embed_docspace: Build a Starspace model for content-based recommendation

In ruimtehol: Learn Text 'Embeddings' with 'Starspace'

Build a Starspace model for content-based recommendation

Description

Build a Starspace model for content-based recommendation (docspace). For example a user clicks on a webpage and this webpage contains a bunch or words.

Usage

embed_docspace(
  x,
  model = "docspace.bin",
  early_stopping = 0.75,
  useBytes = FALSE,
  ...
)

Arguments

x	a data.frame with user interest containing the columns user_id, doc_id and text The user_id is an identifier of a user The doc_id is just an article or document identifier the text column is a character field which contains words which are part of the doc_id, words should be separated by a space and should not contain any tab characters
model	name of the model which will be saved, passed on to starspace
early_stopping	the percentage of the data that will be used as training data. If set to a value smaller than 1, 1-early_stopping percentage of the data which will be used as the validation set and early stopping will be executed. Defaults to 0.75.
useBytes	set to TRUE to avoid re-encoding when writing out train and/or test files. See writeLines for details
...	further arguments passed on to starspace except file, trainMode and fileFormat

Value

an object of class textspace as returned by starspace.

Examples

library(udpipe)
data(dekamer, package = "ruimtehol")
data(dekamer_theme_terminology, package = "ruimtehol")
## Which person is interested in which theme (aka document)
x <- table(dekamer$aut_person, dekamer$question_theme_main)
x <- as.data.frame(x)
colnames(x) <- c("user_id", "doc_id", "freq")
## Characterise the themes (aka document)
docs <- split(dekamer_theme_terminology, dekamer_theme_terminology$theme)
docs <- lapply(docs, FUN=function(x){
  data.frame(theme = x$theme[1], text = paste(x$term, collapse = " "),
             stringsAsFactors=FALSE)
})
docs <- do.call(rbind, docs)

## Build a model
train <- merge(x, docs, by.x = "doc_id", by.y = "theme")
train <- subset(train, user_id %in% sample(levels(train$user_id), 4))
set.seed(123456789)
model <- embed_docspace(train, dim = 10, early_stopping = 1)
plot(model)

ruimtehol documentation built on Jan. 7, 2023, 1:25 a.m.