• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.269-274
• /
• 2023

질의응답 (Question Answering)은 주어진 질문을 이해하여 그에 맞는 답변을 생성하는 자연어 처리 분야의 핵심적인 기계 독해 작업이다. 현재 대다수의 자연어 이해 작업은 사전학습 언어 모델에 미세 조정 (finetuning)하는 방식으로 학습되고, 질의응답 역시 이러한 방법으로 진행된다. 하지만 미세 조정을 통한 전이학습은 사전학습 모델의 크기가 커질수록 전이학습이 잘 이루어지지 않는다는 단점이 있다. 게다가 많은 양의 파라미터를 갱신한 후 새로운 가중치들을 저장하여야 한다는 용량의 부담이 존재한다. 본 연구는 최근 대두되는 deep prompt tuning 방법론을 한국어 추출형 질의응답에 적용하여, 미세 조정에 비해 학습시간을 단축시키고 적은 양의 파라미터를 활용하여 성능을 개선했다. 또한 한국어 추출형 질의응답에 최적의 prompt 길이를 최적화하였으며 오류 분석을 통한 정성적인 평가로 deep prompt tuning이 모델 예측에 미치는 영향을 조사하였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.164-167
• /
• 2022

In table Machine comprehension, the knowledge required for language models or the structural form of tables changes depending on the domain, showing a greater performance degradation compared to text data. In this paper, we propose a pre-learning data construction method and an adversarial learning method through meaningful tabular data selection for constructing a pre-learning table language model robust to these domain changes in table machine reading. In order to detect tabular data sed for decoration of web documents without structural information from the extracted table data, a rule through heuristic was defined to identify head data and select table data was applied. An adversarial learning method between tabular data and infobax data with knowledge information about entities was applied. When the data was refined compared to when it was trained with the existing unrefined data, F1 3.45 and EM 4.14 increased in the KorQuAD table data, and F1 19.38, EM 4.22 compared to when the data was not refined in the Spec table QA data showed increased performance.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.3
• /
• pp.85-90
• /
• 2021

In this paper, fall detection using individual wearable devices for older people is considered. To design a low-cost wearable device for reliable fall detection, we present a comprehensive analysis of two representative models. One is a machine learning model composed of a decision tree, random forest, and Support Vector Machine(SVM). The other is a deep learning model relying on a one-dimensional(1D) Convolutional Neural Network(CNN). By considering data segmentation, preprocessing, and feature extraction methods applied to the input data, we also evaluate the considered models' validity. Simulation results verify the efficacy of the deep learning model showing improved overall performance.

• Journal of Korea Multimedia Society
• /
• v.17 no.2
• /
• pp.232-239
• /
• 2014

As online texts have been rapidly growing, their automatic classification gains more interest with machine learning methods. Nevertheless, comparatively few research could be found, aiming for Korean texts. Evaluating them with statistical methods are also rare. This study took a sample of tweets and used machine learning methods to classify emotions with features of morphemes and n-grams. As a result, about 76% of emotions contained in tweets was correctly classified. Of the two methods compared in this study, Support Vector Machines were found more accurate than Na$\ddot{i}$ve Bayes. The linear model of SVM was not inferior to the non-linear one. Morphological features did not contribute to accuracy more than did the n-grams.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.3
• /
• pp.361-370
• /
• 2017

In this paper, correlation analysis was performed between questionnaire and machine learning based aggressive tendency measurements. this study is part of a aggressive driver detection using machine learning and questionnaire. To collect two types tendency from questionnaire and measurements system, we constructed experiments environments and acquired the data from 30 drivers. In experiment, the machine learning based aggressive tendency measurements system was designed using a driver behavior detection model. And the model was constructed using accelerate and brake position data and hidden markov model method through supervised learning. We performed a correlation analysis between two types tendency using Pearson method. The result was represented to high correlation. The results will be utilize for fusing questionnaire and machine learning. Furthermore, It is verified that the machine learning based aggressive tendency is unique to each driver. The aggressive tendency of driver will be utilized as measurements for advanced driver assistance system such as attention assist, driver identification and anti-theft system.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.27 no.6
• /
• pp.1385-1395
• /
• 2017

