ZORE Files

This directory provide the Java interface for Chinese word segmentation, POS tagging and dependency parsing using zpar under the POS tag set and dependency tag set of Peking University Multi-view Treebank (PMT) on 64-bit windows and linux (fedora). 

The dll files are compiled with the cross-platform, open-source IDE "Code::blocks" (Version 13.12) with MinGW64 c++ compiler on Windows 8.1 and linux (fedora), respectively.

You may use the dlls for Java interface directly without compiling.
Using zpar in 64bit Java (V1.7) on Windows.
(1)Trained models: One trained model "model_parse_pmt1" for parsing, and two trained model "model_tag_pfr1, model_tag_pfr6" for joint segmentation and tagging are in the dir "model/".
(2)Dlls: Copy the dlls "cn_nlp_Parser.dll,cn_nlp_Tagger.dll" together with the two dlls "libgcc_s_seh-1.dll,libstdc++-6.dll" from "dll/64/" to your java project directory.
(3)Examples: Examples of using ZParser and ZTagger are given in the dir "src/cn". You can use the parser and tagger separately or jointly referring to the usages in the examples. 
(4)User dict for word segmentation and POS tagging: In particular, you can give a userdict usingthe file "userdict.txt". In this file (utf-8 encoding), each line contains a word and a POS tag with a tab between them. If you do not have a proper POS tag for some words, you may use the default tag "n" for these words.
Note: the tagger and parser can not process files whose names contain Chinese characters.

About the models:
(1)The model "model_parser_arceager_mvt_origin_autopos" can be used for Chinese dependency parsing. 

If you use them, please cite the following paper:

@InProceedings{qiu-EtAl:2014:Coling2,
  author    = {Qiu, Likun  and  Zhang, Yue  and  Jin, Peng  and  Wang, Houfeng},
  title     = {Multi-view Chinese Treebanking},
  booktitle = {Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers},
  month     = {August},
  year      = {2014},
  address   = {Dublin, Ireland},
  publisher = {Dublin City University and Association for Computational Linguistics},
  pages     = {257--268},
  url       = {http://www.aclweb.org/anthology/C14-1026}
}
(2)The models "model_tag_science" and "model_tag_pfr6" are trained on the People's Daily Corpus in January 1998 and a few sentences from scientific domain, and the People's Daily Corpus in January to June, 2000, respectively. 2000, respectively.

If you use them in your paper, please cite the following paper:
@InProceedings{qiu-EtAl:2014:Coling2,
  author    = {Qiu, Likun  and  Zhang, Yue  and  Jin, Peng  and  Wang, Houfeng},
  title     = {Multi-view Chinese Treebanking},
  booktitle = {Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers},
  month     = {August},
  year      = {2014},
  address   = {Dublin, Ireland},
  publisher = {Dublin City University and Association for Computational Linguistics},
  pages     = {257--268},
  url       = {http://www.aclweb.org/anthology/C14-1026}
}
@article{yu2003specification,
  title={Specification for corpus processing at Peking University: Word segmentation, {POS} tagging and phonetic notation},
  author={Yu, Shiwen and Duan, Huiming and Zhu, Xuefeng and Swen, Bin and Chang, Baobao},
  journal={Journal of {Chinese} Language and Computing},
  volume={13},
  number={2},
  pages={121--158},
  year={2003}
}