• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.115-119
• /
• 2018

다중 문서 제목 추출은 하나의 주제를 가지는 다중 문서에 대한 제목을 추출하는 것을 말한다. 일반적으로 다중 문서 제목 추출에서는 다중 문서 집합을 단일 문서로 본 다음 키워드를 제목 후보군으로 추출하고, 추출된 후보를 나열하는 형식의 연구가 많이 진행되어져 왔다. 하지만 이러한 방법은 크게 두 가지의 한계점을 가지고 있다. 먼저, 다중 문서를 단순히 하나의 문서로 보는 방법은 전체적인 주제를 반영한 제목을 추출하기 어렵다는 문제점이 있다. 다음으로, 키워드를 조합하는 형식의 방법은 키워드의 단위를 찾는 방법에 따라 추출된 제목이 자연스럽지 못하다는 한계점이 있다. 따라서 본 논문에서는 이 한계점들을 보완하기 위하여 단어 관련성 추정과 Byte Pair Encoding을 이용한 요약 기반의 다중 뉴스 기사 제목 추출 방법을 제안한다. 평가를 위해서는 자동으로 군집된 총 12개의 주제에 대한 다중 뉴스 기사 집합을 사용하였으며 전문 교육을 받은 연구원들이 정성평가를 진행하여 5점 만점 기준 평균 3.68점을 얻었다.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.126-131
• /
• 2018

cQA(Community-based Question Answering) 시스템은 사용자들이 질문을 남기고 답변을 작성하는 시스템이다. cQA는 사용자의 편의를 위해 기존의 축적된 질문을 검색하거나 카테고리로 분류하는 기능을 제공한다. 질문의 길이가 길 경우 검색이나 카테고리 분류의 정확도가 떨어지는 한계가 있는데, 이를 극복하기 위해 cQA 질문을 요약하는 모델을 구축할 필요가 있다. 하지만 이러한 모델을 구축하려면 대량의 요약 데이터를 확보해야 하는 어려움이 존재한다. 본 논문에서는 이러한 어려움을 극복하기 위해 cQA의 질문 제목, 본문으로 데이터를 확보하고 필터링을 통해 요약 데이터 셋을 만들었다. 또한 본문의 대표 단어를 이용하여 추상 요약을 하기 위해 딥러닝 기반의 Pointer-generator model을 사용하였다. 실험 결과, 기존의 추출 요약 방식보다 딥러닝 기반의 추상 요약 방식의 성능이 더 좋았으며 Pointer-generator model이 보다 좋은 성능을 보였다.

• Journal of KIISE:Software and Applications
• /
• v.31 no.4
• /
• pp.505-515
• /
• 2004

In a newspaper, 5W1H information is the most fundamental and important element for writing and understanding articles. Focusing on such a relation between a newspaper article and the 5W1H, we propose a summarization method based on the activation degree of 5W1H. To overcome problems of the lead-based and the title-based methods, both of which are known to be the most effective in newspaper summarization, sufficient 5W1H information is extracted from both a title and a lead sentence. Moreover, for each sentence, its weight is computed by considering various factors, such as activation degree of 5W1H, the number of 5W1H categories, and its length and position. These factors make a great contribution to the selection of more important sentences, and thus to the improvement of readability of the summarized texts. In an experimental evaluation, the proposed method achieved a precision of 74.7% outperforming the lead-based method. In sum, our 5W1H approach was shown to be promising for automatic summarization of newspaper articles.

