topic_cooccurrences: Get topic cooccurrences.
In PolMine/polmineR.topics: tools for topicmodelling

Description Usage Arguments Value Examples

Topic models describe documents as composed of different topics. This property can be used to obtain co-occuurrence statistics of topics.

## S4 method for signature 'TopicModel'
cooccurrences(
  .Object,
  k,
  regex = NULL,
  docs = NULL,
  renumber = NULL,
  method = "chisquare",
  progress = TRUE,
  verbose = TRUE
)

## S4 method for signature 'matrix'
cooccurrences(
  .Object,
  regex = NULL,
  docs = NULL,
  renumber = NULL,
  method = "chisquare",
  progress = TRUE,
  verbose = TRUE
)

`.Object`	Either an object inheriting from the `TopicModel` class (such as `LDA_Gibbs`), or a `matrix` with the topics present in documents. If `.Object` is a `matrix`, each column is expected to represent the top k topics present in a document. The `matrix` returned by the `topics`-method from the `topicmodels` package was used to develop the method, but document-term-matrices derived otherwise tools may work as well.
`k`	An `integer` value, the `k` first topics to consider when deriving the document-topic-matrix from a trained topicmodel.
`regex`	If not `NULL` (default), the procedure will be limited to document names matched by the regular expression stated by `regex`.
`docs`	If not `NULL`, the procedure will be limited to documents matching the character string.
`renumber`	If not `NULL` (the default), topics in the document-topic-matrix will be renumbered according to the argument `renumber`. If `renumber` is an integer vector, the length of this vector is required to match the number of topics in topic model. Each topic i present in the document-topic matrix will be mapped on the value at position i of the vector. If `renumber` is a `list` of integer vectors, these vectors are considered as groups of topics that represent a single implict "super-topic". For each integer vector present in the `list`, the topic numbers present in the document-topic-matrix will be matched on the first value of the vector.
`method`	The statistic to calculate co-occurrences, "chisquare" by default.
`progress`	A `logical` value, whether to show a progress bar.
`verbose`	A `logical` value, whether to output messages on the state of affairs.

A data.table with co-occurrence statistics with at least the following columns:

a: number of the topic of interest
b: number of the co-occurring topic
b_total: number of total occurrences of topic b; if the document-topic matrix has been renumbered, the times at least one of the topics in a group occurs in a docuent
b_total: number of total occurrences of topic a; if the document-topic matrix has been renumbered, the times at least one of the topics in a group occurs in a docuent
count_coi: number of joint occurrences of topics a and b
count_ref: number of occurrences of b without co-occurring of a

If argument method is not NULL, additional columns will be included in the topic co-occurrence table. E.g. if method is "chisquare", a column "exp_coi", will report the expected number of occurrences of b together with a, column "chisquare" will report the value of the chi squared test, and a column "rank_chisquare" will report the rank of the statistical significance of the co-occurrence of a and b according to the chi squared test.

data(BE_lda, BE_labels)
dt <- cooccurrences(BE_lda, k = 3L)
topics_to_drop <- grep("^\\(.*?\\)$", BE_labels)
dt_min <- dt[chisquare >= 10.83][!a %in% topics_to_drop][!b %in% topics_to_drop]
dt_min[, "a_label" := BE_labels[ dt_min[["a"]] ] ]
dt_min[, "b_label" := BE_labels[ dt_min[["b"]] ] ]

# Using the cooccurrence data for generating a network visualisation
if (requireNamespace("igraph")){
g <- igraph::graph_from_data_frame(
  d = data.frame(
    from = dt_min[["a_label"]],
    to = dt_min[["b_label"]],
    n = dt_min[["count_coi"]],
    stringsAsFactors = FALSE
  ),
  directed = TRUE
)
g <- igraph::as.undirected(g, mode = "collapse")
if (interactive()){
  igraph::plot.igraph(
    g, shape = "square", vertex.color = "steelblue",
    label = igraph::V(g)$name, label.family = 11, label.cex = 0.5
  )
}
}

# Example how to use the argument 'renumber' if a concept is represented by 
# several topics 
renumber_li <- list(
  school = grep("Grundschule", BE_labels),
  cummunity = grep("Gemeindeentwicklung", BE_labels),
  traffic = grep("Verkehrsmittel", BE_labels)
)
dt <- cooccurrences(BE_lda, k = 3L, renumber = renumber_li)
dt[a == grep("Grundschule", BE_labels)[1]][chisquare > 10.83]