This repository is for our EMNLP17 paper "Document-Level Multi-Aspect Sentiment Classification as Machine Comprehension". We thank Tao Lei as our code is developed based on his code.
git clone [email protected]:HKUST-KnowComp/DMSCMC.git
cd src
export PYTHONPATH=/path/to/DMSCMC/src- Run experiments
cd src
THEANO_FLAGS="floatX=float32,device=gpu0,cuda.root=/usr/local/cuda,on_unused_input=ignore,optimizer=fast_compile" python DMSCMC.py --emb ../data/tripadvisor/embs --train ../data/tripadvisor/train --dev ../data/tripadvisor/dev --test ../data/tripadvisor/test --save ../model/base --aspect_seeds ../data/tripadvisor/aspect.10.words - Run pre-trained model
cd src
THEANO_FLAGS="floatX=float32,device=gpu0,cuda.root=/usr/local/cuda,on_unused_input=ignore,optimizer=fast_compile" python DMSCMC.py --emb ../data/tripadvisor/embs --train ../data/tripadvisor/train --dev ../data/tripadvisor/dev --test ../data/tripadvisor/test --save ../model/base --aspect_seeds ../data/tripadvisor/aspect.10.words --model ../model/base