• Journal of Intelligence and Information Systems
• /
• v.28 no.4
• /
• pp.309-328
• /
• 2022

This study empirically analyzed a Korean pre-trained language models (PLMs) designed for natural language generation. The performance of two PLMs - BART and GPT - at the task of abstractive text summarization was compared. To investigate how performance depends on the characteristics of the inference data, ten different document types, containing six types of informational content and creation content, were considered. It was found that BART (which can both generate and understand natural language) performed better than GPT (which can only generate). Upon more detailed examination of the effect of inference data characteristics, the performance of GPT was found to be proportional to the length of the input text. However, even for the longest documents (with optimal GPT performance), BART still out-performed GPT, suggesting that the greatest influence on downstream performance is not the size of the training data or PLMs parameters but the structural suitability of the PLMs for the applied downstream task. The performance of different PLMs was also compared through analyzing parts of speech (POS) shares. BART's performance was inversely related to the proportion of prefixes, adjectives, adverbs and verbs but positively related to that of nouns. This result emphasizes the importance of taking the inference data's characteristics into account when fine-tuning a PLMs for its intended downstream task.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.8
• /
• pp.341-354
• /
• 2023

The title of a document is the brief summarization of the document. Readers can easily understand a document if we provide them with its title in their preferred styles and the languages. In this research, we propose a cross-lingual and style-based title generation model using multiple adapters. To train the model, we need a parallel corpus in several languages with different styles. It is quite difficult to construct this kind of parallel corpus; however, a monolingual title generation corpus of the same style can be built easily. Therefore, we apply a zero-shot strategy to generate a title in a different language and with a different style for an input document. A baseline model is Transformer consisting of an encoder and a decoder, pre-trained by several languages. The model is then equipped with multiple adapters for translation, languages, and styles. After the model learns a translation task from parallel corpus, it learns a title generation task from monolingual title generation corpus. When training the model with a task, we only activate an adapter that corresponds to the task. When generating a cross-lingual and style-based title, we only activate adapters that correspond to a target language and a target style. An experimental result shows that our proposed model is only as good as a pipeline model that first translates into a target language and then generates a title. There have been significant changes in natural language generation due to the emergence of large-scale language models. However, research to improve the performance of natural language generation using limited resources and limited data needs to continue. In this regard, this study seeks to explore the significance of such research.

• Annual Conference on Human and Language Technology
• /
• 2017.10a
• /
• pp.116-121
• /
• 2017

cQA(Community-based Question Answering) 시스템은 온라인 커뮤니티를 통해 사용자들이 질문을 남기고 답변을 작성할 수 있도록 만들어진 시스템이다. 신규 질문이 인입되면, 기존에 축적된 cQA 저장소에서 해당 질문과 가장 유사한 질문을 검색하고, 그 질문에 대한 답변을 신규 질문에 대한 답변으로 대체할 수 있다. 하지만, 키워드 매칭을 사용하는 전통적인 검색 방식으로는 문장에 내재된 의미들을 이용할 수 없다는 한계가 있다. 이를 극복하기 위해서는 의미적으로 동일한 문장들로 학습이 되어야 하지만, 이러한 데이터를 대량으로 확보하기에는 어려움이 있다. 본 논문에서는 질문이 제목과 내용으로 분리되어 있는 대량의 cQA 셋에서, 질문 제목과 내용을 의미 벡터 공간으로 사상하고 두 벡터의 상대적 거리가 가깝게 되도록 학습함으로써 의사(pseudo) 유사 의미의 성질을 내재화 하였다. 또한, 질문 제목과 내용의 의미 벡터 표현(representation)을 위하여, semi-training word embedding과 CNN(Convolutional Neural Network)을 이용한 딥러닝 기법을 제안하였다. 유사 질문 검색 실험 결과, 제안 모델을 이용한 검색이 키워드 매칭 기반 검색보다 좋은 성능을 보였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.1
• /
• pp.364-380
• /
• 2016

The most popular approach in object classification is based on the bag of visual-words model, which has several fundamental problems that restricting the performance of this method, such as low time efficiency, the synonym and polysemy of visual words, and the lack of spatial information between visual words. In view of this, an object classification based on weakly supervised E2LSH and saliency map weighting is proposed. Firstly, E2LSH (Exact Euclidean Locality Sensitive Hashing) is employed to generate a group of weakly randomized visual dictionary by clustering SIFT features of the training dataset, and the selecting process of hash functions is effectively supervised inspired by the random forest ideas to reduce the randomcity of E2LSH. Secondly, graph-based visual saliency (GBVS) algorithm is applied to detect the saliency map of different images and weight the visual words according to the saliency prior. Finally, saliency map weighted visual language model is carried out to accomplish object classification. Experimental results datasets of Pascal 2007 and Caltech-256 indicate that the distinguishability of objects is effectively improved and our method is superior to the state-of-the-art object classification methods.

