• Annual Conference on Human and Language Technology
• /
• 2016.10a
• /
• pp.220-225
• /
• 2016

본 논문에서는 격틀과 워드 임베딩을 활용한 유사도 기반 대화 모델링을 제안한다. 기존의 유사도 기반 대화 모델링 방법은 형태소, 형태소 표지, 개체명, 토픽 자질, 핵심단어 등을 대화 말뭉치에서 추출하여 BOW(Bag Of Words) 자질로 사용하였기 때문에 입력된 사용자 발화에 포함된 단어들의 주어, 목적어와 같은 문장성분들의 위치적 역할을 반영할 수 가 없다. 또한, 의미적으로 유사하지만 다른 형태소를 가지는 문장 성분들의 경우 유사도 계산에 반영되지 않는 형태소 불일치 문제가 존재한다. 이러한 문제점을 해결하기 위해서, 위치적 정보를 반영하기 위한 문장성분 기반의 격틀과 형태소 불일치 문제를 해결하기 위한 워드임베딩을 활용하여 개선된 유사도 기반 대화 모델링을 제안한다. 개선된 유사도 기반 대화 모델링은 MRR 성능 약 92%의 성능을 나타낸다.

• Annual Conference on Human and Language Technology
• /
• 2002.10e
• /
• pp.300-305
• /
• 2002

자동 문서 요약 분야에서 대상 문서를 컴퓨터가 이해할 수 있는 형태로 어떻게 파악하고 구조화할 것인가는 중요한 이슈가 되어 왔다. 문서에 출현한 단어들은 Bag of Words 가정처럼 서로 독립적으로 존재하는 것이 아니라 문서가 쓰여진 의도에 따라 서로 간의 의미적, 혹은 지시적으로 연관되어 있다. 이러한 단어간의 연관성은 결속성(cohesion)이라고 표현하며, 이를 이용한 자동 방법으로 Barzilay의 어휘 체인(lexical chain)을 사용한 자동 방법이 대표적이다. 본 연구에서는 단어간의 연관성과 영문 시소러스인 워드넷(wordnet)에서 단어의 위치 정보를 사용하여 어휘 체인의 성능을 개선하였고, 대상 문서의 개념을 어휘 체인에 기반해 표현하여 자동의 성능을 개선하는 방안을 제시한다.

• Annual Conference on Human and Language Technology
• /
• 2015.10a
• /
• pp.104-108
• /
• 2015

많은 양의 뉴스가 생성됨에 따라 이를 효과적으로 정리하는 기법이 최근 활발히 연구되어왔다. 그 중 뉴스클러스터링은 두 뉴스가 동일사건을 다루는지를 판정하는 분류기의 성능에 의존적인데, 대부분의 경우 BoW(Bag-of-Words)기반 벡터유사도를 사용하고 있다. 본 논문에서는 BoW기반의 벡터유사도 뿐 아니라 두 문서에 포함된 사진들의 유사성 및 주제의 관련성을 측정, 이를 분류기의 자질로 추가하여 두 뉴스가 동일사건을 다루는지 판정하는 분류기의 성능을 개선하는 방법을 제안한다. 사진들의 유사성 및 주제의 관련성은 최근 각광을 받는 딥러닝기반 CNN과 신경망기반 문서임베딩을 통해 측정하였다. 실험결과 기존의 BoW기반 벡터유사도에 의한 분류기의 성능에 비해 제안하는 두 자질을 사용하였을 경우 3.4%의 성능 향상을 보여주었다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.596-599
• /
• 2023

본 연구는 브랜드가 시장 트렌드를 파악하고 이를 활용하여 경쟁 우위를 확보하고 성장하는 방법을 탐구하고 있다. 이를 위해 세 가지 핵심 요소를 고려하였다. 첫째, 시장의 트렌드 정보를 파악하기 위해 검색 포털 사이트의 검색어 랭킹 정보를 활용하였다. 둘째, 브랜드 상품과 트렌드의 연관성을 분석하기 위해 상품 타이틀과 리뷰 데이터를 활용하였다. 셋째, 각 상품의 브랜드 중요성을 추정하기 위해 리뷰 수, 리뷰 길이, 표현의 다양성 등을 고려했다. 연구 결과, 브랜드는 시장 트렌드를 더욱 정확하게 이해하고 파악함으로써 경쟁 우위를 확보하고 성장할 수 있는 기회를 제공함을 확인하였다. 더불어, 이를 통해 브랜드는 소비자의 요구를 더욱 효과적으로 충족시키고 고객 경험을 개선하는데 기여할 수 있을 것으로 기대된다.

