Automatic language identification using deep neural networks

By Ignacio Lopez-Moreno, Javier Gonzalez-Dominguez, Oldrich Plchot, David Martinez, Joaquin Gonzalez-Rodriguez, Pedro J. Moreno Mengibar. 2014

265 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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

In Signal Processing

Citations in the corpus

Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, background): “In recent years, with the merit of large labeled datasets, enormous computation capability, and effective network architectures, emerging progress towards end-to-end learning opens up a new area for exploration [19, 20, 21, 22, 23].”
A unified deep neural network for speaker and language recognition (2015, methodology): “The impressive gains in performance obtained using deep neural networks (DNNs) for automatic speech recognition (ASR) [1] have motivated the application of DNNs to other speech technologies such as speaker recognition (SR) and language recognition (LR) [2, 3, 4, 5, 6, 7, 8, 9, 10].”
LID-Senones and Their Statistics for Language Identification (2018, background): “at a frame level [16],[17], these layers process acoustic features frame by frame.”
Utterance-level Aggregation for Speaker Recognition in the Wild (2019, methodology): “Earlier deep neural network (DNN) based speaker recognition systems have naı̈vely used pooling methods that have been successful for visual recognition tasks, such as average pooling [2, 3, 7, 8] or fully connected layers [9, 10] to condense frame-level information into utterance-level representatio”
Knowledge Distillation-Based Representation Learning for Short-Utterance Spoken Language Identification (2020, methodology): “Similar to previous work [13], [14], the RNN and DCNN models achieved better performance than i-vector-based approaches on short-utterance LID tasks, i.”
End-to-End Language Identification Using Attention-Based Recurrent Neural Networks (2016, methodology): “Further, LSTM RNN [13, 14] is adopted to the LID task to model the long-range dependency across the input temporal sequence with respect to feed forward neural networks [12, 15].”
Spoken Language Recognition using X-vectors (2018, background): “In language recognition, a popular approach is to train the neural network to classify languages at the framelevel [7, 8, 9, 10].”
Multilingually trained bottleneck features in spoken language recognition (2017, methodology): “TaggedPSeveral different approaches to allow use of multilingual training corpora in SLR has been proposed. Zissman and Singer (1994) used six phoneme recognizers running in parallel, each producing a language-dependent likelihood based on an N-gram phonotactic model.”
Deep Learning for Audio Signal Processing (2019, background): “speaker recognition [96], language recognition [97] and speech translation [98].”
Spoken Language Identification using ConvNets (2019, background): “Nowadays most of the attempts on spoken language identification rely on neural networks for meaningful feature extraction and classification [15,7].”
Interactive Learning of Teacher-student Model for Short Utterance Spoken Language Identification (2019, background): “Recently, end-to-end approaches with deep neural networks (DNN), recurrent neural networks (RNN), convolutional neural networks (CNN) and attention-based neural networks have been investigated on LID tasks [9, 10, 11, 12, 13].”
Layer-Centric Memory Reuse and Data Migration for Extreme-Scale Deep Learning on Many-Core Architectures (2018, unknown): “Speech Recognition 4-layer LSTM-RNN [34] (sequence length=800) 2975-hour Google 5M LID Corpus [28]) 148MB/49.”
Feature Representation of Short Utterances Based on Knowledge Distillation for Spoken Language Identification (2018, background): “Recently, endto-end approaches with convolutional neural networks (CNN), recurrent neural networks (RNN), and attention-based neural networks have been investigated on LID tasks[11, 12, 13, 14].”
A Semisupervised Approach for Language Identification based on Ladder Networks (2016, methodology): “DNNs have also been applied to directly train language classification systems [4][5].”
A Novel Learnable Dictionary Encoding Layer for End-to-End Language Identification (2018, background): “In order to feed into the network, as is done in [13, 14, 15, 16], the original input feature sequence has to be resized or cropped into multiple small fixed-size segments in frame level.”
Deep Neural Network Approaches to Speaker and Language Recognition (2015, methodology): “While there is a substantial degradation in performance relative to the baseline system at the 30 s and 10 s durations using this technique, there is a slight gain at the short 3 s duration for the 2 layer DNN which is consistent with results reported in [4].”
Automatic language identification using long short-term memory recurrent neural networks (2014, unknown): “In [10] we found Deep feeed forward Neural Networks (DNNs) to surpass i-vector based approaches when dealing with very short test utterances (≤3s) and large amount of training material is available (≥20h per language).”