ASTE-Glove-Bert
ASTE-Data-V2
Data from Position-Aware Tagging for Aspect Sentiment Triplet Extraction (In EMNLP 2020) https://github.com/xuuuluuu/Position-Aware-Tagging-for-ASTE
ASTE-Graph-V2
heterogeneous graph generated by running *global_graph_.py in ./build_graph_glove/ Noted that the graphs are built in advance and saved in ./ASTE-Graph-V2
ASTE-Rele-Sentences
relevant sentences for every sentence in the dataset, which will be used to build graphs.
models/bert_moodels
save the models with glove ts is the model without any graphs ts0 to ts3 respond to baseline1 to baseline3, which use graph0 to graph3
bert_models
save the models with bert there is only one model in bert_init.py that does not use graphs
run with glove
CUDA_VISIBLE_DEVICES=0 python train_with_glove.py --dataset res14
run with bert
CUDA_VISIBLE_DEVICES=1 python train_with_bert.py --dataset res14