Tulpen / NNTranJSTagger

JointPS

The code of A Simple and Effective Neural Model for Joint Word Segmentation and POS Tagging.

Usage

MKL

If MKL is supported in your server, modify the MKL path in CMakeLists.txt first.

modify "set(MKL_ROOT /opt/intel/mkl)" to "set(MKL_ROOT your_mkl_path)"

Run

mkdir build
cd build
cmake .. or cmake .. -DMKL=True(if mkl is supported)
cd ..
./bin/NNJSTagger -l -train data/ctb50/train.corpus -dev data/ctb50/dev.corpus -test data/ctb50/test.corpus -option data/option.debug

Config

config file in ./data/option.debug
seg = true
dropProb = 0.25
adaAlpha = 0.001
charEmbFile = data/char.vec
bicharEmbFile = data/mini.bichar.vec
batchSize = 16

Network Structure

Performance

	CTB5	CTB6	CTB7	PKU	NCC
Model	SEG      POS	SEG      POS	SEG      POS	SEG      POS	SEG      POS
Our Model (No External Embeddings)	97.69    94.16	95.37    90.83	95.32    90.25	95.22    92.62	93.97    89.47
Our Model (Basic Embeddings)	97.93    94.44	95.78    91.79	95.77    91.12	95.82    93.42	94.52    89.82
Our Model (Word-context Embeddings)	98.50    94.95	96.36    92.51	96.25    91.87	96.35    94.14	95.30    90.42

Speed

Intel CPU: i7 6800k, MKL supported; GCC version 5.4.0

CTB6	Sentences	Time	Speed
Train	23k	465.41s	50.3 sents/s
Devel	2.1k	17.74s	117.1 sents/s
Test	2.8k	23.67s	118.1 sents/s

Cite

@Article{zhang2018jointposseg,  
  author    = {Zhang, Meishan and Yu, Nan and Fu, Guohong},  
  title     = {A Simple and Effective Neural Model for Joint Word Segmentation and POS Tagging},  
  journal   = {IEEE/ACM Transactions on Audio, Speech and Language Processing (TASLP)},  
  year      = {2018},  
  volume    = {26},  
  number    = {9},
  pages     = {1528--1538},
  publisher = {IEEE Press},
}

Question

• if you have any question, you can open a issue or email [e-mail]、[e-mail]、bamtercelboo@{gmail.com, 163.com}.

• if you have any good suggestions, you can PR or email me.

Authors

Meishan Zhang, Yu Nan, Zonglin Liu