Efficient Pronunciation Assessment of Taiwanese-Accented English Based on Unsupervised Model Adaptation and Dynamic Sentence Selection

2012 ◽  
pp. 12-30
Author(s):  
Chung-Hsien Wu ◽  
Hung-Yu Su ◽  
Chao-Hong Liu
Keyword(s):  
Mutual Information ◽  
English Teachers ◽  
Experimental Results ◽  
Model Adaptation ◽  
Selection Algorithm ◽  
Accuracy Improvement ◽  
Acoustic Models ◽  
Efficient Approach ◽  
Accented English ◽  
Adaptation Method

This chapter presents an efficient approach to personalized pronunciation assessment of Taiwanese-accented English. The main goal of this study is to detect frequently occurring mispronunciation patterns of Taiwanese-accented English instead of scoring English pronunciations directly. The proposed assessment help quickly discover personalized mispronunciations of a student, thus English teachers can spend more time on teaching or rectifying students’ pronunciations. In this approach, an unsupervised model adaptation method is performed on the universal acoustic models to recognize the speech of a specific speaker with mispronunciations and Taiwanese accent. A dynamic sentence selection algorithm, considering the mutual information of the related mispronunciations, is proposed to choose a sentence containing the most undetected mispronunciations in order to quickly extract personalized mispronunciations. The experimental results show that the proposed unsupervised adaptation approach obtains an accuracy improvement of about 2.1% on the recognition of Taiwanese-accented English speech.

scholarly journals Adversarial Disentanglement with Grouped Observations

2020 ◽  
Vol 34 (06) ◽  
pp. 10243-10250
Author(s):  
Jozsef Nemeth
Keyword(s):  
Mutual Information ◽  
Experimental Results ◽  
Unseen Data ◽  
Data Content ◽  
Training Objective ◽  
Content Information ◽  
Image Datasets

We consider the disentanglement of the representations of the relevant attributes of the data (content) from all other factors of variations (style) using Variational Autoencoders. Some recent works addressed this problem by utilizing grouped observations, where the content attributes are assumed to be common within each group, while there is no any supervised information on the style factors. In many cases, however, these methods fail to prevent the models from using the style variables to encode content related features as well. This work supplements these algorithms with a method that eliminates the content information in the style representations. For that purpose the training objective is augmented to minimize an appropriately defined mutual information term in an adversarial way. Experimental results and comparisons on image datasets show that the resulting method can efficiently separate the content and style related attributes and generalizes to unseen data.

Time sensitive blog retrieval using temporal properties of queries

2016 ◽  
Vol 43 (1) ◽  
pp. 103-121 ◽  
Author(s):  
MohammadSadegh Zahedi ◽  
Abolfazl Aleahmad ◽  
Maseud Rahgozar ◽  
Farhad Oroumchian ◽  
Arastoo Bozorgi
Keyword(s):  
Experimental Analysis ◽  
Experimental Results ◽  
Temporal Property ◽  
Time Intervals ◽  
Efficient Approach ◽  
Temporal Properties ◽  
Time Sensitivity ◽  
Retrieval Algorithms ◽  
First Time ◽  
The Web

Blogs are one of the main user-generated contents on the web and are growing in number rapidly. The characteristics of blogs require the development of specialized search methods which are tuned for the blogosphere. In this paper, we focus on blog retrieval, which aims at ranking blogs with respect to their recurrent relevance to a user’s topic. Although different blog retrieval algorithms have already been proposed, few of them have considered temporal properties of the input queries. Therefore, we propose an efficient approach to improving relevant blog retrieval using temporal property of queries. First, time sensitivity of each query is automatically computed for different time intervals based on an initially retrieved set of relevant posts. Then a temporal score is calculated for each blog and finally all blogs are ranked based on their temporal and content relevancy with regard to the input query. Experimental analysis and comparison of the proposed method are carried out using a standard dataset with 45 diverse queries. Our experimental results demonstrate that, using different measurement criteria, our proposed method outperforms other blog retrieval methods.

