Skip to content

Latest commit

 

History

History
60 lines (44 loc) · 1.88 KB

README.md

File metadata and controls

60 lines (44 loc) · 1.88 KB

MLN-based Relation Extraction

About

Markov Logic Newtwork(MLN)-based Relation Extraction Model. This is the re-implementation of "Han and Sun. Global distant supervision for relation extraction. AAAI. 2016" for Korean Dataset.

Prerequisite

  • Python 3.5+
    • numpy 1.13.0+
    • scikit-learn 0.18.2+
  • Alchemy 1.0

How to use

Configuration

  1. Copy config_sample.py to config.py
  2. Edit variables in config.py fit to your environment.
  • data_path : location of data directory
  • alchemy_path : location of binary file directory of Alchemy 1.0

An example of config.py

# data directory
data_path = './data/'

# alchemy path
alchemy_path = '/home/user0/alchemy/bin/'

Location of data files

  • Training file : ./data/train_data
  • Test file : ./data/test_data

Training

python3 train.py

Test

python3 test.py

After test script is run, you can check the result on ./data/prec_recall_per_prop.txt

Licenses

Maintainer

Kijong Han han0ah@kaist.ac.kr

Publisher

Machine Reading Lab @ KAIST

Citation

  • Kijong Han, Sangha Nam, Younggyun Hahm, Jiseong Kim, Jin-Dong Kim, Key-Sun Choi, "Analysis of Distant Supervision for Relation Extraction Dataset", The 17th International Semantic Web Conference (ISWC 2018), Posters and Demonstrations, 2018.

Acknowledgement

This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIT) (2013-0-00109, WiseKB: Big data based self-evolving knowledge base and reasoning platform)