• Journal of Korean Library and Information Science Society
• /
• v.14
• /
• pp.181-215
• /
• 1987

Reading education is very important in order to promote the refinement, cultivate the emotion and complete the character to the secondary school students. This thesis deals with the establishment of reading education as a formal course in secondary schools, responsibility of teaching and problems related to recommended reading lists. Reading education must separate from the national language education because of literature centered education in reading education. If reading education was separated from the national language education, students can a n.0, pproach to the other cultural boundary besides other own and exchange their information and ideas. So, reading education must be included to the elective subjects in a independent course or become a compulsory subject in secondary school curriculum. The teacher of reading education must become the teacher librarian who has a firm faith and an intellectual accomplishment. But, teacher-librarian has much disadvantages such as the problems of promotion, the division of qualification between elementary school and secondary school, and a short-term training courses for teacher-librarian. Hence, theses problems music be solved in national administrative level. Recommended reading lists must be provided to the student in order to prevent confusion of the sense of value, to estimate their own reading ability by themselves and to establish life long reading plan. Therefore, both Korean Library Association and the Ministry of Education should re-examine and develop recommended reading lists. Finally, problems of a juvenile delinquency in the post industrial society have to be solved through reading education. To solve the juvenile delinquency problems, adolescents should cultivate their moral character and possesses abundant knowledge through reading education. Then, young adults will grow as sound citizen in the society.

• Journal of The Korean Association of Information Education
• /
• v.14 no.4
• /
• pp.547-553
• /
• 2010

Lately, studies on the educational effectiveness of educational programming language which can reduce the learning burden of the learners have been conducted in the programming learning process. This study analyzed the effect of programming education on the learning motivation and academic achievement after training the programming education using Scratch and Dolittle on the preliminary elementary school teachers. As a result, the experimental group trained by Scratch programming education showed significantly higher achievement than the control group by Dolittle Programming. This result can be helpful in selecting educational programming language when the programming education will be trained to the preliminary elementary school teachers.

• Phonetics and Speech Sciences
• /
• v.5 no.3
• /
• pp.83-93
• /
• 2013

Vocal hygiene education is an indirect training approach to improve vocal function by educating all facets of optimal vocal health. Satisfaction levels of participants might be an important component of this indirect therapy for voice disorders. The authors aimed to investigate the satisfaction levels of vocal hygiene education in 51 patients with voice problems. We classified voice disorders of the participants according to three etiological categories (subgroups): organic, neurogenic, and functional. The survey consisted of three parts: 1) a condition of vocal hygiene education, 2) a degree of satisfaction of the present education, and 3) a request for future education. Participants responded to each item of the survey using a five-point Likert scale of 1 to 5 (1 being not at all and 5 being extremely). They also wrote down personal comments of improvement. Participants scored the vocal hygiene education offered by the speech-language pathologists between '3' and '4'. Specifically, the participants were highly satisfied with the specific and comprehensible explanation/instruction given by their speech-language pathologists. However, they were less satisfied with the tuition fee for the therapy sessions. Vocal hygiene education is offered individually to people in a clinical setting. Our results support the notion that vocal hygiene education can be an integral aspect of the treatment of voice problems in most cases.

• Journal of Fisheries and Marine Sciences Education
• /
• v.28 no.3
• /
• pp.818-830
• /
• 2016

The purpose of this study is to examine the effects of the parents listening guidance program, which allows the parents to understand their hearing impaired children and how to listen at home, on the communication skills of the hearing impaired children. The research subjects were 3 hearing impaired children who did not accompany with the intellectual, emotional and behabioral problems, and the listening guidance has been performed for their parents for 3 months through the listening guidance program. The changes in the communication skills in the hearing impaired children were observed comparing before and after the education. In the results, first, the receptive language skill of the hearing impaired children was improved after than before the parents listening guidance. Second, the expressive language skill of the hearing impaired children was improved after than before the parents listening guidance, too. Third, in the communication behavior of the hearing impaired children, the phonation and the speech production were increased together with the gesture after the parents listening guidance. In conclusion, it is deemed that the parents listening guidance program would have positive influence on the communication behavior of the hearing impaired children.

• Journal of Korea Multimedia Society
• /
• v.21 no.7
• /
• pp.745-754
• /
• 2018

Language is an essential tool for verbal and emotional communication among human beings, enabling them to engage in social interactions. Although a majority of hearing-impaired people can speak; however, they are unable to receive feedback on their pronunciation most of them can speak. However, they do not receive feedback on their pronunciation. This results in impaired communication owing to incorrect pronunciation, which causes difficulties in their social interactions. If hearing-impaired people could receive continuous feedback on their pronunciation and phonation through lip-reading training, they could communicate more effectively with people without hearing disabilities, anytime and anywhere, without the use of sign language. In this study, the mouth area is detected from videos of learners speaking monosyllabic words. The grayscale information of the detected mouth area is used to estimate a velocity vector using Optical Flow. This information is then quantified as feature values to classify vowels. Subsequently, a system is proposed that classifies monosyllables by algebraic computation of geometric feature values of lips using the characteristics of articulatory phonation. Additionally, the system provides feedback by evaluating the comparison between the information which is obtained from the sample categories and experimental results.

• Fire Science and Engineering
• /
• v.30 no.4
• /
• pp.135-140
• /
• 2016

To help emergency medical dispatchers receive rapid and accurate identification and corrective action status determination of an emergency call, and to support the automatic processing of a voice recognition system to the Korean emergency medical dispatch system, emergency call records were analyzed. Furthermore, a list of Korean symptoms expression were produced and the characteristics of the symptoms that appear on the actual wording of the telephone records were identified. This language list and its characteristics will be useful for training emergency medical dispatchers.