Differenced maximum mutual information criterion for robust unsupervised acoustic model adaptation

2016 ◽  
Vol 36 ◽  
pp. 24-41
Author(s):  
Marc Delcroix ◽  
Atsunori Ogawa ◽  
Seong-Jun Hahm ◽  
Tomohiro Nakatani ◽  
Atsushi Nakamura
Keyword(s):  
Mutual Information ◽  
Information Criterion ◽  
Acoustic Model ◽  
Model Adaptation ◽  
Maximum Mutual Information

scholarly journals ACAT1 Benchmark of RANS-Informed Analytical Methods for Fan Broadband Noise Prediction: Part II—Influence of the Acoustic Models

Acoustics ◽  
2020 ◽  
Vol 2 (3) ◽  
pp. 617-649
Author(s):  
Sébastien Guérin ◽  
Carolin Kissner ◽  
Pascal Seeler ◽  
Ricardo Blázquez ◽  
Pedro Carrasco Laraña ◽  
...  
Keyword(s):  
Analytical Methods ◽  
Acoustic Waves ◽  
Power Spectra ◽  
Low Frequency ◽  
Broadband Noise ◽  
Experimental Results ◽  
Acoustic Analogy ◽  
Noise Sources ◽  
Acoustic Models ◽  
The Impact

A benchmark dedicated to RANS-informed analytical methods for the prediction of turbofan rotor–stator interaction broadband noise was organised within the framework of the European project TurboNoiseBB. The second part of this benchmark focuses on the impact of the acoustic models. Twelve different approaches implemented in seven different acoustic solvers are compared. Some of the methods resort to the acoustic analogy, while some use a direct approach bypassing the calculation of a source term. Due to differing application objectives, the studied methods vary in terms of complexity to represent the turbulence, to calculate the acoustic response of the stator and to model the boundary and flow conditions for the generation and propagation of the acoustic waves. This diversity of approaches constitutes the unique quality of this work. The overall agreement of the predicted sound power spectra is satisfactory. While the comparison between the models show significant deviations at low frequency, the power levels vary within an interval of ±3 dB at mid and high frequencies. The trends predicted by increasing the rotor speed are similar for almost all models. However, most predicted levels are some decibels lower than the experimental results. This comparison is not completely fair—particularly at low frequency—because of the presence of noise sources in the experimental results, which were not considered in the simulations.

scholarly journals Input-Aware Implication Selection Scheme Utilizing ATPG for Efficient Concurrent Error Detection

Electronics ◽  
2018 ◽  
Vol 7 (10) ◽  
pp. 258 ◽  
Author(s):  
Abdus Hassan ◽  
Umar Afzaal ◽  
Tooba Arifeen ◽  
Jeong Lee
Keyword(s):  
Error Detection ◽  
High Probability ◽  
State Of The Art ◽  
The State ◽  
Experimental Results ◽  
Concurrent Error Detection ◽  
Selection Algorithm ◽  
Probability Of Error ◽  
Selection Scheme ◽  
Selection Strategies

Recently, concurrent error detection enabled through invariant relationships between different wires in a circuit has been proposed. Because there are many such implications in a circuit, selection strategies have been developed to select the most valuable implications for inclusion in the checker hardware such that a sufficiently high probability of error detection ( P d e t e c t i o n ) is achieved. These algorithms, however, due to their heuristic nature cannot guarantee a lossless P d e t e c t i o n . In this paper, we develop a new input-aware implication selection algorithm with the help of ATPG which minimizes loss on P d e t e c t i o n . In our algorithm, the detectability of errors for each candidate implication is carefully evaluated using error prone vectors. The evaluation results are then utilized to select the most efficient candidates for achieving optimal P d e t e c t i o n . The experimental results on 15 representative combinatorial benchmark circuits from the MCNC benchmarks suite show that the implications selected from our algorithm achieve better P d e t e c t i o n in comparison to the state of the art. The proposed method also offers better performance, up to 41.10%, in terms of the proposed impact-level metric, which is the ratio of achieved P d e t e c t i o n to the implication count.

Sign in / Sign up

Export Citation Format

Share Document