This paper aims to develop an IDS (Intrusion Detection System) that takes into account class imbalanced datasets. For this, we first built a set of training data sets from the Kyoto 2006+ dataset in which the amounts of normal data and abnormal (intrusion) data are not balanced. Then, we have run a number of tests to evaluate the effectiveness of machine learning techniques for detecting intrusions. Our evaluation results demonstrated that the Random Forest algorithm achieved the best performances.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.8 no.3
• /
• pp.57-67
• /
• 2018

The ensemble is a unified approach used for getting better performance by using multiple algorithms in machine learning. In this paper, we introduce boosting and bagging, which have been widely used in ensemble techniques, and design a method using support vector regression, radial basis function network, Gaussian process, and multilayer perceptron. In addition, our experiment was performed by adding a recurrent neural network and MOHID numerical model. The drifter data used for our experimental verification consist of 683 observations in seven regions. The performance of our ensemble technique is verified by comparison with four algorithms each. As verification, mean absolute error was adapted. The presented methods are based on ensemble models using bagging, boosting, and machine learning. The error rate was calculated by assigning the equal weight value and different weight value to each unit model in ensemble. The ensemble model using machine learning showed 61.7% improvement compared to the average of four machine learning technique.

• Journal of Convergence for Information Technology
• /
• v.8 no.1
• /
• pp.201-206
• /
• 2018

Recently, as technologies for realizing artificial intelligence have become more common, machine learning is widely used. Machine learning provides insight into collecting large amounts of data, batch processing, and taking final action, but the effects of the work are not immediately integrated into the learning process. In this paper proposed an adaptive learning model to improve the performance of real-time stream analysis as a big business issue. Adaptive learning generates the ensemble by adapting to the complexity of the data set, and the algorithm uses the data needed to determine the optimal data point to sample. In an experiment for six standard data sets, the adaptive learning model outperformed the simple machine learning model for classification at the learning time and accuracy. In particular, the support vector machine showed excellent performance at the end of all ensembles. Adaptive learning is expected to be applicable to a wide range of problems that need to be adaptively updated in the inference of changes in various parameters over time.

• Annual Conference of KIPS
• /
• 2013.11a
• /
• pp.751-753
• /
• 2013

안드로이드 플랫폼은 타 모바일 플랫폼보다 보안에 있어서 더 많은 취약점을 안고 있다. 따라서 현재 발생하고 있는 대부분의 모바일 악성코드는 안드로이드 플랫폼에서 발생하고 있다. 현재 악성코드 탐지 기법 중 기계학습을 도입한 방법은 변종 악성코드의 대처에 유연하다. 하지만 기계학습기법은 불필요한 Feature를 학습데이터로 사용할 경우, 오버피팅이 발생하여 전체적인 성능을 저하시킬 수 있다. 본 논문에서는 안드로이드 플랫폼에서 발생하는 리소스를 모니터링하여 Feature vector를 생성하고, Feature-selection 알고리즘을 통하여 Feature의 수에 따라 기계학습 Classifier를 통한 악성코드 탐지의 성능지표를 보인다. 이를 통하여, 기계학습을 통한 악성코드 탐지에서 Feature-selection의 필요성과 중요성을 설명한다.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.54-59
• /
• 2020

사실 불일치 교정은 기계 요약 시스템이 요약한 결과를 실제 사실과 일치하도록 만드는 작업이다. 실제 요약 생성연구에서 가장 공통적인 문제점은 요약을 생성할 때 잘못된 사실을 생성하는 것이다. 이는 요약 모델이 실제 서비스로 상용화 하는데 큰 걸림돌이 되는 부분 중 하나이다. 본 논문에서는 원문으로부터 개체명을 가져와 사실과 일치하는 문장으로 고치는 방법을 제안한다. 이를 위해서 언어 모델이 개체명에 대한 문맥적 표현을 잘 생성할 수 있도록 학습시킨다. 그리고 학습된 모델을 이용하여 원문과 요약문에 등장한 개체명들의 문맥적 표현 비교를 통해 적절한 단어로 교체함으로써 요약문의 사실 불일치를 해소한다. 제안 모델을 평가하기 위해 추상 요약 데이터를 이용해 학습데이터를 만들어 학습하고, 실제 시나리오에서 적용가능성을 검증하기 위해 모델이 요약한 요약문을 이용해 실험을 수행했다. 실험 결과, 자동 평가와 사람 평가에서 제안 모델이 비교 모델보다 높은 성능을 보여주었다.