Digital Library[ Search Result ]
An Automated Error Detection Method for Speech Transcription Corpora Based on Speech Recognition and Language Models
Jeongpil Lee, Jeehyun Lee, Yerin Choi, Jaehoo Jang, Myoung-Wan Koo
http://doi.org/10.5626/JOK.2024.51.4.362
This research proposes a "machine-in-the-loop" approach for automatic error detection in Korean speech corpora by integrating the knowledge of CTC-based speech recognition models and language models. We experimentally validated its error detection performance through a three-step procedure that leveraged Character Error Rate (CER) from the speech recognition model and Perplexity (PPL) from the language model to identify potential transcription error candidates and verify their text labels. This research focused on the Korean speech corpus, KsponSpeech, resulting in a reduction of the character error rate on the test set from 9.44% to 8.9%. Notably, this performance enhancement was achieved even when inspecting only approximately 11% of the test data, highlighting the higher efficiency of our proposed method than a comprehensive manual inspection process. Our study affirms the potential of this efficient "machine-in-the-loop" approach for a cost-effective error detection mechanism in speech data while ensuring accuracy.
Prediction of Toothbrushing Position Based on Gyro Sensor Data and its Validation Using Unsupervised Learning-based Clustering
DoYoon Kim, MinWook Kwon, SeungJu Baek, HyeRin Yoon, DaeYeon Lim, Eunah Jo, Seungjae Ryu, Young Wook Kim, Jin Hyun Kim
http://doi.org/10.5626/JOK.2023.50.12.1143
Oral health is an important health indicator that is directly related to longevity. For this reason, oral health has become a key component of public health, from infants to the elderly. The foundation of good oral health is good brushing habits. However, the recommended correct brushing method is not easy to adopt, and this harms oral health. This paper proposes a method to distinguish brushing zones using low-cost IMU sensors to track the correct brushing method. We evaluated the accuracy of the brushing zone estimation method using clustering algorithms in machine learning. In this paper, we propose a method for determining the brushing area based on toothbrush posture alone using the gyro sensor of an IMU sensor. In this paper, we propose a method for determining the brushing area using only the gyro sensor of an IMU sensor based on toothbrush posture. We showed that relatively inexpensive 6-axis IMU gyro sensor data could be used to estimate the user’s brushing area with an accuracy of 80.6%. In addition, we applied a clustering algorithm to these data and trained a logistic regression model using the clustered data to estimate the brushing area. The result was obtained with an accuracy of 86.7%, showing that clustering was effective and that the toothbrush posture-based brushing area estimation proposed in this paper was effective. In conclusion, it is expected that the brushing zone estimation algorithm can be implemented as a function of a relatively low-cost toothbrush and that it can help to maintain oral health by analyzing and improving personal brushing habits.
Improvement of Machine Learning-Based Event-Related Desynchronization Accuracy
http://doi.org/10.5626/JOK.2023.50.12.1131
The biometrics field is known for providing fast and accurate identity verification. Recently, motor imagery (MI) brainwaves have gained prominence, accompanied by event-related desynchronization (ERD) signals. The purpose of this study is to optimize existing ERD models to enhance inter-user classification accuracy. We used a well-known common spatial pattern (CSP) and ERD as representative features for MI, and classified them using naïve bayes (NB). To evaluate the reliability of the binary classification results of the SVM, equal error rate (EER) and area under the curve (AUC) were used. The proposed ERD model exhibited superior accuracy compared to CSP and traditional ERD, achieving classification accuracies of 86.4%, 86.3%, and 63%, respectively. Based on this results, the proposed ERD method is presented as a suitable future biometric marker.
Robust Korean Table Machine Reading Comprehension across Various Domains
Sanghyun Cho, Hye-Lynn Kim, Hyuk-chul Kwon
http://doi.org/10.5626/JOK.2023.50.12.1102
Unlike regular text data, tabular data has structural features that allow it to represent compressed information. This has led to their use in a variety of domains, and machine reading comprehension of tables has become an increasingly important aspect of Machine Reading Comprehension(MRC). However, the structure of tables and the knowledge required for each domain are different, and when a language model is trained for a single domain, the evaluation performance of the model in other domains is likely to be reduced, resulting in poor generalization performance. To overcome this, it is important to build datasets of various domains and apply various techniques rather than simply pre-trained models. In this study, we design a language model that learns cross-domain invariant linguistic features to improve domain generalization performance. We applied adversarial training to improve performance on evaluation datasets in each domain and modify the structure of the model by adding an embedding layer and a transformer layer specialized for tabular data. When applying adversarial learning, we found that the model with a structure that does not add table-specific embeddings improves performance. On the other hand, while adding a table-specific transformer layer and having the added layer receive additional table-specific embeddings as input, shows the best performance on data from all domains.
Predicting Significant Blood Marker Values for Pressure Ulcer Forecasting Utilizing Feature Minimization and Selection
Yeonhee Kim, Hoyoul Jung, Jang-Hwan Choi
http://doi.org/10.5626/JOK.2023.50.12.1054
Pressure ulcers are difficult to treat once they occur, and huge economic costs are incurred during the treatment process. Therefore, predicting the occurrence of pressure ulcers is important in terms of patient suffering and economics. In this study, the correlation between the lab codes (features) and pressure ulcers obtained from blood tests of patients with spinal cord injury was analyzed to provide meaningful characteristic information for the prediction of pressure ulcers. We compare and analyze the correlation coefficients of Pearson, Spearman, and Kendall"s tau, which are mainly used in feature selection methods. In addition, the importance of features is calculated using XGBoost and LightGBM, which are machine learning methods based on gradient boosting. In order to verify the performance of this model, we use the long short-term memory (LSTM) model to predict other features using the features occupying the top-5 in importance. In this way, unnecessary features can be minimized in diagnosing pressure ulcers and guidelines can be provided to medical personnel.
