Lattice decoding and rescoring with long-Span neural network language models
By Martin Sundermeyer, Zolt'an Tuske, Ralf Schlüter, Hermann Ney. 2014
50 citations recorded by the source corpus.
Subfields: Artificial Intelligence, Natural Language Processing, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Ralf Schlüter
- Hermann Ney
- Albert Zeyer
- Kazuki Irie
- Jinyu Li
- Shinji Watanabe
- Sanjeev Khudanpur
- Daniel Povey
In Natural Language Processing
- Ralf Schlüter
- Hermann Ney
- Albert Zeyer
- Kazuki Irie
- Markus Kitza
- Eugen Beck
- Wilfried Michel
- Martin Sundermeyer
In Signal Processing
- Ralf Schlüter
- Hermann Ney
- Daniel Povey
- Sanjeev Khudanpur
- Albert Zeyer
- Kazuki Irie
- Shinji Watanabe
- Jinyu Li
Citations in the corpus
- RWTH ASR Systems for LibriSpeech: Hybrid vs Attention - w/o Data Augmentation (2019, methodology): “We use push-forward algorithm [26] with recombination pruning of order 9.”
- Language Modeling with Deep Transformers (2019, methodology): “The standard push-forward lattice rescoring algorithm [20] for longspan language models can be directly applied to self-attention based models.”
- A Pruned Rnnlm Lattice-Rescoring Algorithm for Automatic Speech Recognition (2018, methodology): “In [5], two methods are proposed to merge lattice-states in order to speedup latticerescoring and reduce sizes of the generated lattices; In [6], a refined pruning technique is used in order to reduce the search space when rescoring lattices with long short-term memory (LSTM) neural network language”
- The RWTH/UPB/FORTH System Combination for the 4th CHiME Challenge Evaluation (2016, methodology): “For the RASR back end (R) we carried out lattice rescoring [20] with an LSTM-RNN language model [21, 22] as follows.”
- LSTM Language Models for LVCSR in First-Pass Decoding and Lattice-Rescoring (2019, methodology): “For our rescoring experiments we use push-forward rescoring as described in [17].”
- From Feedforward to Recurrent LSTM Neural Networks for Language Modeling (2015, methodology): “However, as noted in [17] and [25], it is important to obtain a fully rescored lattice incorporating neural network probabilities, because this allows advanced methods such as confusion network rescoring, which we will take into account in our comparative study.”