quanteda.classifiers: Text classification textmodel extensions for quanteda

Installation

To install this package, use the following, which also installs what the R keras package needs in order to run.

# devtools package required to install quanteda from Github 
devtools::install_github("quanteda/quanteda.classifiers") 

keras::install_keras(method = "conda")

Available classifiers

This package contains two experimental methods that are built on top of the keras package. (The SVM models have been moved to quanteda.textmodels.)

Classifier	Command
Multilevel perceptron network	textmodel_mlp()
Convolutional neural network + LSTM model fitted to word embeddings	textmodel_cnnlstmemb()

Available human-annotated corpora

Corpus	Name
Sentence-level corpus of UK party manifestos 1945–2019, partially annotated	data_corpus_manifestosentsUK
Large Movie Review Dataset of 50,000 annotated highly polar movie reviews for training and testing, from Maas et. al. (2011)	data_corpus_LMRD

Demonstration

See this (very preliminary!) performance comparison.

How to cite

Benoit, Kenneth, Patrick Chester, and Stefan Müller (2019). quanteda.classifiers: Models for supervised text classification. R package version 0.2. URL: http://github.com/quanteda/quanteda.svm.

For a BibTeX entry, use the output from citation(package = “quanteda.classifiers”).

Issues

• Please file an issue (with a bug, wish list, etc.) via GitHub.