Learn Text 'Embeddings' with 'Starspace'

Wraps the 'StarSpace' library < https://github.com/facebookresearch/StarSpace> allowing users to calculate word, sentence, article, document, webpage, link and entity 'embeddings'. By using the 'embeddings', you can perform text based multi-label classification, find similarities between texts and categories, do collaborative-filtering based recommendation as well as content-based recommendation, find out relations between entities, calculate graph 'embeddings' as well as perform semi-supervised learning and multi-task learning on plain text. The techniques are explained in detail in the paper: 'StarSpace: Embed All The Things!' by Wu et al. (2017), available at .

This repository contains an R package which wraps the StarSpace C++ library (https://github.com/facebookresearch/StarSpace), allowing the following:

  • Text classification
  • Learning word, sentence or document level embeddings
  • Finding sentence or document similarity
  • Ranking web documents
  • Content-based recommendation (e.g. recommend text/music based on the content)
  • Collaborative filtering based recommendation (e.g. recommend text/music based on interest)
  • Identification of entity relationships


  • For regular users, install the package from your local CRAN mirror install.packages("ruimtehol")
  • For installing the development version of this package: devtools::install_github("bnosac/ruimtehol", build_vignettes = TRUE)

Look to the vignette and the documentation of the functions

vignette("ground-control-to-ruimtehol", package = "ruimtehol")
help(package = "ruimtehol")

Main functionalities

This R package allows to Build Starspace models on your own text / Get embeddings of words/ngrams/sentences/documents/labels / Get predictions from a model (e.g. classification / ranking) / Get nearest neighbours similarity

The following functions are made available.

Function Functionality
starspace Low-level interface to build a Starspace model
starspace_load_model Load a pre-trained model or a tab-separated file
starspace_save_model Save a Starspace model
starspace_embedding Get embeddings of documents/words/ngrams/labels
starspace_knn Find k-nearest neighbouring information for new text
starspace_dictonary Get words/labels part of the model dictionary
predict.textspace Get predictions along a Starspace model
as.matrix Get words and label embeddings
embedding_similarity Cosine/dot product similarity between embeddings - top-n most similar text
embed_wordspace Build a Starspace model which calculates word/ngram embeddings
embed_sentencespace Build a Starspace model which calculates sentence embeddings
embed_articlespace Build a Starspace model for embedding articles - sentence-article similarities
embed_tagspace Build a Starspace model for multi-label classification
embed_docspace Build a Starspace model for content-based recommendation
embed_pagespace Build a Starspace model for interest-based recommendation
embed_entityrelationspace Build a Starspace model for entity relationship completion


Short example showing word embeddings

## Get some training data
download.file("https://s3.amazonaws.com/fair-data/starspace/wikipedia_train250k.tgz", "wikipedia_train250k.tgz")
x <- readLines("wikipedia_train250k.tgz", encoding = "UTF-8")
x <- x[-c(1:9)]
x <- x[sample(x = length(x), size = 10000)]
writeLines(text = x, sep = "\n", con = "wikipedia_train10k.txt")
## Train
model <- starspace(file = "wikipedia_train10k.txt", fileFormat = "labelDoc", dim = 10, trainMode = 3)
Object of class textspace
 dimension of the embedding: 10
 training arguments:
      loss: hinge
      margin: 0.05
      similarity: cosine
      epoch: 5
      adagrad: TRUE
      lr: 0.01
      termLr: 1e-09
      norm: 1
      maxNegSamples: 10
      negSearchLimit: 50
      p: 0.5
      shareEmb: TRUE
      ws: 5
      dropoutLHS: 0
      dropoutRHS: 0
      initRandSd: 0.001
embedding <- as.matrix(model)
embedding[c("school", "house"), ]
              [,1]         [,2]        [,3]        [,4]         [,5]        [,6]       [,7]       [,8]         [,9]       [,10]
school 0.008395348  0.002858619 0.004770191 -0.03791502 -0.016193179 0.008368539 -0.0221493 0.01587386 -0.002012054 0.029385706
house  0.005371093 -0.007831781 0.010563998  0.01040361  0.000616577 0.005770847 -0.0097075 0.01678141 -0.004738560 0.009139475
dictionary <- starspace_dictionary(model)
## Save trained model as a binary file or as TSV so that you can inspect the embeddings e.g. with data.table::fread("wikipedia_embeddings.tsv")
starspace_save_model(model, file = "textspace.ruimtehol",      method = "ruimtehol")
starspace_save_model(model, file = "wikipedia_embeddings.tsv", method = "tsv-data.table")
## Load a pre-trained model or pre-trained embeddings
model <- starspace_load_model("textspace.ruimtehol",      method = "ruimtehol")
model <- starspace_load_model("wikipedia_embeddings.tsv", method = "tsv-data.table", trainMode = 3)
## Get the document embedding
starspace_embedding(model, "get the embedding of a full document")
                                          [,1]        [,2]      [,3]       [,4]      [,5]      [,6]       [,7]      [,8]     [,9]     [,10]
get the embedding of a full document 0.1489144 -0.09543591 0.1242385 -0.1080941 0.6971645 0.3131362 -0.3405705 0.3293449 0.231894 -0.281555

