• The Journal of Information Technology and Database
• /
• v.8 no.1
• /
• pp.117-128
• /
• 2001

Recently there are many researches of text mining to find interesting patterns or association rules from mass textual documents. However, the words extracted from informal documents are tend to be irregular and there are too many general words, so if we use pre-exist method, we would have difficulty in retrieving knowledge information effectively. In this paper, we propose a new feature extraction method to classify mass documents using association rule based on unsupervised learning technique. In experiment, we show the efficiency of suggested method by extracting features and classifying of documents.

• International Journal of Contents
• /
• v.8 no.1
• /
• pp.39-46
• /
• 2012

Character segmentation is a preprocessing step in many offline handwriting recognition systems. In this paper, Chinese characters are categorized into seven different structures. In each structure, the character size with the range of variations is estimated considering typical handwritten samples. The component removal and merge criteria are presented to remove punctuation symbols or to merge small components which are part of a character. Finally, the criteria for segmenting the adjacent characters concerning each other or overlapped are proposed.

• Proceedings of the Korean Society of Disaster Information Conference
• /
• 2016.11a
• /
• pp.304-306
• /
• 2016

본 논문에서는 기계학습 방법에 기반을 둔 119 긴급 신고 전화 전사 데이터에 대한 구급, 구조, 화재 상황 분류 알고리즘을 개발하였다. 신고전화에서 빈번하게 발생하는 비정형 발화 패턴을 효율적으로 정규화하고 자연어 문장 처리 기법에서 일반적으로 사용하는 방법을 적용하여 신고전화 텍스트 데이터를 기계학습에서 사용할 수 있는 특징 벡터로 재구성하였다. 2743개의 신고전화에 대해 선형 서포트 벡터 머신을 이용하여 상황 분류를 수행한 결과, 92% 의 정확도를 얻을 수 있었다.

• Proceedings of the Korea Database Society Conference
• /
• 1998.09a
• /
• pp.421-430
• /
• 1998

지식경영을 위한 다양한 대상 업무중에서 텍스트 데이터의 마이닝은 특히 중요하다. 그 이유는 텍스트 데이터가 양적인 면에서 가장 풍부하고, 또 발견할 수 있는 지식을 가장 많이 포함하고 있기 때문이다. 본 논문에서는 텍스트 데이터베이스에서 지식발견을 위한 한 과정으로 텍스트 데이터베이스 내의 텍스트들을 분류하는 기법을 기술한다. 특히 문서 분류 방법은 데이터베이스의 일부 데이터를 훈련, 예제로 간주하여 교사 학습 알고리즘을 통해 학습한 후 나머지 데이터를 이용해 분류 정확성을 검증 및 향상시킨다. 시험 데이터로는 인터넷의 뉴스그룹의 기사를 이용하였고, 시험 결과 분류의 정확성은 한글 및 영문 모두 최소 70% 이상으로 나타났다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.167-169
• /
• 2000

문서 분류 시스템은 문서에 나타난 용어나 개념의 출현 정보를 이용한다. 개념 기반문서분류는 용어를 사용하지 않고 문서의 단어에 나타난 의미를 이용한다. 단어가 중의성을 가지는 경우 그 뜻을 정확히 가리지 않으면 문서에 출현하지 않은 의미를 이용하게 되므로 문서 분류 시스템의 성능이 저하된다. 본 논문은 개념 기반 문서분류를 위하여 단어 애매성 해소를 시도하였다. 문서에 출현된 의미 정보를 이용하여 의미들간의 공기정보를 구하고 이를 이용하여 단어의 애매성을 해소하였다. 단어의 의미정보는 시소러스 도구를 통해 획득하고 의미들간의 공기정보는 의미들간의 동시 출현 정보를 획득하여 구축하였다. 본 시스템은 문서 분류 등 자연어처리 분야에 이용할 수 있어 효용가치가 높다.

