Sunday, 18 October 2026 · Weekend classic from 2014 · 6 min read
Deep LSTM networks map reversed input sequences to fixed vectors for strong machine translation
This paper introduces a sequence-to-sequence model using two deep LSTMs—one encoding reversed source sentences into fixed vectors, the other decoding target sentences—achieving a BLEU score of 34.8 on WMT'14 English-French, outperforming phrase-based SMT baselines.
Based on the paper Sequence to Sequence Learning with Neural Networks by Ilya Sutskever, Oriol Vinyals, Quoc V. Le
Challenges of mapping variable-length sequences in neural networks
Deep Neural Networks (DNNs) excel in many tasks but traditionally require fixed-size input and output vectors, limiting their use in sequence-to-sequence problems like machine translation where input and output lengths vary.
Recurrent Neural Networks (RNNs) can process sequences but struggle when input and output sequences differ in length and have complex alignments, especially due to long-term dependencies.
Long Short-Term Memory (LSTM) networks address long-term dependency issues, making them promising for sequence-to-sequence tasks by encoding input sequences into fixed-size vectors and decoding output sequences from these vectors.
Sign in to read the whole paper
You have read the opening. The rest is free with a Bezen account, and no card is needed.
In the rest of this paper
- How the method works, step by step, with a flow chart
- The key formula from the paper, explained symbol by symbol
- Two figures from the paper, and how to read them
- How it was tested, what was found, and where it falls short