The following functionalities do similar things. They see what is the closest word or sentence to a provided sentence.

## What is closest term from the dictionary
starspace_knn(model, "What does this bunch of text look like", k = 10)
## What is closest sentence to vector of sentences
predict(model, newdata = "what does this bunch of text look like", 
        basedoc = c("what does this bunch of text look like", 
                    "word abracadabra was not part of the dictionary", 
                    "give me back my mojo",
                    "cosine distance is what i show"))
## Get cosine distance between 2 sentence vectors
  starspace_embedding(model, "what does this bunch of text look like"),
  starspace_embedding(model, "word abracadabra was not part of the dictionary"), 
  type = "cosine")

Short example showing classification modelling (tagspace)

Below Starspace is used for classification

data("dekamer", package = "ruimtehol")
dekamer$x <- strsplit(dekamer$question, "\\W")
dekamer$x <- sapply(dekamer$x, FUN = function(x) paste(setdiff(x, ""), collapse = " "))
dekamer$x <- tolower(dekamer$x)
dekamer$y <- strsplit(dekamer$question_theme, split = ",")
dekamer$y <- lapply(dekamer$y, FUN=function(x) gsub(" ", "-", x))
model <- embed_tagspace(x = dekamer$x, y = dekamer$y,
                        dim = 50, 
                        lr = 0.01, epoch = 40, loss = "softmax", adagrad = TRUE, 
                        similarity = "cosine", negSearchLimit = 50,
                        ngrams = 2, minCount = 2)
text <- c("de nmbs heeft het treinaanbod uitgebreid via onteigening ...",
          "de migranten komen naar europa de asielcentra ...")                   
predict(model, text, k = 3)  
predict(model, "koning filip", k = 10, type = "knn")
predict(model, "koning filip", k = 10, type = "embedding")


  • Why did you call the package ruimtehol? Because that is the translation of StarSpace in WestVlaams.
  • The R wrapper is distributed under the Mozilla Public License 2.0. The package contains a copy of the StarSpace C++ code (namely all code under src/Starspace) which has a BSD license (which is available in file LICENSE.notes) and also has an accompanying PATENTS file which you can inspect here.

Support in text mining

Need support in text mining? Contact BNOSAC: http://www.bnosac.be


CHANGES IN rumitehol VERSION 0.1.2

  • Changes to src/Makevars
    • Added -pthread in PKG_CPPFLAGS and removed usage of SHLIB_PTHREAD_FLAGS

CHANGES IN rumitehol VERSION 0.1.1

  • Initial release based on STARSPACE-2017-2.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


0.1.2 by Jan Wijffels, 2 months ago


Browse source code at https://github.com/cran/ruimtehol

Authors: Jan Wijffels [aut, cre, cph] (R wrapper) , BNOSAC [cph] (R wrapper) , Facebook , Inc. [cph] (Starspace (BSD licensed))

Documentation:   PDF Manual  

MPL-2.0 license

Imports Rcpp, utils, graphics, stats

Suggests udpipe, data.table

Linking to Rcpp, BH

System requirements: C++11, GNU make

See at CRAN