• Speech Sciences
• /
• v.9 no.1
• /
• pp.207-215
• /
• 2002

Perceptual evaluation of duration models of spoken Korean was carried out based on the Classification and Regression Tree (CART) model for text-to-speech conversion. A reference set of durations was produced by a commercial text-to-speech synthesis system for comparison. The duration model which was built in the previous research (Chung & Huckvale, 2001) was applied to a Korean language speech synthesis diphone database, 'Hanmal (HN 1.0)'. The synthetic speech produced by the CART duration model was preferred in the subjective preference test by a small margin and the synthetic speech from the commercial system was superior in the clarity test. In the course of preparing the experiment, a labeled database of spoken Korean with 670 sentences was constructed. As a result of the experiment, a trained duration model for speech synthesis was obtained. The 'Hanmal' diphone database for Korean speech synthesis was also developed as a by-product of the perceptual evaluation.

• Journal of Korean Library and Information Science Society
• /
• v.32 no.3
• /
• pp.307-330
• /
• 2001

This study experimented the performance of categorization methods using the kNN classifier. Most sample based automatic text categorization techniques like the kNN classifier reduces the feature set of the training documents. We sought to find out which percentage reductions in the feature set would result in high performances. In addition, the kNN classifier has to find the k number of training documents most similar to the test documents in the training documents. We sought to verify the most appropriate k value through experiments.

• Journal of Information Technology Services
• /
• v.17 no.2
• /
• pp.165-177
• /
• 2018

Recently, online commerces are becoming more common due to factors such as mobile technology development and smart device dissemination, and online review has a big influence on potential buyer's purchase decision. This study presents a set of analytical methodologies for understanding the meaning of customer reviews of products in online transaction. Using techniques currently developed in deep learning are implemented Hierarchical Attention Network for analyze meaning in online reviews. By using these techniques, we could solve time consuming pre-data analysis time problem and multiple topic problems. To this end, this study analyzes customer reviews of laptops sold in domestic online shopping malls. Our result successfully demonstrates over 90% classification accuracy. Therefore, this study classified the unstructured text data in the semantic analysis and confirmed the practical application possibility of the review analysis process.

• ETRI Journal
• /
• v.40 no.5
• /
• pp.643-652
• /
• 2018

Dysarthria is a degenerative disorder of the central nervous system that affects the control of articulation and pitch; therefore, it affects the uniqueness of sound produced by the speaker. Hence, dysarthric speaker recognition is a challenging task. In this paper, a feature-extraction method based on deep belief networks is presented for the task of identifying a speaker suffering from dysarthria. The effectiveness of the proposed method is demonstrated and compared with well-known Mel-frequency cepstral coefficient features. For classification purposes, the use of a multi-layer perceptron neural network is proposed with two structures. Our evaluations using the universal access speech database produced promising results and outperformed other baseline methods. In addition, speaker identification under both text-dependent and text-independent conditions are explored. The highest accuracy achieved using the proposed system is 97.3%.

• Journal of English Language & Literature
• /
• v.57 no.6
• /
• pp.911-938
• /
• 2011

The present study examined the effect of two variables of task complexity, reasoning demand and time pressure, each from the resourcedirecting and resource-dispersing dimension in Robinson's (2001) framework of task classification. Reasoning demand was operationalized as the two types of texts to read and summarize, expository and argumentative. Time pressure was operationalized as the two modes of performance, oral and written. Six university students summarized the two types of text orally and twenty four students from the same school summarized them in the written form. Results from t test and ANCOVA showed that in the oral mode, reasoning demand tends to heighten the complexity of the language used in the summary in competition with accuracy but such an effect disappeared in the written mode. It was interpreted that the degree of time pressure is not the only difference between the oral and written modes but that the two modes may be fundamentally different cognitive tasks, and that Robinson's (2001) and Skehan's (1998) models were differentially supported by the oral mode of tasks but not by the written mode of the tasks.