This is the code of SVMs for Aspects Based Sentiment Analysis.
- python 3.x
- sklearn
- thundersvm
- numpy
- hyperopt
- matplotlib
- standfordcorenlp
You can run python dataset.py to process the two ABSA datasets. Notice that the path of the dataset is needed to be specified and the flag is_preprocessed should be False if you want to process the data from scratch.
The file search_feature_comb.py is used for searching the best features and parameters for our model. So you can run python search_feature_comb.py after the data preprocessing.
The codes for model visualization are stored in the folder named 'visualization'. t-sne.py is used for clustering model visualization by employing T-SNE. In addition, files named svm_visualize_*.py are for visualizing the trained SVMs in our model.