• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.34 no.3
• /
• pp.451-461
• /
• 2024

Recently, various AI chatbots based on large language models have been released. Chatbots have the advantage of providing users with quick and easy information through interactive prompts, making them useful in various fields such as question answering, writing, and programming. However, a vulnerability in chatbots called "prompt injection attacks" has been proposed. This attack involves injecting instructions into the chatbot to violate predefined guidelines. Such attacks can be critical as they may lead to the leakage of confidential information within large language models or trigger other malicious activities. However, the vulnerability of Korean prompts has not been adequately validated. Therefore, in this paper, we aim to generate malicious Korean prompts and perform attacks on the popular chatbot to analyze their feasibility. To achieve this, we propose a system that automatically generates malicious Korean prompts by analyzing existing prompt injection attacks. Specifically, we focus on generating malicious prompts that induce harmful expressions from large language models and validate their effectiveness in practice.

• Journal of Intelligence and Information Systems
• /
• v.24 no.4
• /
• pp.111-136
• /
• 2018

In this paper, we propose a methodology to extract answer information about queries from various types of unstructured documents collected from multi-sources existing on web in order to expand knowledge base. The proposed methodology is divided into the following steps. 1) Collect relevant documents from Wikipedia, Naver encyclopedia, and Naver news sources for "subject-predicate" separated queries and classify the proper documents. 2) Determine whether the sentence is suitable for extracting information and derive the confidence. 3) Based on the predicate feature, extract the information in the proper sentence and derive the overall confidence of the information extraction result. In order to evaluate the performance of the information extraction system, we selected 400 queries from the artificial intelligence speaker of SK-Telecom. Compared with the baseline model, it is confirmed that it shows higher performance index than the existing model. The contribution of this study is that we develop a sequence tagging model based on bi-directional LSTM-CRF using the predicate feature of the query, with this we developed a robust model that can maintain high recall performance even in various types of unstructured documents collected from multiple sources. The problem of information extraction for knowledge base extension should take into account heterogeneous characteristics of source-specific document types. The proposed methodology proved to extract information effectively from various types of unstructured documents compared to the baseline model. There is a limitation in previous research that the performance is poor when extracting information about the document type that is different from the training data. In addition, this study can prevent unnecessary information extraction attempts from the documents that do not include the answer information through the process for predicting the suitability of information extraction of documents and sentences before the information extraction step. It is meaningful that we provided a method that precision performance can be maintained even in actual web environment. The information extraction problem for the knowledge base expansion has the characteristic that it can not guarantee whether the document includes the correct answer because it is aimed at the unstructured document existing in the real web. When the question answering is performed on a real web, previous machine reading comprehension studies has a limitation that it shows a low level of precision because it frequently attempts to extract an answer even in a document in which there is no correct answer. The policy that predicts the suitability of document and sentence information extraction is meaningful in that it contributes to maintaining the performance of information extraction even in real web environment. The limitations of this study and future research directions are as follows. First, it is a problem related to data preprocessing. In this study, the unit of knowledge extraction is classified through the morphological analysis based on the open source Konlpy python package, and the information extraction result can be improperly performed because morphological analysis is not performed properly. To enhance the performance of information extraction results, it is necessary to develop an advanced morpheme analyzer. Second, it is a problem of entity ambiguity. The information extraction system of this study can not distinguish the same name that has different intention. If several people with the same name appear in the news, the system may not extract information about the intended query. In future research, it is necessary to take measures to identify the person with the same name. Third, it is a problem of evaluation query data. In this study, we selected 400 of user queries collected from SK Telecom 's interactive artificial intelligent speaker to evaluate the performance of the information extraction system. n this study, we developed evaluation data set using 800 documents (400 questions * 7 articles per question (1 Wikipedia, 3 Naver encyclopedia, 3 Naver news) by judging whether a correct answer is included or not. To ensure the external validity of the study, it is desirable to use more queries to determine the performance of the system. This is a costly activity that must be done manually. Future research needs to evaluate the system for more queries. It is also necessary to develop a Korean benchmark data set of information extraction system for queries from multi-source web documents to build an environment that can evaluate the results more objectively.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.191-206
• /
• 2022

Recently, it is a de-facto approach to utilize a pre-trained language model(PLM) to achieve the state-of-the-art performance for various natural language tasks(called downstream tasks) such as sentiment analysis and question answering. However, similar to any other machine learning method, PLM tends to depend on the data distribution seen during the training phase and shows worse performance on the unseen (Out-of-Distribution) domain. Due to the aforementioned reason, there have been many efforts to develop domain-specified PLM for various fields such as medical and legal industries. In this paper, we discuss the training of a finance domain-specified PLM for the Korean language and its applications. Our finance domain-specified PLM, KB-BERT, is trained on a carefully curated financial corpus that includes domain-specific documents such as financial reports. We provide extensive performance evaluation results on three natural language tasks, topic classification, sentiment analysis, and question answering. Compared to the state-of-the-art Korean PLM models such as KoELECTRA and KLUE-RoBERTa, KB-BERT shows comparable performance on general datasets based on common corpora like Wikipedia and news articles. Moreover, KB-BERT outperforms compared models on finance domain datasets that require finance-specific knowledge to solve given problems.

• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.754-758
• /
• 2009

Extracting information in large-scale documents would be very useful not only for information retrieval but also for question answering and summarization. Even though relation extraction is very important area, it is difficult to develop and evaluate a machine learning based system without test collection. The study shows how to build test collection(KREC2008) for the relation extraction system. We extracted technology terms from abstracts of journals and selected several relation candidates between them using Wordnet. Judges who were well trained in evaluation process assigned a relation from candidates. The process provides the method with which even non-experts are able to build test collection easily. KREC2008 are open to the public for researchers and developers and will be utilized for development and evaluation of relation extraction system.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.15 no.1
• /
• pp.56-61
• /
• 2022

In this paper, we classify and present AI algorithms and natural language processing methods used in chatbots. A framework that can be used to implement a chatbot is also described. A chatbot is a system with a structure that interprets the input string by constructing the user interface in a conversational manner and selects an appropriate answer to the input string from the learned data and outputs it. However, training is required to generate an appropriate set of answers to a question and hardware with considerable computational power is required. Therefore, there is a limit to the practice of not only developing companies but also students learning AI development. Currently, chatbots are replacing the existing traditional tasks, and a practice course to understand and implement the system is required. RNN and Char-CNN are used to increase the accuracy of answering questions by learning unstructured data by applying technologies such as deep learning beyond the level of responding only to standardized data. In order to implement a chatbot, it is necessary to understand such a theory. In addition, the students presented examples of implementation of the entire system by utilizing the methods that can be used for coding education and the platform where existing developers and students can implement chatbots.

• Journal of KIISE:Software and Applications
• /
• v.33 no.12
• /
• pp.1073-1089
• /
• 2006

The goal of Natural Language Processing(NLP) is to make a computer understand a natural language and to deliver the meanings of natural language to humans. Word sense Disambiguation(WSD is a very important technology to achieve the goal of NLP. In this paper, we describe a technology for automatic homonyms disambiguation using both Mutual Information(MI) and a Sense-Tagged Compound Noun Dictionary. Previous research work using word definitions in dictionary suffered from the problem of data sparseness because of the use of exact word matching. Our work overcomes this problem by using MI which is an association measure between words. To reflect language features, the rate of word-pairs with MI values, sense frequency and site of word definitions are used as weights in our system. We constructed a Sense-Tagged Compound Noun Dictionary for high frequency compound nouns and used it to resolve homonym sense disambiguation. Experimental data for testing and evaluating our system is constructed from QA(Question Answering) test data which consisted of about 200 query sentences and answer paragraphs. We performed 4 types of experiments. In case of being used only MI, the result of experiment showed a precision of 65.06%. When we used the weighted values, we achieved a precision of 85.35% and when we used the Sense-Tagged Compound Noun Dictionary, we achieved a precision of 88.82%, respectively.

• Journal of KIISE
• /
• v.44 no.8
• /
• pp.813-821
• /
• 2017

Entity Linking find the meaning of an entity mention, which indicate the entity using different expressions, in a user's query by linking the entity mention and the entity in the knowledge base. This task has four challenges, including the difficult knowledge base construction problem, multiple presentation of the entity mention, ambiguity of entity linking, and NIL entity recognition. In this paper, we first construct the entity name dictionary based on Wikipedia to build a knowledge base and solve the multiple presentation problem. We then propose various methods for NIL entity recognition and solve the ambiguity of entity linking by training the support vector machine based on several features, including the similarity of the context, semantic relevance, clue word score, named entity type similarity of the mansion, entity name matching score, and object popularity score. We sequentially use the proposed two methods based on the constructed knowledge base, to obtain the good performance in the entity linking. In the result of the experiment, our system achieved 83.66% and 90.81% F1 score, which is the performance of the NIL entity recognition to solve the ambiguity of the entity linking.

• Journal of Intelligence and Information Systems
• /
• v.29 no.4
• /
• pp.129-164
• /
• 2023

In a situation where the use and introduction of Large Language Models (LLMs) is expanding due to recent developments in generative AI technology, it is difficult to find actual application cases or implementation methods for the use of internal company data in existing studies. Accordingly, this study presents a method of implementing generative AI services using the LLM application architecture using the most widely used LangChain framework. To this end, we reviewed various ways to overcome the problem of lack of information, focusing on the use of LLM, and presented specific solutions. To this end, we analyze methods of fine-tuning or direct use of document information and look in detail at the main steps of information storage and retrieval methods using the retrieval augmented generation (RAG) model to solve these problems. In particular, similar context recommendation and Question-Answering (QA) systems were utilized as a method to store and search information in a vector store using the RAG model. In addition, the specific operation method, major implementation steps and cases, including implementation source and user interface were presented to enhance understanding of generative AI technology. This has meaning and value in enabling LLM to be actively utilized in implementing services within companies.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.4
• /
• pp.1400-1418
• /
• 2020

