How to implement Sentiment Analysis using word embedding and Convolutional Neural Networks on Keras seems to be a relatively simple model, creating their own embedding.
I would think the embeddings could also be initialised or frozen with pre-existing word embeddings.
Here the inputs are padded word sequences (i.e. concatenated word embeddings at the next level). The convolutions would allow it to learn a local context. An alternative would be to use LSTMs. Although the author found the convolutions to be faster.
How to implement Sentiment Analysis using word embedding and Convolutional Neural Networks on Keras seems to be a relatively simple model, creating their own embedding. I would think the embeddings could also be initialised or frozen with pre-existing word embeddings. Here the inputs are padded word sequences (i.e. concatenated word embeddings at the next level). The convolutions would allow it to learn a local context. An alternative would be to use LSTMs. Although the author found the convolutions to be faster.