• Korean Journal of Cognitive Science
• /
• v.26 no.2
• /
• pp.147-165
• /
• 2015

Through short-answer questions, we can reflect the depth of students' understanding and higher-order thinking skills. Scoring for short-answer questions may take long time and may be an issue on consistency of grading. To alleviate such the suffering, automated scoring systems are widely used in Europe and America, but are in the initial stage in research in Korea. In this paper, we propose a semi-automatic scoring system for short Korean free-text responses using semi-supervised learning. First of all, based on the similarity score between students' answers and model answers, the proposed system grades students' answers and the scored answers with high reliability have been included in the model answers through the thorough test. This process repeats until all answers are scored. The proposed system is used experimentally in Korean and social studies in Nationwide Scholastic Achievement Test. We have confirmed that the processing time and the consistency of grades are promisingly improved. Using the system, various assessment methods have got to be developed and comparative studies need to be performed before applying to school fields.

• Journal of KIISE:Information Networking
• /
• v.32 no.3
• /
• pp.318-326
• /
• 2005

This paper presents the image feature extraction method and the image classification technique to select the harmful image flowed from the Internet by grade of image contents such as harmlessness, sex-appealing, harmfulness (nude), serious harmfulness (adult) by the characteristic of the image. In this paper, we suggest skin area detection technique to recognize whether an input image is harmful or not. We also propose the ROI detection algorithm that establishes region of interest to reduce some noise and extracts harmful degree effectively and defines the characteristics in the ROI area inside. And this paper suggests the multiple-SVM training method that creates the image classification model to select as 4 types of class defined above. This paper presents the multiple-SVM classification algorithm that categorizes harmful grade of input data with suggested classification model. We suggest the skin likelihood image made of the shape information of the skin area image and the color information of the skin ratio image specially. And we propose the image feature vector to use in the characteristic category at a course of traininB resizing the skin likelihood image. Finally, this paper presents the performance evaluation of experiment result, and proves the suitability of grading image using image feature classification algorithm.

• Investigative Magnetic Resonance Imaging
• /
• v.20 no.2
• /
• pp.81-87
• /
• 2016

Purpose: To analyze the feasibility of three-dimensional (3D) diffusion-weighted (DW) PSIF (reversed FISP [fast imaging with steady-state free precession]) sequence in order to evaluate peripheral nerves in the elbow. Materials and Methods: Ten normal, asymptomatic volunteers were enrolled (6 men, 4 women, mean age 27.9 years). The following sequences of magnetic resonance images (MRI) of the elbow were obtained using a 3.0-T machine: 3D DW PSIF, 3D T2 SPACE (sampling perfection with application optimized contrasts using different flip angle evolution) with SPAIR (spectral adiabatic inversion recovery) and 2D T2 TSE (turbo spin echo) with modified Dixon (m-Dixon) sequence. Two observers used a 5-point grading system to analyze the image quality of the ulnar, median, and radial nerves. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of each nerve were measured. We compared 3D DW PSIF images with other sequences using the Wilcoxon-signed rank test and Friedman test. Inter-observer agreement was measured using intraclass correlation coefficient (ICC) analysis. Results: The mean 5-point scores of radial, median, and ulnar nerves in 3D DW PSIF (3.9/4.2/4.5, respectively) were higher than those in 3D T2 SPACE SPAIR (1.9/2.8/2.8) and 2D T2 TSE m-Dixon (1.7/2.8/2.9) sequences (P < 0.05). The mean SNR in 3D DW PSIF was lower than 3D T2 SPACE SPAIR, but there was no difference between 3D DW PSIF and 2D T2 TSE m-Dixon in all of the three nerves. The mean CNR in 3D DW PSIF was lower than 3D T2 SPACE SPAIR and 2D T2 TSE m-Dixon in the median and ulnar nerves, but no difference among the three sequences in the radial nerve. Conclusion: The three-dimensional DW PSIF sequence may be feasible to evaluate the peripheral nerves around the elbow in MR imaging. However, further optimization of the image quality (SNR, CNR) is required.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.3
• /
• pp.133-140
• /
• 2022

We investigate the performance of deep learning-based Korean language models on a task of predicting the score range of Korean essays written by foreign students. We construct a data set containing a total of 304 essays, which include essays discussing the criteria for choosing a job ('job'), conditions of a happy life ('happ'), relationship between money and happiness ('econ'), and definition of success ('succ'). These essays were labeled according to four letter grades (A, B, C, and D), and a total of eleven essay score range prediction experiments were conducted (i.e., five for predicting the score range of 'job' essays, five for predicting the score range of 'happiness' essays, and one for predicting the score range of mixed topic essays). Three deep learning-based Korean language models, KoBERT, KcBERT, and KR-BERT, were fine-tuned using various training data. Moreover, two traditional probabilistic machine learning classifiers, naive Bayes and logistic regression, were also evaluated. Experiment results show that deep learning-based Korean language models performed better than the two traditional classifiers, with KR-BERT performing the best with 55.83% overall average prediction accuracy. A close second was KcBERT (55.77%) followed by KoBERT (54.91%). The performances of naive Bayes and logistic regression classifiers were 52.52% and 50.28% respectively. Due to the scarcity of training data and the imbalance in class distribution, the overall prediction performance was not high for all classifiers. Moreover, the classifiers' vocabulary did not explicitly capture the error features that were helpful in correctly grading the Korean essay. By overcoming these two limitations, we expect the score range prediction performance to improve.