这篇论文来自Yoshua Bengio的实验室，而标题并不起眼「Neural Machine Translation by Jointly Learning to Align and Translate」。 或许很多人都没有听说过这篇论文，但它是NLP中最伟大的里程碑之一，已经被引用了29K次（相比之下，Transformer为77K）。

Transformer in Artificial Intelligence powers over 90% of modern AI models today. Introduced by researchers at Google in 2017, the Transformer architecture changed machine learning forever. It helps ...