• The KIPS Transactions:PartB
• /
• v.18B no.6
• /
• pp.365-374
• /
• 2011

Recently, malicious video classification and filtering techniques are of practical interest as ones can easily access to malicious multimedia contents through the Internet, IPTV, online social network, and etc. Considerable research efforts have been made to developing malicious video classification and filtering systems. However, the malicious video classification and filtering is not still being from mature in terms of reliable classification/filtering performance. In particular, the most of conventional approaches have been limited to using only the spatial features (such as a ratio of skin regions and bag of visual words) for the purpose of malicious image classification. Hence, previous approaches have been restricted to achieving acceptable classification and filtering performance. In order to overcome the aforementioned limitation, we propose new malicious video classification framework that takes advantage of using both the spatial and temporal features that are readily extracted from a sequence of video frames. In particular, we develop the effective temporal features based on the motion periodicity feature and temporal correlation. In addition, to exploit the best data fusion approach aiming to combine the spatial and temporal features, the representative data fusion approaches are applied to the proposed framework. To demonstrate the effectiveness of our method, we collect 200 sexual intercourse videos and 200 non-sexual intercourse videos. Experimental results show that the proposed method increases 3.75% (from 92.25% to 96%) for classification of sexual intercourse video in terms of accuracy. Further, based on our experimental results, feature-level fusion approach (for fusing spatial and temporal features) is found to achieve the best classification accuracy.

• Journal of Digital Convergence
• /
• v.17 no.3
• /
• pp.107-117
• /
• 2019

In the era of SNS, many people relies on it to express their emotions about various kinds of products and services. Therefore, for the companies eagerly seeking to investigate how their products and services are perceived in the market, emotion mining tasks using dataset from SNSs become important much more than ever. Basically, emotion mining is a branch of sentiment analysis which is based on BOW (bag-of-words) and TF-IDF. However, there are few studies on the emotion mining which adopt feature selection (FS) methods to look for optimal set of features ensuring better results. In this sense, this study aims to propose FS methods to conduct emotion mining tasks more effectively with better outcomes. This study uses Twitter and SemEval2007 dataset for the sake of emotion mining experiments. We applied three FS methods such as CFS (Correlation based FS), IG (Information Gain), and ReliefF. Emotion mining results were obtained from applying the selected features to nine classifiers. When applying DT (decision tree) to Tweet dataset, accuracy increases with CFS, IG, and ReliefF methods. When applying LR (logistic regression) to SemEval2007 dataset, accuracy increases with ReliefF method.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.9
• /
• pp.359-366
• /
• 2021

A number of construction companies in Korea invest considerable human and financial resources to construct a system for managing apartment defect data and for categorizing repair tasks. Thus, this study proposes machine learning models to automatically classify defect complaint text-data into one of the sub categories of 'finishing work' (i.e., one of the defect repair tasks). In the proposed models, we employed two word representation methods (Bag-of-words, Term Frequency-Inverse Document Frequency (TF-IDF)) and two machine learning classifiers (Support Vector Machine, Random Forest). In particular, we conducted both binary- and multi- classification tasks to classify 9 sub categories of finishing work: home appliance installation work, paperwork, painting work, plastering work, interior masonry work, plaster finishing work, indoor furniture installation work, kitchen facility installation work, and tiling work. The machine learning classifiers using the TF-IDF representation method and Random Forest classification achieved more than 90% accuracy, precision, recall, and F1 score. We shed light on the possibility of constructing automated defect classification systems based on the proposed machine learning models.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.45-52
• /
• 2012

This paper proposes a new enhancing document clustering method using the important terms of cluster and the wikipedia. The proposed method can well represent the concept of cluster topics by means of selecting the important terms in cluster by the semantic features of NMF. It can solve the problem of "bags of words" to be not considered the meaningful relationships between documents and clusters, which expands the important terms of cluster by using of the synonyms of wikipedia. Also, it can improve the quality of document clustering which uses the expanded cluster important terms to refine the initial cluster by re-clustering. The experimental results demonstrate that the proposed method achieves better performance than other document clustering methods.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.5
• /
• pp.30-38
• /
• 2011

This paper proposes a new enhancing document clustering method using a synonym of cluster topic and the similarity. The proposed method can well represent the inherent structure of document cluster set by means of selecting terms of cluster topic based on the semantic features by NMF. It can solve the problem of "bags of words" by using of expanding the terms of cluster topics which uses the synonyms of WordNet. Also, it can improve the quality of document clustering which uses the cosine similarity between the expanded cluster topic terms and document set to well cluster document with respect to the appropriation cluster. The experimental results demonstrate that the proposed method achieves better performance than other document clustering methods.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.12 no.5
• /
• pp.243-253
• /
• 2012

An opinion mining technique which was developed from document classification in area of data mining now becomes a common interest in domestic as well as international industries. The core of opinion mining is to decide precisely whether an opinion document is a positive or negative one. Although many related approaches have been previously proposed, a classification accuracy was not satisfiable enough to applying them in practical applications. A opinion documents written in Korean are not easy to determine a polarity automatically because they often include various and ungrammatical words in expressing subjective opinions. Proposed in this paper is a new approach of classification of opinion documents, which considers only a frequency of word patterns and excludes the grammatical factors as much as possible. In proposed method, we express a document into a bag of words and then apply a learning algorithm using a frequency of word patterns, and finally decide the polarity of the document using a score function. Additionally, we also present the experiment results for evaluating the accuracy of the proposed method.