Title: Introduction to the second issue on machine translation for low-resource languages
Abstract: In this second issue of the special topic on machine translation (MT) for lowresource languages, we include four papers that introduce knowledge in addition to those automatically learned from parallel corpora.The paper by Armengol-Estapé and Costa-Jussà is a survey on the use of semantic and linguistic information in neural MT (NMT).It reviews methods making use of these features in NMT models and explores the use of linked data from BabelNet and Babelfy.Preliminary experiments show improved MT performance on both high (German-to-English) and low (English-to-Nepali) resource language pairs.It would be interesting to see more research efforts in this direction for other languages.