This is the last project of our course: Internet-based Information Extraction. We need to extract all employee relations from a text.

1. install anaconda3

2. create a new environment with python2.7 in anaconda

   • conda create -n re-py27 python=2.7

   • conda activate re-py27

3. install pyltp

   • pip install pyltp (会自动编译，所以会花十来分钟)

4. download LTP 3.4 model for pyltp
   把模型解压到项目文件夹下，并重命名为ltp_data

5. install libsvm
   linux 平台无法使用conda/pip直接安装，也无法通过wheel文件安装，windows平台可以通过wheel安装

   • download source code
     • git clone https://github.com/cjlin1/libsvm
   • compile
     • cd libsvm
     • make
     • cd python
     • make
   • copy files to re-py27 environment

     • copy

       • libsvm/python/commonutil.py
       • libsvm/python/svm.py
       • libsvm/python/svmtil.py

       to /home/zhouyiyuan/anaconda3/envs/re-py27/lib/python2.7/

   • copy libsvm/libsvm.so.2 to /home/zhouyiyuan/anaconda3/envs/re-py27/lib/

6. download Chinese word vectors list

   • choose any one txt file from the github, then make a new dir and extract the txt file there
     • mkdir word_list
     • mv \your path\download_wordlist.txt \word_list

• pyltp
  用于分词/词性标注/命名实体识别

  • 教程 https://www.jianshu.com/p/f78453f5d1ca
  • 在线文档 https://pyltp.readthedocs.io/zh_CN/latest/api.html
  • 标注规则 https://ltp.readthedocs.io/zh_CN/latest/appendix.html#id4

• libsvm
  svm api: 训练，预测， 读取数据

  • 教程 https://www.cnblogs.com/Finley/p/5329417.html

1. run the following command to train and test:

bash train_and_test.sh

2. to train only:

   bash train.sh

3. to test only:

   bash test.sh

• train

• test

• feature extraction