AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

A Dependency-Based Neural Reordering Model for Statistical Machine Translation

Christian Hadiwinoto, Hwee Tou Ng

AAAI 2017 · Architectures and efficiency › RNNs and normalization · approach: Non-autoregressive translation · 8 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace sparse dependency swap features with a 10-system neural reordering ensemble initialized with dependency-driven embeddingsPhrase-based Chinese-to-English SMT · NIST averageBLEU: +0.57 BLEU pointImproved
Replace sparse dependency swap features with a single neural classifier initialized with dependency-driven embeddingsPhrase-based Chinese-to-English SMT · NIST averageBLEU: +0.32 BLEU pointImproved
Add dependency-driven word embedding initialization (ablation)Neural reordering classifier · NIST averageBLEU: 39.06 -> 39.55Improved
Replace dependency-driven embeddings with random initialization (ablation)Neural reordering classifier · NIST averageBLEU: 39.55 -> 38.12Hurt
Replace dependency-driven embeddings with skip-gram embeddings (ablation)Neural reordering classifier · NIST averageBLEU: 39.55 -> 39.06No clear effect
Use a 10-system ensemble instead of a single neural classifier (ablation)Dependency-driven neural reordering · NIST averageBLEU: 39.30 -> 39.55Improved
Add neural reordering without conventional phrase-based reordering models (ablation)Phrase-based Chinese-to-English SMT · NIST averageBLEU: +1.05 BLEU pointImproved
Use sparse dependency distortion and swap features without conventional phrase-based reordering models (ablation)Phrase-based Chinese-to-English SMT · NIST averageBLEU: 37.74 vs 37.89Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy