AP19-OLR Challenge: Three Tasks and Their Baselines

Tang, Zhiyuan; Wang, Dong; Song, Liming

doi:10.1109/apsipaasc47483.2019.9023321

Cited by 19 publications

(22 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…6], with parameters CMiss = CFA = 1 and PTarget = 0.5, on the predicted language scores. Results From Table 4, we compare the Cavg results of models 1, 2, and 3 to the respective baseline results of AP19-OLR (0.126) [12], MGB-3 (0.218) [27], and DoSL (0.013) [19], and note that our results are within 0.1, 1.8, and 2.4 percentage points. We note that model 1 outperforms other models on AP19-OLR and DoSL, while model 2 is best on its reference dataset MGB-3 and the best overall model on average.…”

Section: End-to-end Experimentsmentioning

confidence: 93%

“…We did not have access to the NIST LRE datasets. AP19-OLR Oriental Language Recognition challenge 2019 (AP19-OLR), contains speech in 10 languages mainly spoken in Asia and one out-of-set (OOS) mixture of European languages [12]. The dataset includes 261 hours of training data and 5 hours of test data.…”

Section: Datasetsmentioning

confidence: 99%

“…The choice of baselines for MGB-3 and DoSL were motivated by the success previously reported for these two datasets. The baseline architecture for AP19-OLR was defined as the competition baseline by [12] and we choose the same baseline. We enumerate the model architectures used in this paper as follows:…”

Section: End-to-end Experimentsmentioning

confidence: 99%

“…tional neural network (CNN) feature extractor front-ends, attention mechanisms and long short-term memory (LSTM) layers [6], as well as a larger time-delayed deep neural network (TDNN) structure with residual, skip connections [11]. In the fourth oriental language recognition (AP19-OLR) challenge, an x-vector based SLI model was given as the baseline [12]. In addition to supporting end-to-end SLI, the x-vector architecture can also be used to learn a fixed-length language embedding representation for variable length utterances [4].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Releasing a Toolkit and Comparing the Performance of Language Embeddings Across Various Spoken Language Identification Datasets

Lindgren¹,

Jauhiainen²,

Kurimo³

2020

Interspeech 2020

View full text Add to dashboard Cite

In this paper, we propose a software toolkit for easier end-toend training of deep learning based spoken language identification models across several speech datasets. We apply our toolkit to implement three baseline models, one speaker recognition model, and three x-vector architecture variations, which are trained on three datasets previously used in spoken language identification experiments. All models are trained separately on each dataset (closed task) and on a combination of all datasets (open task), after which we compare if the open task training yields better language embeddings. We begin by training all models end-to-end as discriminative classifiers of spectral features, labeled by language. Then, we extract language embedding vectors from the trained end-to-end models, train separate Gaussian Naive Bayes classifiers on the vectors, and compare which model provides best language embeddings for the backend classifier. Our experiments show that the open task condition leads to improved language identification performance on only one of the datasets. In addition, we discovered that increasing x-vector model robustness with random frequency channel dropout significantly reduces its end-to-end classification performance on the test set, while not affecting back-end classification performance of its embeddings. Finally, we note that two baseline models consistently outperformed all other models.

show abstract

Section: End-to-end Experimentsmentioning

confidence: 93%

Section: Datasetsmentioning

confidence: 99%

Section: End-to-end Experimentsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Releasing a Toolkit and Comparing the Performance of Language Embeddings Across Various Spoken Language Identification Datasets

Lindgren¹,

Jauhiainen²,

Kurimo³

2020

Interspeech 2020

View full text Add to dashboard Cite

show abstract

“…EXPERIMENTS3.1. Experimental conditionsExperiments on SLID task are carried out based on data corpus from Oriental Language Recognition (OLR) 2020 Challenge[20,21, 22]. In training, there are around 110 k utterances ( more than 100 hours), from 10 languages.…”

mentioning

confidence: 99%

Unsupervised Neural Adaptation Model Based on Optimal Transport for Spoken Language Identification

Shen

Tsao

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Due to the mismatch of statistical distributions of acoustic speech between training and testing sets, the performance of spoken language identification (SLID) could be drastically degraded. In this paper, we propose an unsupervised neural adaptation model to deal with the distribution mismatch problem for SLID. In our model, we explicitly formulate the adaptation as to reduce the distribution discrepancy on both feature and classifier for training and testing data sets. Moreover, inspired by the strong power of the optimal transport (OT) to measure distribution discrepancy, a Wasserstein distance metric is designed in the adaptation loss. By minimizing the classification loss on the training data set with the adaptation loss on both training and testing data sets, the statistical distribution difference between training and testing domains is reduced. We carried out SLID experiments on the oriental language recognition (OLR) challenge data corpus where the training and testing data sets were collected from different conditions. Our results showed that significant improvements were achieved on the cross domain test tasks.

show abstract

Development of accurate automated language identification model using polymer pattern and tent maximum absolute pooling techniques

Tuncer

Doğan

Akbal

et al. 2022

Neural Comput & Applic

View full text Add to dashboard Cite

AP19-OLR Challenge: Three Tasks and Their Baselines

Cited by 19 publications

References 13 publications

Releasing a Toolkit and Comparing the Performance of Language Embeddings Across Various Spoken Language Identification Datasets

Releasing a Toolkit and Comparing the Performance of Language Embeddings Across Various Spoken Language Identification Datasets

Unsupervised Neural Adaptation Model Based on Optimal Transport for Spoken Language Identification

Development of accurate automated language identification model using polymer pattern and tent maximum absolute pooling techniques

Contact Info

Product

Resources

About