Dovetail Usage Prediction Model for Resource-Efficient Virtual Machine Placement in Cloud Computing Environment
Hyeongbin Kang, Hyeon-Jin Yu, Jungbin Kim, Heeseok Jeong, Jae-Hyuck Shin, Seo-Young Noh
http://doi.org/10.5626/JOK.2023.50.12.1041
As IT services have migrated to the cloud, efficient resource management in cloud computing environments has become an important issue. Consequently, research has been conducted on virtual machine placement(VMP), which can increase resource efficiency without the need for additional equipment in data centers. This paper proposes the use of a usage prediction model as a method for selecting and deploying hosts suitable for virtual machine placement. The dovetail usage prediction model, which improves the shortcomings of the existing usage prediction models, measures indicators such as CPU, disk, and memory usage of virtual machines running on hosts and extracts features using a deep learning model by converting them into time series data. By utilizing this approach in virtual machine placement, hosts can be used efficiently while ensuring appropriate load balancing of the virtual machines.
A Comparative Analysis of the Motion Recognition Rate by Direction of Push-up Activity Using ELM Algorithm
Sangwoong Kim, Jaeyeong Ryu, Jiwoo Jeong, Dongyeong Kim, Youngho Chai
http://doi.org/10.5626/JOK.2023.50.12.1031
In this paper, we propose a motion recognition system for each direction of push-up activity using ELM algorithm. In the proposed system, a recognized motion consists of three parts. The first part is the process of reading motion data. In the process, the data acquired from the motion capture system is entered into the system"s memory. Then, the system extracts a feature vector from the motion data. The 3D position data converted from the quaternion data value of the motion data is projected onto the X-Y plane, Y-Z plane and Z-X plane of the system, and the values are used as the final feature vector. Feature vectors projected on each plane train different ELM, and a total of three ELM are learned. Finally, by inputting test data to each learned ELM, the final recognition result value is derived. First, before obtaining motion data, as the data set to be trained, general push-ups performed in the correct posture were selected. Second, the upper chest did not go down all the way. Third, only the buttocks came up when bending and lifting. Four, when bending your elbows move away from your upper chest. Finally, mix these motions to build a test dataset.
Applying Deep Neural Networks and Random Forests to Predict the Pathogenicity of Single Nucleotide Variants in Hereditary Cancer-associated Genes
Da-Bin Lee, Seonhwa Kim, Moonjong Kang, Changbum Hong, Kyu-Baek Hwang
http://doi.org/10.5626/JOK.2023.50.9.746
The recent proliferation of genetic testing has made it possible to explore an individual"s genetic variants and use pathogenicity information to diagnose and prevent genetic diseases. However, the number of identified variants with pathogenicity information is quite small. A method for predicting the pathogenicity of variants by machine learning was proposed to address this problem. In this study, we apply and compare deep neural networks with random forests and logistic regression, which have been widely used in previous studies, to predict variant pathogenicity. The experimental data consisted of 1,068 single-nucleotide variants in genes associated with hereditary cancers. Experiments on 100 random data-sets generated for hyperparameter selection showed that random forests performed best in terms of area under the precision-recall curve. On 15 holdout gene data-sets, deep neural networks performed best on average, but the difference in performance from the second-best random forest was not significant. Logistic regression was also statistically significantly worse than that of either model. In conclusion, we found that deep neural networks and random forests were generally better than logistic regression at predicting the pathogenicity of single-nucleotide variants associated with hereditary cancer.
Machine Learning-Based Approach for Predicting Drug-Induced Liver Injury of Chemical Compounds
http://doi.org/10.5626/JOK.2023.50.9.777
Drug-induced liver injury (DILI) is one of the factors constraining the distribution of investigational products on the market. Therefore, DILI risk of compounds should be assessed in advance. Although in vivo and in vitro methods can be used to test drug safety, both methods are labor-intensive, time consuming and expensive. In this study, we suggested random forest, light gradient boosting machine, logistic regression models to overcome the above problems. These models used molecular structure and physicochemical features as input to predict the DILI as output. The optimal model was random forest, which performed well for evaluation metrics overall. The proposed model is expected to help drug development process by identifying potential DILI of drug candidates in advance.
Vision-based Position Deviation Fault Injection Method for Building a Collaborative Robot Motion Fault Dataset
Donghee Yun, Dongyeon Yoo, Jungwon Lee
http://doi.org/10.5626/JOK.2023.50.9.795
The data-based fault detection method, which collects data from internal and external sensors in real-time and predicts fault, is being applied to collaborative robots, which are key facilities in smart factories. The data-based fault detection method requires a large amount of data for learning, and in particular, a large amount of data labeled as a fault state is essential. However, it is difficult to obtain large amounts of actual fault data in industrial settings. Therefore, in this study, the output of the collaborative robot fault state based on a vision sensor was analyzed and compared with the output of the normal state, and a fault injection method was proposed based on the deviation between the analyzed output signals. Collaborative robot data collected in the actual fault state could be replaced with data collected in the proposed fault injection state. The comparison of the performance of the model trained with fault injection data and trained with actual fault data confirmed that there was almost no difference, with an average of 0.97 and 0.98 accuracy, thus verifying the effectiveness of the proposed fault injection method.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr