• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.87-92
• /
• 2019

본 논문에서는 기계 독해를 이용한 웹 기반 오픈 도메인 한국어 질의응답 시스템에 대하여 서술한다. 하나의 사용자 질의에 대하여, 본 논문에서 제안된 시스템은 기 존재하는 검색 엔진을 이용하여 실시간으로 최대 1,500 개의 문서를 기계 독해 방식으로 분석하고, 각 문서별로 얻어진 답을 종합함으로써 최종 답변을 도출한다. 실험 결과, 제안된 시스템은 평균적으로 2초 이내의 실행 시간을 보였으며, 사람과 비교하여 86%의 성능을 나타내었다. 본 논문에서 제안된 시스템의 데모는 http://nlp-api.kakao.com에서 확인 가능하다.

• Journal of KIISE:Computing Practices and Letters
• /
• v.9 no.5
• /
• pp.530-540
• /
• 2003

As the amount of information on Internet sites increases, it becomes more necessary to provide information in efficient ways. However, information search methods based on Boolean combination of keywords that most sites provide are difficult to express user's intention adequately so that there are numerous unexpected search results. This paper proposes a conversational agent that provides users with accurate information in a friendly manner through natural language conversation. The agent recognizes user's intention by applying finite state automata to natural language queries, utilizes the intention for structured pattern matching with response knowledge, and thus provides answers that are robust against changes in word order and consistent with the user's intention. To show its practical utility, the agent is applied to the problem of introducing a Web site. The results show that the conversational agent has the ability to provide accurate and friendly responses.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.243-244
• /
• 2021

대학교에 민원, 문의 등 업무용 전화가 많이 오가는 상황에서 사용자가 원하는 부서, 교직원의 전화번호를 알아내기 위해 직접 검색하는 과정에 대한 솔루션을 제공하기 위해 본 논문에서는 대학 교직원들의 전화번호와 부서의 정보를 저장하고 있는 CSV 파일을 챗봇 시스템에서 요구하는 모양과 특성에 맞게 데이터를 가공하고 알맞은 정보를 제공하기 위해 사용자의 질의 문장을 해체 분석하여 필요 정보에 대하여 답변을 해주는 대학 교직원 번호 안내 챗봇 시스템을 개발하였다.

• Journal of Intelligence and Information Systems
• /
• v.9 no.3
• /
• pp.99-115
• /
• 2003

In this paper, we present a knowledge-based object-oriented spatial database system called KOBOS. A knowledge-based approach is introduced to the object-oriented spatial database system for data modeling and approximate query answering. For handling the structure of spatial objects and the approximate spatial operators, we propose three levels of object-oriented data model: (1) a spatial shape model; (2) a spatial object model; (3) an internal description model. We use spatial type abstraction hierarchies(STAHs) to provide the range of the approximate spatial operators. We then propose SOQL, a spatial object-oriented query language. SOQL provides an integrated mechanism for the graphical display of spatial objects and the retrieval of spatial and aspatial objects. To support an efficient hybrid query evaluation, we use the top-down spatial query processing method.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.459-461
• /
• 2020

매일 게시되는 다양한 프로야구 관련 기사에는 경기 결과, 각종 기록, 선수의 부상 등 다양한 정보가 뒤섞여있어, 사용자가 원하는 정보를 찾아내는 과정이 매우 번거롭다. 본 논문에서는 문서 검색과 기계 독해를 이용하여 야구 분야에 대한 Q&A 시스템을 제안한다. 기사를 형태소 분석하고 BM25 알고리즘으로 얻은 문서 가중치로 사용자 질의에 적합한 기사들을 선정하고 KorQuAD 1.0과 직접 구축한 프로야구 질의응답 데이터셋을 이용해 학습시킨 BERT 모델 기반 기계 독해로 답변 추출을 진행한다. 야구 특화 데이터 셋을 추가하여 학습시켰을 때 F1 score, EM 모두 15% 내외의 정확도 향상을 보였다.

