• The Journal of the Acoustical Society of Korea
• /
• v.18 no.3E
• /
• pp.21-25
• /
• 1999

This paper proposes methods that replace a missing triphone with a new one selected or created by existing triphones, and compares the results. The recognition system uses DMS (Dynamic Multisection) model for acoustic modeling. DMS is one of the statistical recognition techniques proper to a small - or mid - size vocabulary system, while HMM (Hidden Markov Model) is a probabilistic technique suitable for a middle or large system. Accordingly, it is reasonable to use an effective algorithm that is proper to DMS, rather than using a complicated method like a polyphone clustering technique employed in HMM-based systems. In this paper, four methods of filling missing triphones are presented. The result shows that a proposed replacing algorithm works almost as well as if all the necessary triphones existed. The experiments are performed on the 500＋ word DMS speech recognizer.

• MALSORI
• /
• no.43
• /
• pp.151-161
• /
• 2002

This study presents performance analysis results of speaker adaptation for keyword spotting system. In this paper, we implemented MLLR (Maximum Likelihood Linear Regression) method on our middle size vocabulary keyword spotting system. This system was developed for directory services of universities and colleges. The experimental results show that speaker adaptation reduces the false alarm rate to 1/3 with the preservation of the mis-detection ratio. This improvement is achieved when speaker adaptation is applied to not only keyword models but also non-keyword models.

• MALSORI
• /
• no.43
• /
• pp.89-99
• /
• 2002

Recently, researches into ways of improving large vocabulary continuous speech recognition and speech synthesis are being carried out intensively as the field of speech information technology is progressing rapidly. In the field of speech recognition, developments of stochastic methods such as HMM require large amount of speech data for training, and also in the field of speech synthesis, recent practices show that synthesis of better quality can be produced by selecting and connecting only the variable size of speech data from the large amount of speech data. In this paper we design and discuss linguistic contents for speech copora for speech recognition and synthesis to be shared in common.

• Journal of Computing Science and Engineering
• /
• v.6 no.2
• /
• pp.151-160
• /
• 2012

We present a methodology that automatically selects indexing algorithms for each heading in Medical Subject Headings (MeSH), National Library of Medicine's vocabulary for indexing MEDLINE. While manually comparing indexing methods is manageable with a limited number of MeSH headings, a large number of them make automation of this selection desirable. Results show that this process can be automated, based on previously indexed MEDLINE citations. We find that AdaBoostM1 is better suited to index a group of MeSH hedings named Check Tags, and helps improve the micro F-measure from 0.5385 to 0.7157, and the macro F-measure from 0.4123 to 0.5387 (both p < 0.01).

• Proceedings of the IEEK Conference
• /
• 2002.07a
• /
• pp.98-101
• /
• 2002

This paper reports a parallel architecture in a HMM based speech recognition system for a low-power LSI design. The proposed architecture calculates output probability of continuous HMM (CHMM) by using concurrent and pipeline processing. They enable to reduce memory access and have high computing efficiency. The novel point is the efficient use of register arrays that reduce memory access considerably compared with any conventional method. The implemented system can achieve a real time response with lower clock in a middle size vocabulary recognition task (100-1000 words) by using this technique.

• ETRI Journal
• /
• v.32 no.5
• /
• pp.795-800
• /
• 2010

Embedding a large vocabulary speech recognition system in mobile devices requires a reduced acoustic model obtained by eliminating redundant model parameters. In conventional optimization methods based on the minimum description length (MDL) criterion, a binary Gaussian tree is built at each state of a hidden Markov model by iteratively finding and merging similar mixture components. An optimal subset of the tree nodes is then selected to generate a downsized acoustic model. To obtain a better binary Gaussian tree by improving the process of finding the most similar Gaussian components, this paper proposes a new distance measure that exploits the difference in likelihood values for cases before and after two components are combined. The mixture weight of Gaussian components is also introduced in the component merging step. Experimental results show that the proposed method outperforms MDL-based optimization using either a Kullback-Leibler (KL) divergence or weighted KL divergence measure. The proposed method could also reduce the acoustic model size by 50% with less than a 1.5% increase in error rate compared to a baseline system.

• International Journal of Internet, Broadcasting and Communication
• /
• v.12 no.2
• /
• pp.45-50
• /
• 2020

A Language model with neural networks commonly trained with likelihood loss. Such that the model can learn the sequence of human text. State-of-the-art results achieved in various language generation tasks, e.g., text summarization, dialogue response generation, and text generation, by utilizing the language model's next token output probabilities. Monotonous and boring outputs are a well-known problem of this model, yet only a few solutions proposed to address this problem. Several decoding techniques proposed to suppress repetitive tokens. Unlikelihood training approached this problem by penalizing candidate tokens probabilities if the tokens already seen in previous steps. While the method successfully showed a less repetitive generated token, the method has a large memory consumption because of the training need a big vocabulary size. We effectively reduced memory footprint by encoding words as sequences of subword units. Finally, we report competitive results with token level unlikelihood training in several automatic evaluations compared to the previous work.

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.2E
• /
• pp.66-71
• /
• 2009

In practical applications of speech recognition, one of the fundamental questions might be on the number of training data that should be provided for a specific task. Though plenty of training data would undoubtedly enhance the system performance, we are then faced with the problem of heavy cost. Therefore, it is of crucial importance to determine the least number of training data that will afford a certain level of accuracy. For this purpose, we investigate the effect of the number of training data on the speaker-independent speech recognition of isolated words by using FVQ/HMM. The result showed that the error rate is roughly inversely proportional to the number of training data and grows linearly with the vocabulary size.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.8
• /
• pp.1110-1118
• /
• 1995

Over the past few years, a wide variety of techniques have been developed which make a reliable recognition of speech signal. Multi-layer perceptron(MLP) which has excellent pattern recognition properties is one of the most versatile networks in the area of speech recognition. This paper describes an automatic speech recognition system which use both MLP and lexicon. In this system., the recognition is performed by a network search algorithm which matches words in lexicon to MLP output scores. We also suggest a recognition algorithm which incorperat durational information of each phone, whose performance is comparable to that of conventional continuous HMM(CHMM). Performance of the system is evaluated on the database of 26 vocabulary size from 9 speakers. The experimental results show that the proposed algorithm achieves error rate of 7.3% which is 5.3% lower rate than 12.6% of CHMM.

• Asia Pacific Journal of Corpus Research
• /
• v.4 no.2
• /
• pp.57-74
• /
• 2023

Having a firm grasp of technical terms is essential for learners of Japanese for Specific Purposes (JSP). This research aims to analyze Japanese nursing care vocabulary based on objective corpus-based frequency and subjectively rated word familiarity. For this purpose, we constructed a text corpus centered on the National Examination for Certified Care Workers to extract nursing care keywords. The Log-Likelihood Ratio (LLR) was used as the statistical criterion for keyword identification, giving a list of 300 keywords as target words for a further word recognition survey. The survey involved 115 participants of whom 51 were certified care workers (CW group) and 64 were individuals from the general public (GP group). These participants rated the familiarity of the target keywords through crowdsourcing. Given the limited sample size, Bayesian linear mixed models were utilized to determine word familiarity rates. Our study conducted a comparative analysis of word familiarity between the CW group and the GP group, revealing key terms that are crucial for professionals but potentially unfamiliar to the general public. By focusing on these terms, instructors can bridge the knowledge gap more efficiently.