Coupled Multi-Layer Attentions for Co-Extraction of Aspect and Opinion Terms
This is an instruction file for successfully running the CMLA model of the paper "Coupled Multi-Layer Attentions for Co-Extraction of Aspect and Opinion Terms" published in AAAI 2017: http://www.aaai.org/Conferences/AAAI/2017/PreliminaryPapers/15-Wang-W-14441.pdf
This code makes use of theano for the implementation of GRU.
Please follow these steps to run the model:
-
Upload your sentence file and label file in util/data_semEval/
-
Under the folder 'util', run $ python 10seqLabel.py to generate structured input data.
-
Under the folder 'util', run $ python 20word_embedding.py to generate pre-trained word embeddings
-
Under the main folder, run $ python train_GRU_dropout_attention.py to train a model as well as the evaluation.