• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.236-238
• /
• 2019

신경망 기계 번역(Neural Machine Translation)은 주로 지도 학습(Supervised learning)을 이용한 End-to-end 방식의 연구가 이루어지고 있다. 그러나 지도 학습 방법은 데이터가 부족한 경우에는 낮은 성능을 보이기 때문에 BERT와 같은 대량의 단일 언어 데이터로 사전학습(Pre-training)을 한 후에 미세조정(Finetuning)을 하는 Transfer learning 방법이 자연어 처리 분야에서 주로 연구되고 있다. 최근에 발표된 MASS 모델은 언어 생성 작업을 위한 사전학습 방법을 통해 기계 번역과 문서 요약에서 높은 성능을 보였다. 본 논문에서는 영어-한국어 기계 번역 성능 향상을 위해 MASS 모델을 신경망 기계 번역에 적용하였다. 실험 결과 MASS 모델을 이용한 영어-한국어 기계 번역 모델의 성능이 기존 모델들보다 좋은 성능을 보였다.

• Journal of Science Education
• /
• v.35 no.2
• /
• pp.138-148
• /
• 2011

Science class is the starting point of science education where science teacher and students meet together to teach and learn science. Therefore successful science class is the main concern of various educational partakers, such as teachers, students, parents, school administrators, and science education researchers. In order to find what is important among the teaching abilities of science teachers, a survey was conducted to the various educational partakers. As the result of survey, students chose 'progressing a lesson' and 'teacher's attitude' are important, and in-service teachers chose 'analyzing learner characteristics and curriculum' and 'teacher-student interaction' are important, while the pre-service teachers and parents chose 'teacher-student interaction' and 'progressing a lesson' are important. In addition, the school administrators chose 'setting up teaching-learning strategy' and science education researchers chose 'understanding curriculum and developing teaching-learning method' is important along with 'analyzing learner characteristics and curriculum.' In the communication on science education among different educational partakers, these result can be used as the basic information to understand the different perspectives on a science class.

• Journal of The Korean Association For Science Education
• /
• v.38 no.2
• /
• pp.147-160
• /
• 2018

The purpose of this study is to construct a Teachers' Learning Community (TLC) with three teachers from the same school and to develop a joint teaching plan for students through the TLC. The conclusions from this study are as follows: The TLC with the same grade helps teachers to implement STEAM classes, where teachers overcame difficulties of STEAM lessons and successfully implemented them. Teachers in this study expressed difficulties of STEAM lessons including lack of time, difficulties of STEAM lesson implementation, and difficulties of developing a good STEAM lesson. Teachers worked together to develop a common teaching plan, to overcome the burden of teaching, and to plan better lessons through discussions and cross-checking. In addition, teachers newly discovered difficulties of lesson implementation as they watch each other teaching using a joint lesson plan. Teachers will conduct a better lesson as they improve these difficulties, where a better lesson means having students reach learning goals and learn from the lesson. Teachers in TLC felt that their lesson improved and they themselves growing through a series of courses of watching and learning each other's lessons.

• Journal of The Korean Association For Science Education
• /
• v.34 no.1
• /
• pp.10-20
• /
• 2014

This study compares the perceptions of elementary gifted child and science teacher in a science class for the gifted. In order to explore the research problem, students and teachers answered a written semi-structured questionnaire and participated in interviews regarding the gifted science class. The data was collected and analyzed. Science teachers recognized the characteristics of a good science class, especially in terms of educational content and teaching methodology. First, they suggested promoting inquiry skills, presenting a challenging task in atypical topic selection, student-centered curriculum, and controlling the pace of learning to recognize individual differences. Second, in terms of the science class skills and attitudes category, teachers recommended raising mutual satisfaction through vigorous interaction within a permissible atmosphere. Finally, science teachers need to strive for continued professional growth. Gifted children, meanwhile, want to investigate a wide range of topics without time constraints. Additionally, they may have to explore challenging topics further. They prefer to act like scientists in that they enjoy group activities, communication and cooperation. In particular, they want to be evaluated by others in a totally embedded assessment. Gifted children also expect teachers to understand the life circumstances and needs of the students. In addition, they asked for teachers to respect individual experiments and to show them how to safely use new equipment or research methods. As a result, gifted children and science teachers have to recognize the differences of opinion concerning a good science class for the gifted. This study can help formulate strategies to establish quality management of materials in gifted science classes.

• Journal of the Korean School Mathematics Society
• /
• v.12 no.3
• /
• pp.189-209
• /
• 2009

One of the characteristics in math -abstract concept- makes the students find difficulties in understanding general ideas about math. This study is about how much do the modeling lessons using the technical instruments which is based on the realistic mathematical theory influence on understanding the mathematical concept. This study is based on one of the contents the first grade of middle school students study, the function, especially the meaning of it. Some brilliant students being the objects of this study, mathematically experimental modeling lesson was planned, conducted. Survey on the students' attitudes about math before and after the modeling classes and Questionnaire survey on the effectiveness about the modeling class were conducted and their attitudes were recorded also. This study tells that students show very meaningful changes before and after the modeling class and scientific knowledge seems to be very helpful for the students to understand the mathematical concept and solve the problems. When scientific research and development get together with mathematics, students will be more motivated and be able to form the right mathematical concept easily.

