• ETRI Journal
• /
• v.46 no.1
• /
• pp.48-58
• /
• 2024

This paper presents the development of language tutoring systems for nonnative speakers by leveraging advanced end-to-end automatic speech recognition (ASR) and proficiency evaluation. Given the frequent errors in non-native speech, high-performance spontaneous speech recognition must be applied. Our systems accurately evaluate pronunciation and speaking fluency and provide feedback on errors by relying on precise transcriptions. End-to-end ASR is implemented and enhanced by using diverse non-native speaker speech data for model training. For performance enhancement, we combine semisupervised and transfer learning techniques using labeled and unlabeled speech data. Automatic proficiency evaluation is performed by a model trained to maximize the statistical correlation between the fluency score manually determined by a human expert and a calculated fluency score. We developed an English tutoring system for Korean elementary students called EBS AI Peng-Talk and a Korean tutoring system for foreigners called KSI Korean AI Tutor. Both systems were deployed by South Korean government agencies.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.59-70
• /
• 2024

We introduce a high-performance named entity recognition (NER) model for written and spoken language. To overcome challenges related to labeled data scarcity and domain shifts, we use transfer learning to leverage our previously developed KorBERT as the base model. We also adopt a meta-pseudo-label method using a teacher/student framework with labeled and unlabeled data. Our model presents two modifications. First, the student model is updated with an average loss from both human- and pseudo-labeled data. Second, the influence of noisy pseudo-labeled data is mitigated by considering feedback scores and updating the teacher model only when below a threshold (0.0005). We achieve the target NER performance in the spoken language domain and improve that in the written language domain by proposing a straightforward rollback method that reverts to the best model based on scarce human-labeled data. Further improvement is achieved by adjusting the label vector weights in the named entity dictionary.

• Analytical Science and Technology
• /
• v.24 no.2
• /
• pp.127-134
• /
• 2011

Quantitative HPLC analysis for the determination of in herbal liquid preparations was improved from the general test method besides the Korean Pharmacopeia. Good chromatographic separation of samples containing parabens, interferences, and other pharmaceutical excipients was effectively achieved by using acetonitrile water (containing 1% glacial acetic acid) mixture (30:70 v/v) as mobile phase. To monitor preservatives (benzoic acid, parabens, sorbic acid, dehydroacetic acid, and their salts) in herbal liquid preparations, a group of 47 samples was divided into two different group: preservative labeled group and unlabeled group. From the results, the contents of preservatives in 31 samples of preservative labeled group fell under KFDA regulations, and the contents of dehydroacetic acid in 6 samples of preservative labeled group were not followed by KFDA regulations. Preservatives were detected in 3 samples out of 10 samples in preservative unlabeled group.

• Journal of Life Science
• /
• v.15 no.1 s.68
• /
• pp.55-60
• /
• 2005

The murine dopamine receptor regulating factor (DRRF) gene is transcribed from a TATA-less promoter that has several putative Sp1 binding sites. The present investigation identifies functional transcription factors that modulate the expression of this gene, In the $D_2-expressing$ NB41A3 cells, Spl potently activates transcription from the DRRF promoter in pCAT-DRRF-1153/+17, but DRRF effectively inhibits it. Deletion of the 31 bp fragment between -1153 and -1122 decreased transcription down to about $60\%$. This fragment contains a functional API binding site. In addition, deletion of the 129 bp region between -901 and -772 further decreased transcription. The latter region has a functional AP2 binding site. Using a DRRF_AP1 (bases -1153 to -1121) probe, a specific retarded band was observed, and the unlabeled AP1 consensus competitor could effectively compete away this retarded band. In addition, using a DRRF_AP2 (bases -873 to -846), a specific retarded band was observed, and the unlabeled AP2 consensus competitor could effectively compete away this retarded band. The present observations suggest that Spl and DRRF regulate the DRRF promoter and that both API and AP2 also modulate this gene.

