##Goal: compare the difference between Linear LSTM and TreeLSTM , does Linear LSTM can learn the "tree structure"?
##Subtask1: ###Slot1: LibSVM to classify OTE from sentence ###Slot2: 還在想 ###Slot3: * auto-encoder(Linear LSTM) * TreeLSTM concat sentence embedding and aspect info to predict polarity
##SubTask2: Use Slot1's classifier to choose possible category from "text" Use this category and those sentence with it to predict the polarity,collect them to text's {OTE,polarity}
##SubTask3: similar to subtask2
##Problem to be solved: * How to concat categry and sentence embedding? * How to use the target word(relate to category)'s info in sentence embedding ?(concat :p) * How to use models in specific domain to predict unseen domain