2011 IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), 2011
AbstractUbiquitous computing technologies have been developed fast and various decision support ... more AbstractUbiquitous computing technologies have been developed fast and various decision support systems were proposed. Consecutively, in recently, people are working on an implicit service agent of mobile phone to make people to be provided useful services without ...
In this paper, we propose a new research trend analysis using important word clusters and its rel... more In this paper, we propose a new research trend analysis using important word clusters and its relationship. Journals published many papers every month or week and new scientific contributions were exponentially cumulated to their database. If can analysis important words and related relationships of the papers, a change of research trend in a domain is an interesting topic in text mining. We use a Term Frequency Inverse Document Frequency (TFIDF) to extract meaningful words, the similarity of words measures using WordNet information and a document comparison approach. To measure the similarity from word lists extracted by TFIDF and differences of important word clusters and weights, the approach analyzes the research trend and visualizes the differences of research interest in same research fields. To show usefulness of proposed approach, we illustrate simulations and various results.
This research deals with an issue of preventive medicine in bioinformatics. We can diagnose liver... more This research deals with an issue of preventive medicine in bioinformatics. We can diagnose liver conditions reasonably well to prevent Liver Cirrhosis by classifying liver disorder patients into fatty liver and high risk groups. The classification proceeds in two steps. Classification rules are first built by clustering five attributes (MCV, ALP, ALT, ASP, and GGT) of blood test dataset provided by the UCI Repository. The clusters can be formed by the K-mean method that analyzes multi dimensional attributes. We analyze the properties of each cluster divided into fatty liver, high risk and normal classes. The classification rules are generated by the analysis. In this paper, we suggest a method to diagnosis and predict liver condition to alcoholic patient according to risk levels using the classification rule from the new results of blood test. The K-mean classifier has been found to be more accurate for the result of blood test and provides the risk of fatty liver to normal liver c...
2017 IEEE International Conference on Big Data and Smart Computing (BigComp), 2017
Deep neural networks (DNNs), which show outstanding performance in various areas, consume conside... more Deep neural networks (DNNs), which show outstanding performance in various areas, consume considerable amounts of memory and time during training. Our research led us to propose a controlled dropout technique with the potential of reducing the memory space and training time of DNNs. Dropout is a popular algorithm that solves the overfitting problem of DNNs by randomly dropping units in the training process. The proposed controlled dropout intentionally chooses which units to drop compared to conventional dropout, thereby possibly facilitating a reduction in training time and memory usage. In this paper, we focus on validating whether controlled dropout can replace the traditional dropout technique to enable us to further our research aimed at improving the training speed and memory efficiency. A performance comparison between controlled dropout and traditional dropout is carried out by implementing an image classification experiment on data comprising handwritten digits from the MNI...
2016 International Conference on Big Data and Smart Computing (BigComp), 2016
In Korea, authors of the newspaper article tend to express their intention indirectly, that is, t... more In Korea, authors of the newspaper article tend to express their intention indirectly, that is, they choose a method to leave out some important facts, or sometimes uses biased terms to support their opinion. Since they're not expressing their opinion directly, detecting the political bias is a difficult task. In this paper, we propose a method to detect political bias in the Korean articles by first building word vectors and sentence vectors, and second do a DBN-Training with those vectors and finally do a regression with SVM to calculate the bias. We used our own dataset which is scored with the political bias before doing the regression.
2019 IEEE International Conference on Big Data and Smart Computing (BigComp), 2019
This paper introduce a question understanding system to respond appropriate answers in a dialog s... more This paper introduce a question understanding system to respond appropriate answers in a dialog system for banking services. The question understanding system provides an automated response service in a specific domain (e.g. banking). This can increase response rate of a customer counseling service, and improve business efficiency and expertise. The question understanding system classify domains, specific categories, and speech acts of questions. Finally, the system analyze meanings and intents of the questions, and searching correct answers even various input sentences. In this paper, we describe methods of keyword tokenizing, pattern recognition, sentence embedding, analyzing dialogue intention, and searching similar FAQs. Through these methods, we have developed the question understanding unit in a real interactive system for financial services for real insurance companies and banks, and analyze the usefulness of the system through practical system implementation examples.
2017 IEEE International Conference on Big Data and Smart Computing (BigComp), 2017
Sentence similarity measurement is an important technology that can be apply to various applicati... more Sentence similarity measurement is an important technology that can be apply to various applications in the natural language processing. Recently, an encoder-decoder model using recurrent neural network (RNN) has achieved remarkable results. This paper proposes a model for measuring Korean sentence similarity based on sense-based morpheme embedding and gated recurrent units (GRU) encoder. We evaluate the measurement model consist of experimentally optimized morpheme embedding and sentence encoding models. In the measurement of sentence similarity, the proposed model encoded using the pre-trained morpheme embedding improves the performance compared with the character-embedding model. In addition, it can be used effectively in the question and answering (Q&A) system.
