• Journal of KIISE
• /
• v.44 no.1
• /
• pp.51-56
• /
• 2017

The research area for solving question answering (QA) problems using artificial intelligence models is in a methodological transition period, and one such architecture, the dynamic memory network (DMN), is drawing attention for two key attributes: its attention mechanism defined by neural network operations and its modular architecture imitating cognition processes during QA of human. In this paper, we increased accuracy of the inferred answers, by adapting an automatic data augmentation method for lacking amount of training data, and by improving the ability of time perception. The experimental results showed that in the 1K-bAbI tasks, the modified DMN achieves 89.21% accuracy and passes twelve tasks which is 13.58% higher with passing four more tasks, as compared with one implementation of DMN. Additionally, DMN's word embedding vectors form strong clusters after training. Moreover, the number of episodic passes and that of supporting facts shows direct correlation, which affects the performance significantly.

• 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.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.7
• /
• pp.81-88
• /
• 2024

Despite the rapid growth of the generative AI market and significant interest from domestic companies and institutions, concerns about the provision of inaccurate information and potential information leaks have emerged as major factors hindering the adoption of generative AI. To address these issues, this paper designs and implements a question-answering system based on the Retrieval-Augmented Generation (RAG) architecture. The proposed method constructs a knowledge database using Korean sentence embeddings and retrieves information relevant to queries through optimized searches, which is then provided to the generative language model. Additionally, it allows users to directly manage the knowledge database to efficiently update changing business information, and it is designed to operate in a private network to reduce the risk of corporate confidential information leakage. This study aims to serve as a useful reference for domestic companies seeking to adopt and utilize generative AI.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.12
• /
• pp.485-496
• /
• 2018

So far, the deep learning, a field of artificial intelligence, has achieved remarkable results in solving problems from unstructured data. However, it is difficult to comprehensively judge situations like humans, and did not reach the level of intelligence that deduced their relations and predicted the next situation. Recently, deep neural networks show that artificial intelligence can possess powerful relational reasoning that is core intellectual ability of human being. In this paper, to analyze and observe the performance of Relation Networks (RN) among the neural networks for relational reasoning, two types of RN-based deep neural network models were constructed and compared with the baseline model. One is a visual question answering RN model using Sort-of-CLEVR and the other is a text-based question answering RN model using bAbI task. In order to maximize the performance of the RN-based model, various performance improvement experiments such as hyper parameters tuning have been proposed and performed. The effectiveness of the proposed performance improvement methods has been verified by applying to the visual QA RN model and the text-based QA RN model, and the new domain model using the dialogue-based LL dataset. As a result of the various experiments, it is found that the initial learning rate is a key factor in determining the performance of the model in both types of RN models. We have observed that the optimal initial learning rate setting found by the proposed random search method can improve the performance of the model up to 99.8%.

• Proceedings of the Korea Contents Association Conference
• /
• 2006.05a
• /
• pp.140-143
• /
• 2006

In this paper, we describe a question-answering system which extracts answers for the superlative questions which include the superlative words such as "the most", "the best", "the first", "the largest", "the least", and so on. The superlative questions are composed of four main components and others. Four main components are the superlative word, answer type, regional information, and a verb modified by the superlative word. We classify the superlative words into two types as to whether the verb has to be needed to be a question or not. The superlative word, answer type and regional information are essential elements to extract answer for all superlative questions. But the verb may be an essential element by the type of superlative word. Our system analyzes input question, and finds four main components of the superlative question. Also, our system searches relative documents and candidate sentences using them, and extracts answers from candidate sentences. Empirical result shows that our system has high precision and high recall for the superlative questions.

• Journal of KIISE:Computing Practices and Letters
• /
• v.10 no.6
• /
• pp.529-536
• /
• 2004

The main idea of this paper is to combine position information in sentence and query type classification to make the documents ranking to query more accessible. First, the use of conceptual graphs for the representation of document contents In information retrieval is discussed. The method is based on well-known strategies of text comparison, such as Dice Coefficient, with position-based weighted term. Second, we introduce a method for learning query type classification that improves the ability to retrieve answers to questions from Question Answering system. Proposed methods employ naive bayes classification in machine learning fields. And, we used a collection of approximately 30,000 question-answer pairs for training, obtained from Frequently Asked Question(FAQ) files on various subjects. The evaluation on a set of queries from international TREC-9 question answering track shows that the method with machine learning outperforms the underline other systems in TREC-9 (0.29 for mean reciprocal rank and 55.1% for precision).