In the development of commercial promotion, chatbot is known as one of significant skill by application of natural language processing (NLP). Conventional design methods are using bag-of-words model (BOW) alone based on Google database and other online corpus. For one thing, in the bag-of-words model, the vectors are Irrelevant to one another. Even though this method is friendly to discrete features, it is not conducive to the machine to understand continuous statements due to the loss of the connection between words in the encoded word vector. For other thing, existing methods are used to test in state-of-the-art online corpus but it is hard to apply in real applications such as telemarketing data. In this paper, we propose an improved chatbot design way using hybrid bag-of-words model and skip-gram model based on the real telemarketing data. Specifically, we first collect the real data in the telemarketing field and perform data cleaning and data classification on the constructed corpus. Second, the word representation is adopted hybrid bag-of-words model and skip-gram model. The skip-gram model maps synonyms in the vicinity of vector space. The correlation between words is expressed, so the amount of information contained in the word vector is increased, making up for the shortcomings caused by using bag-of-words model alone. Third, we use the term frequency-inverse document frequency (TF-IDF) weighting method to improve the weight of key words, then output the final word expression. At last, the answer is produced using hybrid retrieval model and generate model. The retrieval model can accurately answer questions in the field. The generate model can supplement the question of answering the open domain, in which the answer to the final reply is completed by long-short term memory (LSTM) training and prediction. Experimental results show which the hybrid word vector expression model can improve the accuracy of the response and the whole system can communicate with humans.

• Journal of Intelligence and Information Systems
• /
• v.26 no.4
• /
• pp.1-25
• /
• 2020

Sentiment Analysis (SA) is a Natural Language Processing (NLP) task that analyzes the sentiments consumers or the public feel about an arbitrary object from written texts. Furthermore, Aspect-Based Sentiment Analysis (ABSA) is a fine-grained analysis of the sentiments towards each aspect of an object. Since having a more practical value in terms of business, ABSA is drawing attention from both academic and industrial organizations. When there is a review that says "The restaurant is expensive but the food is really fantastic", for example, the general SA evaluates the overall sentiment towards the 'restaurant' as 'positive', while ABSA identifies the restaurant's aspect 'price' as 'negative' and 'food' aspect as 'positive'. Thus, ABSA enables a more specific and effective marketing strategy. In order to perform ABSA, it is necessary to identify what are the aspect terms or aspect categories included in the text, and judge the sentiments towards them. Accordingly, there exist four main areas in ABSA; aspect term extraction, aspect category detection, Aspect Term Sentiment Classification (ATSC), and Aspect Category Sentiment Classification (ACSC). It is usually conducted by extracting aspect terms and then performing ATSC to analyze sentiments for the given aspect terms, or by extracting aspect categories and then performing ACSC to analyze sentiments for the given aspect category. Here, an aspect category is expressed in one or more aspect terms, or indirectly inferred by other words. In the preceding example sentence, 'price' and 'food' are both aspect categories, and the aspect category 'food' is expressed by the aspect term 'food' included in the review. If the review sentence includes 'pasta', 'steak', or 'grilled chicken special', these can all be aspect terms for the aspect category 'food'. As such, an aspect category referred to by one or more specific aspect terms is called an explicit aspect. On the other hand, the aspect category like 'price', which does not have any specific aspect terms but can be indirectly guessed with an emotional word 'expensive,' is called an implicit aspect. So far, the 'aspect category' has been used to avoid confusion about 'aspect term'. From now on, we will consider 'aspect category' and 'aspect' as the same concept and use the word 'aspect' more for convenience. And one thing to note is that ATSC analyzes the sentiment towards given aspect terms, so it deals only with explicit aspects, and ACSC treats not only explicit aspects but also implicit aspects. This study seeks to find answers to the following issues ignored in the previous studies when applying the BERT pre-trained language model to ACSC and derives superior ACSC models. First, is it more effective to reflect the output vector of tokens for aspect categories than to use only the final output vector of [CLS] token as a classification vector? Second, is there any performance difference between QA (Question Answering) and NLI (Natural Language Inference) types in the sentence-pair configuration of input data? Third, is there any performance difference according to the order of sentence including aspect category in the QA or NLI type sentence-pair configuration of input data? To achieve these research objectives, we implemented 12 ACSC models and conducted experiments on 4 English benchmark datasets. As a result, ACSC models that provide performance beyond the existing studies without expanding the training dataset were derived. In addition, it was found that it is more effective to reflect the output vector of the aspect category token than to use only the output vector for the [CLS] token as a classification vector. It was also found that QA type input generally provides better performance than NLI, and the order of the sentence with the aspect category in QA type is irrelevant with performance. There may be some differences depending on the characteristics of the dataset, but when using NLI type sentence-pair input, placing the sentence containing the aspect category second seems to provide better performance. The new methodology for designing the ACSC model used in this study could be similarly applied to other studies such as ATSC.