For dialogue systems, it is critical to detect the out-of-domain (OOD) utterances in a conversati... more For dialogue systems, it is critical to detect the out-of-domain (OOD) utterances in a conversation. We detect OOD sentences occurring in a dialogue based on sentence distances. The sentence distances are measured by sentence embedding vectors using RNN(Recurrent Neural Network) encoders with the attention mechanism. Our approach improves the accuracy of the out-of-domain detection(OOD) method, and we apply this method to develop a chatbot system for customer services.
In this paper, we present two approaches to generate prognosis from general blood test results. T... more In this paper, we present two approaches to generate prognosis from general blood test results. The first approach is a knowledge-based approach using ripple-down rules (RDR). The knowledge-based approach with RDR converts knowledge of pathologists into a knowledge base with the minimum intervention of knowledge engineers. The second approach is a machine-learning(ML)-based approach using decision tree, random forest and deep neural network (DNN). The ML-based approach learns patterns of attributes from various cases of general blood test. Our experimental results show that there are indeed some important patterns of the attributes in general blood test results, and they are adequately encoded by the both approaches.
It is still a long way to communicate humans and machines emotionally. There are some tries to pr... more It is still a long way to communicate humans and machines emotionally. There are some tries to provide sentimental conversations among humans and machines. Computational humor is one of research topics in computational linguistics and artificial intelligence. We introduce a new method to generate jokes in a sentence related temporal and spatial contexts for continuous conversations with images. We propose a novel model based on a recurrent neural network with natural language processing (NLP) and understanding (NLU) methods. The method generates jokes in a sentence considering temporal and spatial context. The method can joke to trend sensitive users according to different points of humor that vary from region to region. Through this, the user can feel the interest of the conversational service with humorous responses or contents. We apply the method to some applications such as psychiatric counseling and stress management to enhance the applicability of conversational service.
2017 18th IEEE International Conference on Mobile Data Management (MDM)
There are early studies to attempt users for psychiatric counseling with chatbot. They lead to ch... more There are early studies to attempt users for psychiatric counseling with chatbot. They lead to changes in drinking habit based on intervention approach via chat bot. The application does not consider the user's psychiatric status through the conversations, continuous user monitoring, and ethical judgment in the intervention. We contend that more accurate and continuous emotion recognition gives better satisfaction to users who need mental health care. In addition, appropriate clinical psychological response based on ethical responses is as well. We suggest a conversational service for psychiatric counseling that is adapted methodologies to understand counseling contents based on of high-level natural language understanding (NLU), and emotion recognition based on multi-modal approach. The methodologies enable continuous observation of emotional changes sensitively. In addition, the case-based counseling response model that combines ethical judgment model provides a suitable response to clinical psychiatric counseling.
2017 IEEE International Conference on Big Data and Smart Computing (BigComp)
Early study tries to use chatbot for counseling services. They changed drinking habit of who bein... more Early study tries to use chatbot for counseling services. They changed drinking habit of who being consulted by leading them via intervene chatbot. However, the application did not concerned about psychiatric status through continuous conversation with user monitoring. Furthermore, they had no ethical judgment method that about the intervention of the chatbot. We argue that more reasonable and continuous emotion recognition will make better mental healthcare experiment. It will be more proper clinical psychiatric consolation in ethical view as well. This paper suggests a introduce a novel chatbot system for psychiatric counseling service. Our system understands content of conversation based on recent natural language processing (NLP) methods with emotion recognition. It senses emotional flow through the continuous observation of conversation. Also, we generate personalized counseling response from user input, to do this, we use additional constrains to generation model for the proper response generation which can detect conversational context, user emotion and expected reaction.
Studies in health technology and informatics, 2017
There are earlier studies for psychiatric counseling using chat bots. These studies have not cons... more There are earlier studies for psychiatric counseling using chat bots. These studies have not considered the user's emotional status and ethical judgment to provide interventions. This paper proposes an intelligent assistant for psychiatric counseling that understands dialogues using high-level features of natural language understanding, and multi-modal emotion recognition. A response generation model using machine leaning provides suitable responses for clinical psychiatric counseling.
2016 International Conference on Big Data and Smart Computing (BigComp), 2016
Recommending travel destinations on the basis of users' travel intentions is a research topic... more Recommending travel destinations on the basis of users' travel intentions is a research topic being studied recently in the field of intention analysis. This study considers travel intentions from a large number of travel-related reviews containing the reviewers' purpose for visiting the points of interest (POIs). We analyze travel intentions of 83,207 POIs using 6,791,427 reviews in www.TripAdvisor.com with domain-tailored word embedding model. Building an attraction network based on travel intentions helps to recommend travel destinations to travelers and reviewers. We present three prediction methods to recommend travel destinations with an attraction network and description logic. We also present the evaluation results of recommendations from some prediction scenarios. Consequently, the travel intention classification is commensurate with an analysis of intentions from textual data, and the attraction network is useful for recommending travel destinations on the basis of short-and long-term user preferences.
