• MALSORI
• /
• no.56
• /
• pp.207-223
• /
• 2005

The goal of language model adaptation is to improve the background language model with a relatively small adaptation corpus. This study presents a language model adaptation technique where additional text data for the adaptation do not exist. We propose the information retrieval (IR) technique with N-gram language modeling to collect the adaptation corpus from baseline text data. We also propose to use a dynamic language model interpolation coefficient to combine the background language model and the adapted language model. The interpolation coefficient is estimated from the word hypotheses obtained by segmenting the input speech data reserved for held-out validation data. This allows the final adapted model to improve the performance of the background model consistently The proposed approach reduces the word error rate by $13.6\%$ relative to baseline 4-gram for two-hour broadcast news speech recognition.

• ETRI Journal
• /
• v.33 no.5
• /
• pp.712-719
• /
• 2011

In this paper, we study the problem of domain adaptation for structural support vector machines (SVMs). We consider a number of domain adaptation approaches for structural SVMs and evaluate them on named entity recognition, part-of-speech tagging, and sentiment classification problems. Finally, we show that a prior model for structural SVMs outperforms other domain adaptation approaches in most cases. Moreover, the training time for this prior model is reduced compared to other domain adaptation methods with improvements in performance.

• MALSORI
• /
• v.68
• /
• pp.33-47
• /
• 2008

In this paper, we propose a hybrid acoustic and pronunciation model adaptation method based on context dependency for Korean-English speech recognition. The proposed method is performed as follows. First, in order to derive pronunciation variant rules, an n-best phoneme sequence is obtained by phone recognition. Second, we decompose each rule into a context independent (CI) or a context dependent (CD) one. To this end, it is assumed that a different phoneme structure between Korean and English makes CI pronunciation variabilities while coarticulation effects are related to CD pronunciation variabilities. Finally, we perform an acoustic model adaptation and a pronunciation model adaptation for CI and CD pronunciation variabilities, respectively. It is shown from the Korean-English speech recognition experiments that the average word error rate (WER) is decreased by 36.0% when compared to the baseline that does not include any adaptation. In addition, the proposed method has a lower average WER than either the acoustic model adaptation or the pronunciation model adaptation.

• Phonetics and Speech Sciences
• /
• v.7 no.3
• /
• pp.131-138
• /
• 2015

In this paper, we propose L1-norm regularization for state vector adaptation of subspace Gaussian mixture model (SGMM). When you design a speaker adaptation system with GMM-HMM acoustic model, MAP is the most typical technique to be considered. However, in MAP adaptation procedure, large number of parameters should be updated simultaneously. We can adopt sparse adaptation such as L1-norm regularization or sparse MAP to cope with that, but the performance of sparse adaptation is not good as MAP adaptation. However, SGMM does not suffer a lot from sparse adaptation as GMM-HMM because each Gaussian mean vector in SGMM is defined as a weighted sum of basis vectors, which is much robust to the fluctuation of parameters. Since there are only a few adaptation techniques appropriate for SGMM, our proposed method could be powerful especially when the number of adaptation data is limited. Experimental results show that error reduction rate of the proposed method is better than the result of MAP adaptation of SGMM, even with small adaptation data.

• Journal of Korean Society for Atmospheric Environment
• /
• v.30 no.3
• /
• pp.270-280
• /
• 2014

The objective of this research is to quantitatively valuate the economic value of analysis model related to climate change mitigation and adaptation. Due to the fact that the subject of this research, which is the Korean climate change mitigation and adaptation model, has not been actualized, a conjoint analysis applying stated preference data has utilized. As results, among the many attributes considered in this research, the value of the attribute related to reflecting Korea's current situation is analyzed to be largest in both greenhouse gas (GHG) mitigation model and climate change adaptation model. Additionally, if all the considered functional aspects are assumed to be feasible, the economic value of the Korean GHG mitigation model is assumed to be 60.3 billion Korean won(KRW) and the Korean climate change adaptation model is assumed to be 51 billion KRW.