• Annual Conference on Human and Language Technology
• /
• 2002.10e
• /
• pp.277-284
• /
• 2002

본 논문은 신문기사에서 중요한 문장을 추출(Extract)하는데 있어서, 기존에 기장 우수한 방법인 전문기반 방법(Lead-based method)과 제목을 이용한 유사도 측정방법(Title-based method)의 문제점을 해결하기 위해서, 육하원칙 활성화도를 이용하여 신문기사를 효과적으로 요약할 수 있는 방법과 알고리즘을 제안하였다. 본 연구에서는 먼저, 제목(Title)과 전문(Lead)에서 중복출현하지 않는 육하원칙 구성성분을 결합하고, 본문은 각 문장에서 육하원칙 구성성분의 재사용성과 육하원칙 구성성분의 범주 증감을 파악하여 육하원칙 활성화도를 구하고, 전문기반 방법을 응용하여 각 문장의 상대적인 중요도에 따라 최종적인 가중치를 부여함으로써, 신문기사에서 중요한 문장을 효과적으로 추출할 수 있는 가중치 계산식을 제안하였다. 실험문서는 조선일보 웹사이트에서 제공하는 신문기사 100건을 대상으로 하였으며, 요약율이 30%일 경우 제안한 방법의 정확률은 74.7%로 기존의 전문기반(Lead-based method)방법보다 6.7% 향상되었다.

• Smart Media Journal
• /
• v.10 no.1
• /
• pp.39-46
• /
• 2021

This study is about a method of extracting a summary from a news article in consideration of the importance of each sentence constituting the article. We propose a method of calculating sentence importance by extracting the probabilities of topic sentence, similarity with article title and other sentences, and sentence position as characteristics that affect sentence importance. At this time, a hypothesis is established that the Topic Sentence will have a characteristic distinct from the general sentence, and a deep learning-based classification model is trained to obtain a topic sentence probability value for the input sentence. Also, using the pre-learned ELMo language model, the similarity between sentences is calculated based on the sentence vector value reflecting the context information and extracted as sentence characteristics. The topic sentence classification performance of the LSTM and BERT models was 93% accurate, 96.22% recall, and 89.5% precision, resulting in high analysis results. As a result of calculating the importance of each sentence by combining the extracted sentence characteristics, it was confirmed that the performance of extracting the topic sentence was improved by about 10% compared to the existing TextRank algorithm.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.81-83
• /
• 2022

Understanding research trends is necessary to select research topics and explore related works. Most researchers search representative keywords of interesting domains or technologies to understand research trends. However some conferences in artificial intelligence or data mining fields recently publish hundreds to thousands of papers for each year. It makes difficult for researchers to understand research trend of interesting domains. In our paper, we propose an automatic technology keyphrase extraction method to support researcher to understand research trend for each conference or journal. Keyphrase extraction that extracts important terms or phrases from a text, is a fundamental technology for a natural language processing such as summarization or searching, etc. Previous keyphrase extraction technologies based on pretrained language model extract keyphrases from long texts so performances are degraded in short texts like titles of papers. In this paper, we propose a techonolgy keyphrase extraction model that is robust in short text and considers the importance of the word.

• The Journal of Korean Association of Computer Education
• /
• v.6 no.2
• /
• pp.23-31
• /
• 2003

In general, people use a key word or a phrase as the name of field or subject word in document. This paper has focused on keyword extraction. First of all, we investigate that an author suggests keywords that are not occurred as contents words in literature, and present generation rules to combine compound keywords based on concept of lexical information. Moreover, we present a new importance measurement to avoid useless keywords that are not related to documents' contents. To verify the validity of extraction result, we collect titles and abstracts from research papers about natural language and/or voice processing studies, and obtain the 96% precision in a top rank of extraction result.

• Smart Media Journal
• /
• v.8 no.2
• /
• pp.58-65
• /
• 2019

Automatic text summarization is a process of shortening a text document by either extraction or abstraction. The abstraction approach inspired by deep learning methods scaling to a large amount of document is applied in recent work. Abstractive text summarization involves utilizing pre-generated word embedding information. Low-frequent but salient words such as terminologies are seldom included to dictionaries, that are so called, out-of-vocabulary(OOV) problems. OOV deteriorates the performance of Encoder-Decoder model in neural network. In order to address OOV words in abstractive text summarization, we propose a copy mechanism to facilitate copying new words in the target document and generating summary sentences. Different from the previous studies, the proposed approach combines accurate pointing information and selective copy mechanism based on bidirectional RNN and bidirectional LSTM. In addition, neural network gate model to estimate the generation probability and the loss function to optimize the entire abstraction model has been applied. The dataset has been constructed from the collection of abstractions and titles of journal articles. Experimental results demonstrate that both ROUGE-1 (based on word recall) and ROUGE-L (employed longest common subsequence) of the proposed Encoding-Decoding model have been improved to 47.01 and 29.55, respectively.

