To cite this tool, please use the following reference:

Matthieu Constant and Joakim Nivre. 2016. A Transition-based System for Joint Lexical and Syntactic Analysis. 54th Annual Meeting of the Association for Computational Linguistics (ACL'16)

You first need to download the JSAP-2.1 library.

For both training and parsing, the tool requires a conll file as input. It is necessary to have the 7th and 9th columns (HEAD) identical and 8th and 10th columns (LABEL) identical as well.

java -Xmx8g -cp lgtools.jar:JSAP-2.1.jar fr/upem/lgtools/parser/Parser -m model -t train-dataset -p

• model is the path of the model you want to build
• train-dataset is the path of the training dataset in the conll format
• "-p" idenicates that we apply projective parsing algorithm

java -Xmx8g -cp lgtools.jar:JSAP-2.1.jar fr/upem/lgtools/parser/Parser -m model -i test-dataset -o output

• model is the path of the model to be applied
• test-dataset is the path of the input dataset to be tested
• output is the path of the output (parsed text)

The tool also prints out evaluation score on the test data set:

EVAL Fscore(MWE) Fscore(Fixed_MWE) Nan Nan UAS LAS