Construct a topic model

make_topic_model() moves a table of texts through the necessary steps of preparation before building a topic model. The function applies seven steps:

1. identifies text divisions by the doc_id column

2. divides each of the texts into same-sized chunks of sample_size words (default is 1000 words)

3. unnests text table into a table with one word per row

4. removes stop words and proper nouns (identified as any word that only appears with a capitalized first letter)

5. counts word frequencies for each chunk

6. converts the table of frequencies into a document term matrix

7. builds a topic model with k topics

Usage

make_topic_model(df, by = doc_id, sample_size = 1000, k = 15, cache = TRUE)

Arguments

df

A data frame with nested text in a "text" column.

by

The column for identifying each document. By default, the "title" column will be used.

sample_size

The sample size for each document chunk. By default, samples will include 1000 words.

k

The number of topics to search for. By default, 15 topics will be sought.

cache

Whether to cache the resulting model as an RDS file in the "data/" folder. Set to TRUE by default. Delete this RDS file to create a new model.

Value

A topic model.

Examples

if (FALSE) { # \dontrun{
mysteries <- load_texts("mystery-novels", word = FALSE)

mysteries_lda <- mysteries |>
  make_topic_model(k = 10)
  } # }