• Journal of The Korean Association For Science Education
• /
• v.43 no.2
• /
• pp.111-124
• /
• 2023

The aim of this study was to explore science teachers' perceptions of good science teaching. To this end, the Korean Teaching Observation Protocol (KTOP), which was developed for the purpose of observing and improving science teaching, was utilized. In the first survey, teachers were asked whether they thought each item in the KTOP was important for good science teaching, the extent to which they implemented these items, and the level of difficulty in implementing them. The second survey asked teachers what they believed to be the reasons and solutions for the KTOP items that they had responded as difficult to implement. The responses obtained from 63 teachers in the first survey and 35 teachers in the second survey were categorized based on the characteristics of the responses. The categorized contents were then summarized and discussed for their features. As a result, science teachers responded that all items in KTOP, except for one, are important for good science teaching. However, it was also shown that the level of execution was low in cases where implementation was difficult. For the 13 KTOP items that were considered important but difficult to implement and showed relatively low implementation level, many respondents (69%) attributed the reason to both students and teachers. However, the most common response (60%) was that the teacher should solve those difficulties. From this, it was found that understanding and supporting teachers, as well as enhancing their competencies, are more important for good science teaching than external factors. We hope that this research findings will help to better understand the specific difficulties that science teachers face in their classes and contribute to practical efforts that aim to address these challenges.

• Journal of KIISE
• /
• v.44 no.7
• /
• pp.686-691
• /
• 2017

Telemarketing has become the center of marketing action of the industry in the information society. Recently, machine learning has emerged in many areas, especially, financial prediction. Financial data consists of lots of unlabeled data in most parts, and therefore, it is difficult for humans to perform their labeling. In this paper, we propose a fusion method of semi-supervised learning for automatic labeling of unlabeled data to predict telemarketing. Specifically, we integrate labeling results of label propagation and co-training with a decision tree. The data with lower reliabilities are removed, and the data are extracted that have consistent label from two labeling methods. After adding them to the training set, a decision tree is learned with all of them. To confirm the usefulness of the proposed method, we conduct the experiments with a real telemarketing dataset in a Portugal bank. Accuracy of the proposed method is 83.39%, which is 1.82% higher than that of the conventional method, and precision of the proposed method is 19.37%, which is 2.67% higher than that of the conventional method. As a result, we have shown that the proposed method has a better performance as assessed by the t-test.

• Journal of KIISE:Software and Applications
• /
• v.34 no.2
• /
• pp.160-168
• /
• 2007

For developing a text classifier using supervised learning, a manually labeled corpus of large size is required. However, it takes a lot of time and human effort. Recently a research paradigm was proposed to use a raw corpus and a small amount of seed information instead of manually labeled corpus. In this paper we introduce an unsupervised learning method that makes it possible to achieve better performance than other related works. The characteristics of our approach is that average mutual information is used to learn representative words and their weights and then update of the weights is done using a technique inspired by the works in information retrieval. By iterating this teaming process it was shown that a high performance system can be developed.

• Journal of KIISE
• /
• v.43 no.1
• /
• pp.71-79
• /
• 2016

Disambiguation of homographs is an important job in Korean semantic processing and has been researched for long time. Recently, machine learning approaches have demonstrated good results in accuracy and speed. Other knowledge-based approaches are being researched for untrained words. This paper proposes a hybrid method based on the machine learning approach that uses a lexical semantic network. The use of a hybrid approach creates an additional corpus from subcategorization information and trains this additional corpus. A homograph tagging phase uses the hypernym of the homograph and an additional corpus. Experimentation with the Sejong Corpus and UWordMap demonstrates the hybrid method is to be effective with an increase in accuracy from 96.51% to 96.52%.

• Journal of Science Education
• /
• v.33 no.2
• /
• pp.248-257
• /
• 2009

The purpose of this study is to investigate the learning achievement of students who has taken the play-based and activity-oriented programs in the class. These new science class programs and instrument to measure student learning achievement were developed for the "Constellation Favorite" lesson of 4th grade students and Those are verified by science education experts, teachers in serve. The subjects of this study consist of 121 students in four classes who were selected on the basis of midterm examination result before teaching treatment. They were divided into two categories with two classes per each category, one class for experimental group and the other class for control group. The experimental groups were given five classes including five play-based class programs. The control groups were given five conventional instruction-type classes. The students were given questionnaires to test their interest ratings. The results show about 17% increase of their learning achievement with (p<0.005) by independent sample T-test method and 40% increase of interest rating in play-based class sample compared to normal class sample. In conclusion, the play-based science class program is found to enhance the students' learning achievement and to interest the students more effectively.