), congress (100-111). Built
to showcase multiple (crossed) content covariates and over-time prevalence.
Source
Congressional Record, Hein-bound edition (Gentzkow, Shapiro & Taddy), congresses 100-111. The underlying floor speeches are U.S. government works (public domain).
Examples
data(congress)
fit <- stm(congress, K = 12, prevalence = ~ party + s(congress),
content = ~ party + chamber)
#> faSTM: crossing 2 content covariates (party, chamber) into a saturated content model with 4 groups.
#> faSTM fits with its own variational optimizer, so its topic decomposition differs from a stm::stm() run on the same data (a different but valid optimum, not a relabeling). Post-fit numbers match stm given the same model; see the Validation article. Silence with options(faSTM.quiet = TRUE).
#> faSTM: fitting K=12 on 1679 docs (batch)...