2019/06/04 by Fengshun Xiao, Xiao, Fengshun, Jiangtong Li +7
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #cs.CL
paper · pdf · doi:10.48550/arxiv.1906.01282
Accepted by ACL 2019
arxiv created 2019/06/04 · arxiv updated 2019/06/05
Neural machine translation (NMT) takes deterministic sequences for source representations. However, either word-level or subword-level segmentations have multiple choices to split a source sequence with different word segmentors or different subword vocabulary sizes. We hypothesize that the diversity in segmentations may affect the NMT performance. To integrate different segmentations with the state-of-the-art NMT model, Transformer, we propose lattice-based encoders to explore effective word or subword representation in an automatic way during training. We propose two methods: 1) lattice positional encoding and 2) lattice-aware self-attention. These two methods can be used together and show complementary to each other to further improve translation performance. Experiment results show superiorities of lattice-based encoders in word-level and subword-level representations over conventional Transformer encoder.