Conditional Random Fields for Labelling Sequential Data in Natural Language Processing.

Labelling Sequential Data in Natural Language Processing

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

• Fit a Conditional Random Field model (1st-order linear-chain Markov)
• Use the model to get predictions alongside the model on new data
• The focus of the implementation is in the area of Natural Language Processing where this R package allows you to easily build and apply models for named entity recognition, text chunking, part of speech tagging, intent recognition or classification of any category you have in mind.

For users unfamiliar with Conditional Random Field (CRF) models, you can read this excellent tutorial https://homepages.inf.ed.ac.uk/csutton/publications/crftut-fnt.pdf

Installation

• The package is on CRAN, so just install it with the command install.packages("crfsuite")
• For installing the development version of this package: devtools::install_github("bnosac/crfsuite", build_vignettes = TRUE)

Model building and tagging

For detailed documentation on how to build your own CRF tagger for doing NER / Chunking. Look to the vignette.

library(crfsuite)
vignette("crfsuite-nlp", package = "crfsuite")

Short example

library(crfsuite)

## Get example training data + enrich with token and part of speech 2 words before/after each token
x <- ner_download_modeldata("conll2002-nl")
x <- crf_cbind_attributes(x, 
                          terms = c("token", "pos"), by = c("doc_id", "sentence_id"), 
                          from = -2, to = 2, ngram_max = 3, sep = "-")

## Split in train/test set
crf_train <- subset(x, data == "ned.train")
crf_test <- subset(x, data == "testa")

## Build the crf model
attributes <- grep("token|pos", colnames(x), value=TRUE)
model <- crf(y = crf_train$label, 
             x = crf_train[, attributes], 
             group = crf_train$doc_id, 
             method = "lbfgs", options = list(max_iterations = 25, feature.minfreq = 5, c1 = 0, c2 = 1)) 
model

## Use the model to score on existing tokenised data
scores <- predict(model, newdata = crf_test[, attributes], group = crf_test$doc_id)

table(scores$label)
 B-LOC B-MISC  B-ORG  B-PER  I-LOC I-MISC  I-ORG  I-PER      O 
   261    211    182    693     24    205    209    605  35297 

Build custom CRFsuite models

The package itself does not contain any models to do NER or Chunking. It's a package which facilitates creating your own CRF model for doing Named Entity Recognition or Chunking on your own data with your own categories.

In order to facilitate creating training data of your own text, a shiny app is made available in this R package which allows you to easily tag your own chunks of text, using your own categories. More details about how to launch the app, which data is needed for building a model, how to start to build and use your model - read the vignette in detail: vignette("crfsuite-nlp", package = "crfsuite").

Support in text mining

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