A Fast and Accurate Vietnamese Word Segmenter

The implementation of RDRsegmenter, as described in our paper:

@InProceedings{NguyenNVDJ2018,
author={Dat Quoc Nguyen and Dai Quoc Nguyen and Thanh Vu and Mark Dras and Mark Johnson},
title={{A Fast and Accurate Vietnamese Word Segmenter}},
booktitle={Proceedings of the 11th International Conference on Language Resources and Evaluation (LREC 2018)},
pages={2582--2587},
year={2018}
}

Please CITE our paper whenever RDRsegmenter is used to produce published results or incorporated into other software.

RDRsegmenter has also been incorporated into our Java NLP annotation pipeline VnCoreNLP for Vietnamese. VnCoreNLP provides rich linguistic annotations through key NLP components of word segmentation, POS tagging, named entity recognition and dependency parsing.

Usage

Supposed that Java 1.8+ is already set to run via command line or terminal (for example: adding Java to the path environment variable in Windows OS). To compile RDRsegmenter, users simply run the following command:

$ javac -encoding UTF-8 RDRsegmenter.java

Then users can run RDRsegmenter to segment all raw text corpora in a directory, that share the same file name suffix:

$ java RDRsegmenter PATH-TO-DIRECTORY FILE-NAME-SUFFIX
// Segment a raw text corpus in a directory is as follows: 
// java RDRsegmenter PATH-TO-DIRECTORY CORPUS-FILE-NAME

.WSeg-suffixed files will be generated in the same directory containing the raw text corpora.

See Readme.md in train directory for details of retraining RDRsegmenter.

API

// Add absolute path to "VnVocab" in line 14 in "Vocabulary.java" if necessary
// Add absolute path to "Model.RDR" in line 25 in "RDRsegmenter.java" if necessary
RDRsegmenter segmenter = new RDRsegmenter(); //Model loading

// Perform word segmentation on a text string
String outStr1 = segmenter.segmentRawString("Tính thuế thu nhập cá nhân (TNCN) từ tiền lương.");
String outStr2 = segmenter.segmentTokenizedString("Tính thuế thu nhập cá nhân ( TNCN ) từ tiền lương .");

// Perform word segmentation on text corpora
segmenter.segmentDirectory("PATH-TO-DIRECTORY", "FILE-NAME-SUFFIX");
segmenter.segmentRawCorpus("PATH-TO-RAW-TEXT-CORPUS");
segmenter.segmentTokenizedCorpus("PATH-TO-TOKENIZED-TEXT-CORPUS");