Relation Classification via Convolutional Deep Neural Network

TensorFlow implementation of the paper,

dataset: SemEval2010 task8

word embeddings: senna

to run the code:

./run

Environment(have tested)

• tensorflow 1.4.0
• python 3.5
• linux,macOs or windows

How to run ?

• to train model

  ./run

  where num_epochs=200 --word_dim=50have been set in 'run' file.

• to test model

  excute

  python src/train.py --num_epochs=200 --word_dim=50 --test

  then you can get a 'results.txt' file in /data/resuts.txt

• to calculate F1 score

  perl src/scorer.pl data/test_keys.txt data/results.txt

Problem

when you use Spyder or PyCharm to run this code, you may encounter this error:

ArgumentError: argument --train_file: conflicting option string: --train_file

solution:

1. restart spyder

2. or add annotation for all definitions of tf.flags.FLAGS .

such as # flags.DEFINE_string("train_file", "data/train.cln", "original training file")

Difference

1. delete 'the hidden layer 2' as the paper mentioned
2. use muti - window size(w=3,w=4,w=5) in convolution layer
3. delete Wordnet lexical feature