We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Duplex Sequence-to-Sequence Learning for Reversible Machine Translation

Abstract: Sequence-to-sequence (seq2seq) problems such as machine translation are bidirectional, which naturally derive a pair of directional tasks and two directional learning signals. However, typical seq2seq neural networks are {\em simplex} that only model one unidirectional task, which cannot fully exploit the potential of bidirectional learning signals from parallel data. To address this issue, we propose a {\em duplex} seq2seq neural network, REDER (Reversible Duplex Transformer), and apply it to machine translation. The architecture of REDER has two ends, each of which specializes in a language so as to read and yield sequences in that language. As a result, REDER can simultaneously learn from the bidirectional signals, and enables {\em reversible machine translation} by simply flipping the input and output ends, Experiments on widely-used machine translation benchmarks verify that REDER achieves the first success of reversible machine translation, which helps obtain considerable gains over several strong baselines.
Comments: Under review, 10 pages
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2105.03458 [cs.CL]
  (or arXiv:2105.03458v1 [cs.CL] for this version)

Submission history

From: Zaixiang Zheng [view email]
[v1] Fri, 7 May 2021 18:21:57 GMT (7630kb,D)

Link back to: arXiv, form interface, contact.