Skip to content

Latest commit

 

History

History
24 lines (17 loc) · 1.38 KB

README.md

File metadata and controls

24 lines (17 loc) · 1.38 KB
Raw

law-turk

Predicting Decisions of Turkish Higher Courts.

Code to reproduce results given in "Natural Language Processing in Law: Prediction of outcomes in the Higher Courts of Turkey" by Emre Mumcuoğlu, Ceyhun E. Öztürk, Haldun M. Ozaktas and Aykut Koç (https://www.sciencedirect.com/science/article/abs/pii/S0306457321001692, https://doi.org/10.1016/j.ipm.2021.102684).

Requirements

  • scikit-learn==0.24.2
  • tensorflow==2.3.0 or tensorflow==2.8.0
  • gensim==3.8.3

Tested versions of the required packages are given above. The code was tested in Python 3.8.

The deep learning models require the use of word embeddings. Download a Turkish word embedding model into data. You can use the one we used at https://github.com/akoksal/Turkish-Word2Vec

Use

Simply call predict.py with appropriate arguments.

  • Court name: Should be one of constitutional, civil, criminal, administrative, taxation, constitutional_right1, constitutional_right2, constitutional_right3, constitutional_right4, constitutional_right5, constitutional_right6, constitutional_right7.
  • Model name: Should be one of Dummy, DT, RF, SVM, GRU, LSTM, BiLSTM.
  • Mode: Either training or test. Use test mode to print test results after training.
  • Optional argument --attention: Whether to use attention mechanism in deep learning models.

An example call:

python3 predict.py constitutional BiLSTM training --attention