• International Journal of Contents
• /
• v.1 no.2
• /
• pp.39-43
• /
• 2005

In this paper, we describe a generation mechanism of semantic-based queries for high accuracy information retrieval and question answering. It is difficult to offer the correct retrieval result because general information retrieval systems do not analyze the semantic of user's natural language question. We analyze user's question semantically and extract semantic features, and we .generate semantic-based queries using them. These queries are generated using the se-mantic-based question analysis grammar and the query generation rule. They are represented as semantic features and grammatical morphemes that consider semantic and syntactic structure of user's questions. We evaluated our mechanism using 100 questions whose answer type is a person in the TREC-9 corpus and Web. There was a 0.28 improvement in the precision at 10 documents when semantic-based queries were used for information retrieval.

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.04b
• /
• pp.931-933
• /
• 2004

본 논문에서는 한국어 질의응답시스템에서 정확한 정답추출을 위한 구문 정보에 기반한 질의분석을 제안한다. 질의분석은 세부 정답 유형 결정, 세분화된 키워드 추출을 통해 정확한 정답추출을 목적으로 한다. 술어 유형 정보를 이용하여 대분류 수준의 정답 유형으로 질의분석을 수행하고. 구문 구조 정보를 이용하여 중요 키워드와 일반 키워드를 추출한다 마지막으로 정답 유형 자질 명사를 이용하여 세부 정답 유형을 결정한다. 실험을 통해 세부 정답 유형 결정에서 정확률 59%, 세분화된 키워드 추출에서 정확을 66%를 보였다.

• International Journal of Computer Science & Network Security
• /
• v.21 no.9
• /
• pp.281-291
• /
• 2021

The Islamic jurisprudential legal system represents an essential component of the Islamic religion, that governs many aspects of Muslims' daily lives. This creates many questions that require interpretations by qualified specialists, or Muftis according to the main sources of legislation in Islam. The Islamic jurisprudence is usually classified into branches, according to which the questions can be categorized and classified. Such categorization has many applications in automated question-answering systems, and in manual systems in routing the questions to a specialized Mufti to answer specific topics. In this work we tackle the problem of automatic categorisation of Islamic jurisprudential legal questions using deep learning techniques. In this paper, we build a hierarchical deep learning model that first extracts the question text features at two levels: word and sentence representation, followed by a text classifier that acts upon the question representation. To evaluate our model, we build and release the largest publicly available dataset of Islamic questions and answers, along with their topics, for 52 topic categories. We evaluate different state-of-the art deep learning models, both for word and sentence embeddings, comparing recurrent and transformer-based techniques, and performing extensive ablation studies to show the effect of each model choice. Our hierarchical model is based on pre-trained models, taking advantage of the recent advancement of transfer learning techniques, focused on Arabic language.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.253-255
• /
• 2000

어떠한 질문에 대한 구체적 해답을 얻고 싶은 경우, 일반적인 정보 검색이 가지는 문제점은 검색 결과가 사용자가 찾고자 하는 답이라 하기 보다는 해답을 포함하는(또는 포함하지 않는) 문서의 집합이라는 점이다. 사용자가 후보문서를 모두 읽을 필요 없이 빠르게 원하는 정보를 얻기 위해서는 검색의 결과로 문서집합을 제시하기 보다는 실제 원하는 답을 제공하는 시스템의 필요성이 대두된다. 이를 위해 기존의 TF-IDF(Term Frequency-Inversed Document Frequency)기반의 정보검색의 방삭에 자연언어처리(Natural Language Processing)를 이용한 질문의 분류와 문서의 사전 표지(Tagging)를 사용할 수 있다. 본 연구에서는 매년 NIST(National Institute of Standards & Technology)와 DARPA(Defense Advanced Research Projects Agency)주관으로 열리는 TREC(Text REtrieval Conference)중 1999년에 열린 TREC-8의 사용자의 질문(Question)에 대한 답(Answer)을 찾는 ‘Question & Answer’문제의 실험 환경에서 질문을 특징별로 분류하고 검색 대상의 문서에 대한 사전 표지를 이용한 정보검색 시스템으로 사용자의 질문(Question)에 대한 해답을 보다 정확하고 효율적으로 제시할 수 있음을 실험을 통하여 보인다.