PROJECT STATEMENT
Deep Learning structure, consisting of a word embedding layer, a LSTM layer and a classification layer, to perform sentiment classification on movie review domain.
Introduction PyTorch: https://pytorch.org/ GloVe: https://nlp.stanford.edu/projects/glove/ ▪ I strongly recommend, also recommended by pytorch.org, to use Anaconda as the package manager since it installs all dependencies. ▪ You may also need other prerequisite packages, such as numpy, and you can simply use pip install numpy to install them under Anaconda. ▪ The package on e-learning is tested using Windows (OS), Python 3.7 (Language), None (CUDA). ▪ The package uses glove.840B.300d.txt as the pre-trained GloVe word vectors
Package The package contains the following elements: ▪ Dataset o GloVe
- Glove.840B.300d.txt : Not included, you need to download it from internet, unzip it and add it to the directory. o Stsa
- Label.dev -> validation labels
- Label.test -> testing labels
- Label.train -> training labels
- S1.dev -> validation data
- S1.test -> testing data
- S1.train -> training data
▪ Savedir
o Model.pickle -> save the best model on validation data
▪ Data.py -> data processing
▪ Models.py -> main structure of the model
▪ Multis.py -> Utils class
▪ Train_nli.py -> training process of the model