• Journal of Broadcast Engineering
• /
• v.24 no.3
• /
• pp.472-484
• /
• 2019

This paper proposes a time-domain sound event detection algorithm using DNN (Deep Neural Network). In this system, time domain sound waveform data which is not converted into the frequency domain is used as input to the DNN. The overall structure uses CRNN structure, and GLU, ResNet, and Squeeze-and-excitation blocks are applied. And proposed structure uses structure that considers features extracted from several layers together. In addition, under the assumption that it is practically difficult to obtain training data with strong labels, this study conducted training using a small number of weakly labeled training data and a large number of unlabeled training data. To efficiently use a small number of training data, the training data applied data augmentation methods such as time stretching, pitch change, DRC (dynamic range compression), and block mixing. Unlabeled data was supplemented with insufficient training data by attaching a pseudo-label. In the case of using the neural network and the data augmentation method proposed in this paper, the sound event detection performance is improved by about 6 %(based on the f-score), compared with the case where the neural network of the CRNN structure is used by training in the conventional method.

• Journal of Internet Computing and Services
• /
• v.24 no.6
• /
• pp.91-98
• /
• 2023

Recently, according to development of artificial intelligence, a wide range of industry being automatic and optimized. Also we can find out some research of using supervised learning for deteceting defect of railway in domestic rail industry. However, there are structures other than rails on the track, and the fastener is a device that binds the rail to other structures, and periodic inspections are required to prevent safety accidents. In this paper, we present a method of reducing cost for labeling using semi-supervised and transfer model trained on rail fastener data. We use Resnet50 as the backbone network pretrained on ImageNet. At first we randomly take training data from unlabeled data and then labeled that data to train model. After predict unlabeled data by trained model, we adopted a method of adding the data with the highest probability for each class to the training data by a predetermined size. Futhermore, we also conducted some experiments to investigate the influence of the number of initially labeled data. As a result of the experiment, model reaches 92% accuracy which has a performance difference of around 5% compared to supervised learning. This is expected to improve the performance of the classifier by using relatively few labels without additional labeling processes through the proposed method.

• Korean Journal of Radiology
• /
• v.24 no.6
• /
• pp.541-552
• /
• 2023

Objective: Detection of pneumoperitoneum using abdominal radiography, particularly in the supine position, is often challenging. This study aimed to develop and externally validate a deep learning model for the detection of pneumoperitoneum using supine and erect abdominal radiography. Materials and Methods: A model that can utilize "pneumoperitoneum" and "non-pneumoperitoneum" classes was developed through knowledge distillation. To train the proposed model with limited training data and weak labels, it was trained using a recently proposed semi-supervised learning method called distillation for self-supervised and self-train learning (DISTL), which leverages the Vision Transformer. The proposed model was first pre-trained with chest radiographs to utilize common knowledge between modalities, fine-tuned, and self-trained on labeled and unlabeled abdominal radiographs. The proposed model was trained using data from supine and erect abdominal radiographs. In total, 191212 chest radiographs (CheXpert data) were used for pre-training, and 5518 labeled and 16671 unlabeled abdominal radiographs were used for fine-tuning and self-supervised learning, respectively. The proposed model was internally validated on 389 abdominal radiographs and externally validated on 475 and 798 abdominal radiographs from the two institutions. We evaluated the performance in diagnosing pneumoperitoneum using the area under the receiver operating characteristic curve (AUC) and compared it with that of radiologists. Results: In the internal validation, the proposed model had an AUC, sensitivity, and specificity of 0.881, 85.4%, and 73.3% and 0.968, 91.1, and 95.0 for supine and erect positions, respectively. In the external validation at the two institutions, the AUCs were 0.835 and 0.852 for the supine position and 0.909 and 0.944 for the erect position. In the reader study, the readers' performances improved with the assistance of the proposed model. Conclusion: The proposed model trained with the DISTL method can accurately detect pneumoperitoneum on abdominal radiography in both the supine and erect positions.

