Investigating ‘Aspect’ in NMT and SMT: Translating the English Simple Past and Present Perfect

By Eva Vanmassenhove, Jinhua Du, Andy Way. 2017

22 citations recorded by the source corpus.

Subfields: Machine Translation, Computational Linguistics & Semantics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Machine Translation

1. Yonatan Belinkov
2. James R. Glass
3. Jesse Vig
4. Hassan Sajjad
5. Fahim Dalvi
6. Nadir Durrani

In Computational Linguistics & Semantics

1. Yonatan Belinkov
2. R. Thomas McCoy
3. James R. Glass
4. Paul Smolensky
5. Tal Linzen
6. Paul Soulos

Citations in the corpus

• Probing Classifiers: Promises, Shortcomings, and Advances (2021, background): “Indeed, several studies pointed out this limitation (Belinkov and Glass 2019), including reports on a mismatch between performance of the probe, P ERF ( g, f, D O , D P ) , and performance of the original model, P ERF ( f, D O ) ( Vanmassenhove, Du, and Way 2017).”
• Analysis Methods in Neural Language Processing: A Survey (2018, background): “For example, Vanmassenhove et al. (2017)”
• Probing Classifiers: Promises, Shortcomings, and Alternatives (2021, background): “Indeed, several studies pointed out this limitation (Belinkov and Glass 2019), including reports on a mismatch between performance of the probe, P ERF ( g, f, D O , D P ) , and performance of the original model, P ERF ( f, D O ) (Vanmassenhove, Du, and Way 2017).”
• Discovering the Compositional Structure of Vector Representations with Role Learning Networks (2019, background): “s will fail despite having the necessary information to succeed in their representations, showing that the ability of a classifier to extract that information does not mean that the model is using it (Vanmassenhove et al., 2017). DISCOVER bridges the gap between representation and behavior: It reveal”
• Edinburgh Research Explorer Understanding Learning Dynamics Of Language Models with SVCCA (unknown): “resented without being used, as shown by Vanmassenhove et al. (2017), who found that NMT systems often predicted tense according to a diagnostic classifier but did not produce the correct tense as output.”