Bidirectional recurrent neural networks
Двунаправленные рекуррентные нейронные сети
1997-01-01
SCID: 54.1/spbuzmsd
Discuss with AI
TIMIT phoneme classificationbidirectional recurrent neural networkrecurrent neural networksequence posterior probability estimationtraining in positive and negative time direction
Figures from the paper
Abstract (AI)
In the first part of this paper, a regular recurrent neural network (RNN) is extended to a bidirectional recurrent neural network (BRNN). The BRNN can be trained without the limitation of using input information just up to a preset future frame. This is accomplished by training it simultaneously in positive and negative time direction. Structure and training procedure of the proposed network are explained. In regression and classification experiments on artificial data, the proposed structure gives better results than other approaches. For real data, classification experiments for phonemes from the TIMIT database show the same tendency. In the second part of this paper, it is shown how the proposed bidirectional structure can be easily modified to allow efficient estimation of the conditional posterior probability of complete symbol sequences without making any explicit assumption about the shape of the distribution. For this part, experiments on real data are reported.
Key Findings
1
A regular RNN is extended to a bidirectional recurrent neural network (BRNN) trained simultaneously in positive and negative time directions.
2
Classification experiments on real phoneme data (TIMIT) show the BRNN outperforms alternative methods with the same tendency as artificial-data results.
3
Experiments on real data demonstrate the modified BRNN's ability to estimate sequence-level conditional posteriors effectively.
4
In regression and classification experiments on artificial data, the BRNN yields better results than other approaches.
5
The BRNN removes the limitation of using input information only up to a preset future frame by incorporating information from both past and future.
6
The bidirectional structure can be modified to efficiently estimate conditional posterior probabilities of complete symbol sequences without explicit distributional assumptions.
Research Object
Bidirectional recurrent neural network (BRNN)
Research Subject
Structure, training procedure, and performance (regression/classification and conditional posterior sequence estimation) of BRNNs compared to other approaches
Publication Details
Publication Date
1997-01-01
Journal
Publisher
ISSN
Cited by
10095
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai2
Cited by20
Neural Machine Translation by Jointly Learning to Align and Translate2014
Speech recognition with deep recurrent neural networks2013
A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures2019
Object Detection With Deep Learning: A Review2019
CatBoost for big data: an interdisciplinary review2020
Deep Learning Enabled Semantic Communication Systems2021
Deep learning in bioinformatics2016
Machine Learning and Deep Learning Methods for Intrusion Detection Systems: A Survey2019
Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems2015
FinBERT : A Large Language Model for Extracting Information from Financial Text*2022
A Comprehensive Study on Cross-View Gait Based Human Identification with Deep CNNs2016
Improving sentiment analysis via sentence type classification using BiLSTM-CRF and CNN2016
Attention in Natural Language Processing2020
Modeling Coverage for Neural Machine Translation2016
Semantic Communications for Future Internet: Fundamentals, Applications, and Challenges2022
Explainable Artificial Intelligence Applications in Cyber Security: State-of-the-Art in Research2022
A Comprehensive Review of Deep Learning: Architectures, Recent Advances, and Applications2024
Graph Neural Networks for Natural Language Processing: A Survey2023
An Analysis on Financial Statement Fraud Detection for Chinese Listed Companies Using Deep Learning2022
RenAIssance: A Survey Into AI Text-to-Image Generation in the Era of Large Model2024