Distributional semantics in R with the wordspace package

wordspace logo
current version: 0.2-8 (22 Aug 2022)
News:  [2023-01-15] CRAN compatibility versions: wordspace v0.2-8 and sparsesvd v0.2-2
[2020-01-07] wordspace v0.2-6 relased on CRAN
[2019-07-15] sparsesvd v0.2 relased on CRAN

Coling 2014 Poster
Poster presented at Coling 2014

Introduction

The wordspace package turns R into an interactive laboratory for empirical research on distributional semantic models (DSM). It consists of a small set of carefully designed functions, most of which (i) encapsulate non-trivial R operations in a user-friendly manner or (ii) provide efficient and memory-lean C implementations of key operations.

The package includes various example data sets, ranging from verb-noun co-occurrence data to standard evaluation tasks. Additional large pre-compiled models will be made available from the Web site.

See Evert (2014) for a more detailed description of the package, including benchmarks and a sample session. You can also read the package vignette (PDF, HTML) and try out all the code examples.

If you use wordspace in published research, it would be nice of you to cite the following paper:

Back to the top


Download & installation

The wordspace package and its dependencies are available from CRAN, the Comprehensive R Archive Network. It can easily be installed with a GUI installer or with the R command

  install.packages("wordspace")
  
Alternatively, you can download a local copy of the current version below and install it manually. Note that there are two separate packages, which need to be installed in this order: sparsesvd (a SVDLIBC wrapper) and wordspace itself.

The wordspace package is free and open source. Like R itself, it is distributed under the GNU General Public License, version 3. SVDLIBC and the sparsesvd package are made available under a BSD license.

Back to the top


Contact information

Developer and maintainer:

You should also subscribe to the wordspace mailing list, where you can get help and discuss ideas with other users.

Back to the top