Weigoss / ccks2018task2

Joint Intent Detection and Slot Filling with Rules

• authors: Shiya Ren, Huaming Wang, Dongming Yu, Yuan Li and Zhixing Li
• The specific details of this task can be viewed at https://biendata.com/competition/CCKS2018_2/.

1. Prerequisites

• Python 3.6
• TensorFlow 1.5.0
• keras 2.1.6

2. Setup

2.1 Reproduce the submitted result

• if you want to use our trained model directly. The predictive label in which the intent is predicted is the best result of the submitted results.

python mainFunc.py -o data/result/result_mlnlp-ymm.txt

2.2 Reproduce results from scratch

• if you want to retrain sequence labeling model by youself.

python mainFunc.py --pycrf -o data/result/result_mlnlp-ymm.txt

• if you want to retrain classification model by youself.

python mainFunc.py --classify -o data/result/result_mlnlp-ymm.txt

• if you want to retrain both of the models.

python mainFunc.py --all -o data/result/result_mlnlp-ymm.txt

• The final result will be stored in "data/result/result_mlnlp-ymm.txt" You can open it as you like.