• Journal of Digital Convergence
• /
• v.15 no.3
• /
• pp.313-318
• /
• 2017

In general, users of design software such as Photoshop, search for information online when they want to obtain related knowledge. However, it is very difficult to find the exact information they want about designs because available knowledge sharing systems are very broad in what they manage, and it is rare that such systems would provide any design-specific Q&A or exchange of information functionality. In the paper we development a video based system for sharing design knowledge that supplies Q&A, lecture, knowledge trade function etc. utilizing multimedia like text, image and video reflecting the characteristics of design knowledge. The system are expected to contribute to the competitiveness of products through sharing design knowledge. In the near future the system will need to expand to a framework that can be shared with a variety of knowledge, as well as design knowledge.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.683-689
• /
• 2023

자연어 처리 분야는 최근에 큰 발전을 보였으며, 특히 초대규모 언어 모델의 등장은 이 분야에 큰 영향을 미쳤다. GPT와 같은 모델은 다양한 NLP 작업에서 높은 성능을 보이고 있으며, 특히 챗봇 분야에서 중요하게 다루어지고 있다. 하지만, 이러한 모델에도 여러 한계와 문제점이 있으며, 그 중 하나는 모델이 기대하지 않은 결과를 생성하는 것이다. 이를 해결하기 위한 다양한 방법 중, Retrieval-Augmented Generation(RAG) 방법이 주목받았다. 이 논문에서는 지식베이스와의 통합을 통한 도메인 특화형 질의응답 시스템의 효율성 개선 방안과 벡터 데이터 베이스의 수정을 통한 챗봇 답변 수정 및 업데이트 방안을 제안한다. 본 논문의 주요 기여는 다음과 같다: 1) QA Pair Passage RAG을 활용한 새로운 RAG 시스템 제안 및 성능 향상 분석 2) 기존의 LLM 및 RAG 시스템의 성능 측정 및 한계점 제시 3) RDBMS 기반의 벡터 검색 및 업데이트를 활용한 챗봇 제어 방법론 제안

• The Journal of the Korea Contents Association
• /
• v.19 no.8
• /
• pp.465-472
• /
• 2019

Although there are various cultural events and cultural contents produced in the region, there is a lack of distribution and spread of regional information to expand related economic consumption. This study combined local advertising by local advertisers with the knowledge search method in question and answer format from a location-based service perspective for the purpose of spreading and using local cultural information. The approach looked at domestic and international cases of knowledge search based on region and location-based advertising research, presented community model of location inquiry based information service and revenue model of local advertisement. Through this, this study designed a question and answer based community and operational structure model of local advertising, and developed an information service system in the form of prototyping. By extending the distribution of question and answer data among users to location information, it is meaningful that a business service model was presented that combines local cultural content information and the demand for user access with the revenue model of local advertising.

• Journal of KIISE
• /
• v.45 no.2
• /
• pp.134-140
• /
• 2018

A chat system is a computer program that understands user's miscellaneous utterances and generates appropriate responses. Sometimes a chat system needs to answer users' simple information-seeking questions. However, previous generative chat systems do not consider how to embed knowledge entities (i.e., subjects and objects in triple knowledge), essential elements for question-answering. The previous chat models have a disadvantage that they generate same responses although knowledge entities in users' utterances are changed. To alleviate this problem, we propose a knowledge entity embedding method for improving question-answering accuracies of a generative chat system. The proposed method uses a Siamese recurrent neural network for embedding knowledge entities and their synonyms. For experiments, we implemented a sequence-to-sequence model in which subjects and predicates are encoded and objects are decoded. The proposed embedding method showed 12.48% higher accuracies than the conventional embedding method based on a convolutional neural network.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.1
• /
• pp.27-36
• /
• 2019

In this paper, we propose a method to find the most similar answer to the user's response from the question-answer database in order to avoid generating a redundant question in retrieval-based automatic question generation system. As a question of the most similar answer to user's response may already be known to the user, the question should be removed from a set of question candidates. A similarity detector calculates a similarity between two answers by utilizing the same words, paraphrases, and sentential meanings. Paraphrases can be acquired by building a phrase table used in a statistical machine translation. A sentential meaning's similarity of two answers is calculated by an attention-based convolutional neural network. We evaluate the accuracy of the similarity detector on an evaluation set with 100 answers, and can get the 71% Mean Reciprocal Rank (MRR) score.