Implementation of ABCNN(Attention-Based Convolutional Neural Network) on Pytorch
Attention-Based Convolutional Neural Network for Modeling Sentence Pairs
pip3 install -r requirements.txt
to install project dependencies.python3 main.py
.├── README.md
├── sample_data/ # empty directory because of license
├── abcnn.py # model
├── dataset.py # data load
├── main.py
├── options.toml # options
├── requirements.txt
└── train.py # training function
[model]
embeddeddimension = 200 # embedding vector size
strlenmax = 15 #sentence length
filterwidth = 1
filterchannel = 130
layersize = 2
inception = true # variety receptive field
distance = 'cosine' # cosine or manhattan
jype1, konlpy, mecab are for korean dataset you don't have to use dataset.py and these libraries.