• Journal of KIISE:Software and Applications
• /
• v.26 no.10
• /
• pp.1204-1218
• /
• 1999

문서영상 검색을 위한 디지털도서관의 대부분은 논문제목과/또는 논문요약으로부터 만들어진 색인에 근거한 제한적인 검색기능을 제공하고 있다. 본 논문에서는 영문 문서영상전체에 대한 검색을 위한 단어 영상 형태 특징기반의 단어검색시스템을 제안한다. 본 논문에서는 검색의 효율성과 정확도를 높이기 위해 1) 기존의 단어검색시스템에서 사용된 특징들을 조합하여 사용하며, 2) 특징의 개수 및 위치뿐만 아니라 특징들의 순서를 포함하여 매칭하는 방법을 사용하며, 3) 특징비교에 의해 검색결과를 얻은 후에 여과목적으로 문자인식을 부분적으로 적용하는 2단계의 검색방법을 사용한다. 제안된 시스템의 동작은 다음과 같다. 문서 영상이 주어지면, 문서 영상 구조가 분석되고 단어 영역들의 조합으로 분할된다. 단어 영상의 특징들이 추출되어 저장된다. 사용자의 텍스트 질의가 주어지면 이에 대응되는 단어 영상이 만들어지며 이로부터 영상특징이 추출된다. 이 참조 특징과 저장된 특징들과 비교하여 유사한 단어를 검색하게 된다. 제안된 시스템은 IBM-PC를 이용한 웹 환경에서 구축되었으며, 영문 문서영상을 이용하여 실험이 수행되었다. 실험결과는 본 논문에서 제안하는 방법들의 유효성을 보여주고 있다. Abstract Most existing digital libraries for document image retrieval provide a limited retrieval service due to their indexing from document titles and/or the content of document abstracts. This paper proposes a word spotting system for full English document image retrieval based on word image shape features. In order to improve not only the efficiency but also the precision of a retrieval system, we develop the system by 1) using a combination of the holistic features which have been used in the existing word spotting systems, 2) performing image matching by comparing the order of features in a word in addition to the number of features and their positions, and 3) adopting 2 stage retrieval strategies by obtaining retrieval results by image feature matching and applying OCR(Optical Charater Recognition) partly to the results for filtering purpose. The proposed system operates as follows: given a document image, its structure is analyzed and is segmented into a set of word regions. Then, word shape features are extracted and stored. Given a user's query with text, features are extracted after its corresponding word image is generated. This reference model is compared with the stored features to find out similar words. The proposed system is implemented with IBM-PC in a web environment and its experiments are performed with English document images. Experimental results show the effectiveness of the proposed methods.

• Journal of KIISE:Software and Applications
• /
• v.34 no.6
• /
• pp.539-548
• /
• 2007

Mobile content adaptation is a technology of effectively representing the contents originally built for the desktop PC on wireless mobile devices. Previous approaches for Web content adaptation are mostly device-dependent. Also, the content transformation to suit to a smaller device is done manually. Furthermore, the same contents are provided to different users regardless of their individual preferences. As a result, the user has difficulty in selecting relevant information from a heavy volume of contents since the context information related to the content is not provided. To resolve these problems, this paper proposes an enhanced method of Web content adaptation for mobile devices. In our system, the process of Web content adaptation consists of 4 stages including block filtering, block title extraction, block content summarization, and personalization through learning. Learning is initiated when the user selects the full content menu from the content summary page. As a result of learning, personalization is realized by showing the information for the relevant block at the top of the content list. A series of experiments are performed to evaluate the content adaptation for a number of Web sites including online newspapers. The results of evaluation are satisfactory, both in block filtering accuracy and in user satisfaction by personalization.