textmodel_lsa: Latent Semantic Analysis model

In wordvector: Word and Document Vector Models

Latent Semantic Analysis model

Description

Train a Latent Semantic Analysis model (Deerwester et al., 1990) on a quanteda::tokens object.

Usage

textmodel_lsa(
  x,
  dim = 50,
  min_count = 5L,
  engine = c("RSpectra", "irlba", "rsvd"),
  weight = "count",
  verbose = FALSE,
  ...
)

Arguments

x	a quanteda::tokens object.
dim	the size of the word vectors.
min_count	the minimum frequency of the words. Words less frequent than this in x are removed before training.
engine	select the engine perform SVD to generate word vectors.
weight	weighting scheme passed to quanteda::dfm_weight().
verbose	if TRUE, print the progress of training.
...	additional arguments.

Value

Returns a textmodel_wordvector object with the following elements:

values	a matrix for word vectors values.
weights	a matrix for word vectors weights.
frequency	the frequency of words in x.
engine	the SVD engine used.
weight	weighting scheme.
concatenator	the concatenator in x.
call	the command used to execute the function.
version	the version of the wordvector package.

References

Deerwester, S. C., Dumais, S. T., Landauer, T. K., Furnas, G. W., & Harshman, R. A. (1990). Indexing by latent semantic analysis. JASIS, 41(6), 391–407.

Examples

library(quanteda)
library(wordvector)

# pre-processing
corp <- corpus_reshape(data_corpus_news2014)
toks <- tokens(corp, remove_punct = TRUE, remove_symbols = TRUE) %>% 
   tokens_remove(stopwords("en", "marimo"), padding = TRUE) %>% 
   tokens_select("^[a-zA-Z-]+$", valuetype = "regex", case_insensitive = FALSE,
                 padding = TRUE) %>% 
   tokens_tolower()

# train LSA
lsa <- textmodel_lsa(toks, dim = 50, min_count = 5, verbose = TRUE)

# find similar words
head(similarity(lsa, c("berlin", "germany", "france"), mode = "words"))
head(similarity(lsa, c("berlin" = 1, "germany" = -1, "france" = 1), mode = "values"))
head(similarity(lsa, analogy(~ berlin - germany + france)))

wordvector documentation built on April 12, 2025, 2:23 a.m.