• Biomolecules & Therapeutics
• /
• v.32 no.1
• /
• pp.154-161
• /
• 2024

Amyotrophic lateral sclerosis (ALS) is a fatal motor neuron disorder that causes progressive paralysis. L-Citrulline is a nonessential neutral amino acid produced by L-arginine via nitric oxide synthase (NOS). According to previous studies, the pathogenesis of ALS entails glutamate toxicity, oxidative stress, protein misfolding, and neurofilament disruption. In addition, L-citrulline prevents neuronal cell death in brain ischemia; therefore, we investigated the change in the transport of L-citrulline under various pathological conditions in a cell line model of ALS. We examined the uptake of [¹⁴C]L-citrulline in wild-type (hSOD1wt/WT) and mutant NSC-34/ SOD1^G93A (MT) cell lines. The cell viability was determined via MTT assay. A transport study was performed to determine the uptake of [¹⁴C]L-citrulline. Quantitative real-time polymerase chain reaction (qRT-PCR) analysis was performed to determine the expression levels of rat large neutral amino acid transported 1 (rLAT1) in ALS cell lines. Nitric oxide (NO) assay was performed using Griess reagent. L-Citrulline had a restorative effect on glutamate induced cell death, and increased [¹⁴C]L-citrulline uptake and mRNA levels of the large neutral amino acid transporter (LAT1) in the glutamate-treated ALS disease model (MT). NO levels increased significantly when MT cells were pretreated with glutamate for 24 h and restored by co-treatment with L-citrulline. Co-treatment of MT cells with L-arginine, an NO donor, increased NO levels. NSC-34 cells exposed to high glucose conditions showed a significant increase in [¹⁴C]L-citrulline uptake and LAT1 mRNA expression levels, which were restored to normal levels upon co-treatment with unlabeled L-citrulline. In contrast, exposure of the MT cell line to tumor necrosis factor alpha, lipopolysaccharides, and hypertonic condition decreased the uptake significantly which was restored to the normal level by co-treating with unlabeled L-citrulline. L-Citrulline can restore NO levels and cellular uptake in ALS-affected cells with glutamate cytotoxicity, pro-inflammatory cytokines, or other pathological states, suggesting that L-citrulline supplementation in ALS may play a key role in providing neuroprotection.

• Bulletin of the Korean Chemical Society
• /
• v.34 no.9
• /
• pp.2725-2730
• /
• 2013

Single C-reactive protein (CRP) molecules, which are non-specific acute phase markers and products of the innate immune system, were quantitatively detected on a gold-nanopatterned biochip using evanescent field-enhanced fluorescence imaging. The $4{\times}5$ gold-nanopatterned biochip (spot diameter of 500 nm) was fabricated by electron beam nanolithography. Unlabeled CRP molecules in human serum were identified with single-molecule sandwich immunoassay by detecting secondary fluorescence generated by total internal reflection fluorescence (TIRF) microscopy. With decreased standard CRP concentrations, relative fluorescence intensities reduced in the range of 33.3 zM-800 pM. To enhance fluorescence intensities in TIRF images, the distance between biochip surface and CRP molecules was optimally adjusted by considering the quenching effect of gold and the evanescent field intensity. As a result, TIRF only detected one single-CRP molecule on the biochip the first time.

• BMB Reports
• /
• v.29 no.2
• /
• pp.169-174
• /
• 1996

Yeast cytochrome c (cyt c) was modified at cysteine-102 with a thiol-specific spin label and its interaction with liposomes containing acidic phospholipids was studied by electron paramagnetic resonance (EPR) spectroscopy. Association of cyt c with liposomes resulted in a significant reduction in the mobility of the spin label and a fraction of cyt c even seemed to be immobilized. Based on a large spectral change upon binding and the proximity of the spin-label to lysine-86 and -87, we propose these two residues to be the potential binding site at neutral pH. The interaction is electrostatic in nature because the spectral changes were reversed by addition of anions. Dissociation of the bound cyt c by anions, however, became less effective as the lipid/protein ratio increased. This suggests a repulsive lateral interaction among the bound cyt c. Unlabeled cyt c molecules added to preformed cyt c-liposome complex displaced the bound (spin labeled) cyt c and the process was competitive and reversible.