Projects

Background Traditional convolutional neural networks (CNNs) have been hugely successful on a number of problems. In the case of visual processing, they work by sliding small filters (or kernels) across each pixel and its neighbours, and producing a single value. The weights of these filters are trainable, which allows the Read more…

Github repository: github.com/JHart96/keras_elmo_embedding_layer. ELMo embeddings Pre-trained word embeddings have been highly successful in numerous natural language processing tasks.  Traditional word embeddings are essentially lookup tables mapping a word to its corresponding feature vector. As a result, the context of a word is not taken into account when generating its embedding. Allen Read more…