Before the deep learning revolution, natural language processing (NLP) systems were almost fully rule based. Linguists created intricate parsing rules and tried to define our language's grammar to automate tasks such as part of speech tagging or named entity recognition. Human-level translation between different languages and free-form question answering were in the domain of science fiction. NLP systems were hard to maintain and took a long time to develop.
As with computer vision, deep learning took the NLP world by storm. Deep-learning-based NLP algorithms successfully perform near-human-level translation between different languages, can measure the emotional sentiment of a text, can learn to ...