• Korean Journal of Social Welfare
• /
• v.41
• /
• pp.173-200
• /
• 2000

The purpose of this study is to examine the variables that may influence the level of family adaptation to schizophrenic patients using the Family Resiliency Model. The Family Resiliency Model is the most current extension of family stress Model. According to the Family Resiliency Model, The level of family adaptation in the face of a crisis situation is determined by a number of interacting components. The subjects are 151 family members with schizophrenic patient. The result from the research were as follows: 1) The following variables significantly correlated with the family adaptation: income of the family, educational level of the family, intimacy between family and patient, knowledge on schizophrenia, recognition of prognosis on schizophrenia. 2) The factors that compose the Family Resiliency Model significantly correlated with the level of family adaptation. 3) The result of stepwise multiple regression analysis indicated that factors which predict the level of family adaptation were the family control, the quality of family communication, and the support from the extended family, these findings give us significant practical implications for social work intervention.

• Phonetics and Speech Sciences
• /
• v.4 no.3
• /
• pp.111-117
• /
• 2012

In this paper, we compare the performance of several speaker adaptation methods for a HMM-based Korean speech synthesis system with small amounts of adaptation data. According to objective and subjective evaluations, a hybrid method of constrained structural maximum a posteriori linear regression (CSMAPLR) and maximum a posteriori (MAP) adaptation shows better performance than other methods, when only five minutes of adaptation data are available for the target speaker. During the objective evaluation, we find that the duration models are insufficiently adapted to the target speaker as the spectral envelope and pitch models. To alleviate the problem, we propose the duration rectification method and the duration interpolation method. Both the objective and subjective evaluations reveal that the incorporation of the proposed two methods into the conventional speaker adaptation method is effective in improving the performance of the duration model adaptation.

• Journal of the Korean Home Economics Association
• /
• v.37 no.10
• /
• pp.107-122
• /
• 1999

Based on the theoretical framework of the Double ABCX model of family stress and adaptation, this study was to analysis the causal relationships of stressors (psychological sexual, and physical wife abuse), personal resources (self-efficacy and self-esteem) and social support(emotional and informational support), appraisal(positive appraisal and downward comparisons) with adaptation (psychological well-being and somatic symptoms). Self-administered questionnaire method was used to collect data from 264 wives residing in Chonbuk-do and Kyonggi-do area. The causal model was tested and modified by the maximum likelihood method using UISREL 7 program. Results showed that wife abuse had effect on adaptation indirectly through the latent variables of personal resource and appraisal, which influenced the adaptation directly. In addition, social support indirectly affect the adaptation through personal resource and appraisal. It was also found that all these variables explained 27.6% of the total variance of wives'adaptation, and that personal resources was the most powerful variable in predicting the adaptation of the wives.

• Journal of Power Electronics
• /
• v.15 no.5
• /
• pp.1274-1285
• /
• 2015

This paper deals with model reference adaptive system (MRAS) speed estimators based on a secondary flux for linear induction motors (LIMs). The operation of these estimators significantly depends on an adaptation mechanism. Fixed-gain PI controller is the most common adaptation mechanism that may fail to estimate the speed correctly in different conditions, such as variation in machine parameters and noisy environment. Two adaptation mechanisms are proposed to improve LIM drive system performance, particularly at very low speed. The first adaptation mechanism is based on fuzzy theory, and the second is obtained from an LIM mechanical model. Compared with a conventional PI controller, the proposed adaptation mechanisms have low sensitivity to both variations of machine parameters and noise. The optimum parameters of adaptation mechanisms are tuned using an offline method through chaotic optimization algorithm (COA) because no design criterion is given to provide these values. The efficiency of MRAS speed estimator is validated by both numerical simulation and real-time hardware-in-the-loop (HIL) implementations. Results indicate that the proposed adaptation mechanisms improve performance of MRAS speed estimator.

• ETRI Journal
• /
• v.38 no.3
• /
• pp.487-493
• /
• 2016

Two new methods are proposed for an unsupervised adaptation of a language model (LM) with a single sentence for automatic transcription tasks. At the training phase, training documents are clustered by a method known as Latent Dirichlet allocation (LDA), and then a domain-specific LM is trained for each cluster. At the test phase, an adapted LM is presented as a linear mixture of the now trained domain-specific LMs. Unlike previous adaptation methods, the proposed methods fully utilize a trained LDA model for the estimation of weight values, which are then to be assigned to the now trained domain-specific LMs; therefore, the clustering and weight-estimation algorithms of the trained LDA model are reliable. For the continuous speech recognition benchmark tests, the proposed methods outperform other unsupervised LM adaptation methods based on latent semantic analysis, non-negative matrix factorization, and LDA with n-gram counting.