← ML Research Wiki / 1409.0473

Published as a conference paper at ICLR 2015 NEURAL MACHINE TRANSLATION BY JOINTLY LEARNING TO ALIGN AND TRANSLATE

Dzmitry Bahdanau, Kyunghyun Cho, Yoshua Bengio, Jacobs University BremenGermany, Université de Montréal (2014)

Paper Information
arXiv ID
Venue
International Conference on Learning Representations
Domain
natural language processing
SOTA Claim
Yes

Abstract

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and encode a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

Summary

This paper proposes a novel approach to neural machine translation (NMT) called RNNsearch, which improves upon the encoder-decoder architecture by incorporating a mechanism that allows the model to (soft-)search for parts of the source sentence relevant to each target word being translated. This addresses the bottleneck of using a fixed-length vector in traditional NMT systems. The authors show that this new method can significantly improve translation quality, especially for longer sentences, achieving performance comparable to state-of-the-art phrase-based systems on English-to-French translation. Alignments generated by the model are shown to be intuitive and reliable, allowing for better handling of long input sentences without losing important contextual information. Qualitative evaluations illustrate the model's ability to maintain grammatical correctness and semantic integrity in translations.

Methods

This paper employs the following methods:

  • Encoder-Decoder
  • RNN
  • Bidirectional RNN

Models Used

  • RNN Encoder-Decoder
  • RNNsearch

Datasets

The following datasets were used in this research:

  • WMT '14

Evaluation Metrics

  • BLEU

Results

  • Achieved translation performance comparable to existing state-of-the-art phrase-based systems on English-to-French translation
  • Proposed model more robust to long sentences compared to traditional methods

Limitations

The authors identified the following limitations:

  • Not specified

Technical Requirements

  • Number of GPUs: None specified
  • GPU Type: None specified

Keywords

neural machine translation encoder-decoder attention mechanism sequence-to-sequence alignment

Papers Using Similar Methods

External Resources