Does String-Based Neural MT Learn Source Syntax?

@inproceedings{Shi2016DoesSN,
  title={Does String-Based Neural MT Learn Source Syntax?},
  author={Xing Shi and Inkit Padhi and Kevin Knight},
  booktitle={EMNLP},
  year={2016}
}
We investigate whether a neural, encoderdecoder translation system learns syntactic information on the source side as a by-product of training. We propose two methods to detect whether the encoder has learned local and global source syntax. A fine-grained analysis of the syntactic structure learned by the encoder reveals which kinds of syntax are learned and which are missing. 

Figures, Tables, Results, and Topics from this paper.

Key Quantitative Results

  • We achieve 92.8% accuracy (Table 2), far above the majority class baseline (82.8%).

Explore Further: Topics Discussed in This Paper

Citations

Publications citing this paper.
SHOWING 1-10 OF 104 CITATIONS

An Analysis of Encoder Representations in Transformer-Based Machine Translation

  • BlackboxNLP@EMNLP
  • 2018
VIEW 4 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED

Analysis Methods in Neural Language Processing: A Survey

  • Transactions of the Association for Computational Linguistics
  • 2018
VIEW 10 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED

FILTER CITATIONS BY YEAR

2016
2019

CITATION STATISTICS

  • 14 Highly Influenced Citations

  • Averaged 34 Citations per year from 2017 through 2019

References

Publications referenced by this paper.
SHOWING 1-10 OF 28 REFERENCES

An empirical examination of challenges in Chinese parsing Distributed representations of sentences and documents

Jiwei Li, Xinlei Chen, Eduard Hovy, Dan Jurafsky
  • Proc . ICML .
  • 2016
VIEW 1 EXCERPT

Skip-Thought Vectors

  • NIPS
  • 2015
VIEW 1 EXCERPT

Convolutional neural networks for sentence classification

Yoon Kim.
  • Proc. EMNLP.
  • 2014
VIEW 1 EXCERPT