| Title: | Gaussian Mixture Text Models for Topic Analysis |
| Version: | 0.1.0 |
| Maintainer: | Kohei Watanabe <watanabe.kohei@gmail.com> |
| Description: | Gaussian mixture models and k-means for topic analysis of dense document vectors. The underlying clustering functions rely on the Armadillo library. |
| License: | Apache License (≥ 2.0) |
| Encoding: | UTF-8 |
| Depends: | R (≥ 3.5.0) |
| Imports: | quanteda (≥ 4.0.0), proxyC, wordvector, Rcpp, RcppArmadillo |
| Suggests: | testthat, spelling, rmarkdown, knitr |
| LinkingTo: | Rcpp, RcppArmadillo (≥ 0.7.600.1.0), quanteda, testthat |
| Language: | en-US |
| Config/roxygen2/version: | 8.0.0 |
| NeedsCompilation: | yes |
| Packaged: | 2026-09-06 02:28:38 UTC; watan |
| Author: | Kohei Watanabe 
[aut, cre, cph] |
| Repository: | CRAN |
| Date/Publication: | 2026-09-15 11:20:09 UTC |
Extract words for topics from documents
Description
Identify distinctive words for each topic by applying TF-IDF weights to the original quanteda::dfm.
Usage
terms(x, data, n = 10, ...)
Arguments
x |
a fitted model. |
data |
a quanteda::dfm or quanteda::tokens from which words are extracted for each topic. |
n |
the number of topic words. |
... |
passed to functions. |
Details
To identify distinctive words for topics, original documents must be provided
along with a fitted model because the information about individual words are lost in
document vectors.
The documents in data is grouped by topic and weighted by TF-IDF
to select the most distinctive words for each topic. This technique is
commonly known as c-TF-IDF.
Value
Returns a character matrix with the most distinctive words for each topic.
Gaussian mixture model for topic analysis
Description
Gaussian mixture model for clustering of document vectors based on the Armadillo library.
Usage
textmodel_gmm(
x,
k = 10,
model = NULL,
...,
verbose = quanteda_options("verbose")
)
Arguments
x |
a wordvector::textmodel_doc2vec or a dense matrix of document vectors in the rows. |
k |
the number of topics to identify. |
model |
a fitted model from which initial centroids are extracted. |
... |
passed to the underlying function. |
verbose |
print the progress if |
Details
User can change the number of threads for the parallel computing via
options(GMTM.threads) or OMP_THREAD_LIMIT in the environmental
variable.
Value
Returns a fitted textmodel_gmm object.
Examples
# dummy document vectors with 50 dimensions
mat <- t(replicate(1000, rnorm(50)))
gmm <- textmodel_gmm(mat, k = 10)
table(topics(gmm))
K-means for topic analysis
Description
K-means clustering of document vectors based on the Armadillo library.
Usage
textmodel_kmeans(
x,
k = 10,
model = NULL,
...,
verbose = quanteda_options("verbose")
)
Arguments
x |
a wordvector::textmodel_doc2vec or a dense matrix of document vectors in the rows. |
k |
the number of topics to identify. |
model |
a fitted model from which initial centroids are extracted. |
... |
passed to the underlying function. |
verbose |
print the progress if |
Value
Returns a fitted textmodel_kmeans object.
Examples
# dummy document vectors with 50 dimensions
mat <- t(replicate(1000, rnorm(50)))
km <- textmodel_kmeans(mat, k = 10)
table(topics(km))
Extract topics of documents
Description
Extract topics of documents
Usage
topics(x, ...)
Arguments
x |
a fitted model. |
... |
not used. |
Value
Returns predicted topics as a vector.