2011 IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), 2011
AbstractUbiquitous computing technologies have been developed fast and various decision support ... more AbstractUbiquitous computing technologies have been developed fast and various decision support systems were proposed. Consecutively, in recently, people are working on an implicit service agent of mobile phone to make people to be provided useful services without ...
In this paper, we propose a new research trend analysis using important word clusters and its rel... more In this paper, we propose a new research trend analysis using important word clusters and its relationship. Journals published many papers every month or week and new scientific contributions were exponentially cumulated to their database. If can analysis important words and related relationships of the papers, a change of research trend in a domain is an interesting topic in text mining. We use a Term Frequency Inverse Document Frequency (TFIDF) to extract meaningful words, the similarity of words measures using WordNet information and a document comparison approach. To measure the similarity from word lists extracted by TFIDF and differences of important word clusters and weights, the approach analyzes the research trend and visualizes the differences of research interest in same research fields. To show usefulness of proposed approach, we illustrate simulations and various results.
This research deals with an issue of preventive medicine in bioinformatics. We can diagnose liver... more This research deals with an issue of preventive medicine in bioinformatics. We can diagnose liver conditions reasonably well to prevent Liver Cirrhosis by classifying liver disorder patients into fatty liver and high risk groups. The classification proceeds in two steps. Classification rules are first built by clustering five attributes (MCV, ALP, ALT, ASP, and GGT) of blood test dataset provided by the UCI Repository. The clusters can be formed by the K-mean method that analyzes multi dimensional attributes. We analyze the properties of each cluster divided into fatty liver, high risk and normal classes. The classification rules are generated by the analysis. In this paper, we suggest a method to diagnosis and predict liver condition to alcoholic patient according to risk levels using the classification rule from the new results of blood test. The K-mean classifier has been found to be more accurate for the result of blood test and provides the risk of fatty liver to normal liver c...
2017 IEEE International Conference on Big Data and Smart Computing (BigComp), 2017
Deep neural networks (DNNs), which show outstanding performance in various areas, consume conside... more Deep neural networks (DNNs), which show outstanding performance in various areas, consume considerable amounts of memory and time during training. Our research led us to propose a controlled dropout technique with the potential of reducing the memory space and training time of DNNs. Dropout is a popular algorithm that solves the overfitting problem of DNNs by randomly dropping units in the training process. The proposed controlled dropout intentionally chooses which units to drop compared to conventional dropout, thereby possibly facilitating a reduction in training time and memory usage. In this paper, we focus on validating whether controlled dropout can replace the traditional dropout technique to enable us to further our research aimed at improving the training speed and memory efficiency. A performance comparison between controlled dropout and traditional dropout is carried out by implementing an image classification experiment on data comprising handwritten digits from the MNI...
2016 International Conference on Big Data and Smart Computing (BigComp), 2016
In Korea, authors of the newspaper article tend to express their intention indirectly, that is, t... more In Korea, authors of the newspaper article tend to express their intention indirectly, that is, they choose a method to leave out some important facts, or sometimes uses biased terms to support their opinion. Since they're not expressing their opinion directly, detecting the political bias is a difficult task. In this paper, we propose a method to detect political bias in the Korean articles by first building word vectors and sentence vectors, and second do a DBN-Training with those vectors and finally do a regression with SVM to calculate the bias. We used our own dataset which is scored with the political bias before doing the regression.
2019 IEEE International Conference on Big Data and Smart Computing (BigComp), 2019
This paper introduce a question understanding system to respond appropriate answers in a dialog s... more This paper introduce a question understanding system to respond appropriate answers in a dialog system for banking services. The question understanding system provides an automated response service in a specific domain (e.g. banking). This can increase response rate of a customer counseling service, and improve business efficiency and expertise. The question understanding system classify domains, specific categories, and speech acts of questions. Finally, the system analyze meanings and intents of the questions, and searching correct answers even various input sentences. In this paper, we describe methods of keyword tokenizing, pattern recognition, sentence embedding, analyzing dialogue intention, and searching similar FAQs. Through these methods, we have developed the question understanding unit in a real interactive system for financial services for real insurance companies and banks, and analyze the usefulness of the system through practical system implementation examples.
2017 IEEE International Conference on Big Data and Smart Computing (BigComp), 2017
Sentence similarity measurement is an important technology that can be apply to various applicati... more Sentence similarity measurement is an important technology that can be apply to various applications in the natural language processing. Recently, an encoder-decoder model using recurrent neural network (RNN) has achieved remarkable results. This paper proposes a model for measuring Korean sentence similarity based on sense-based morpheme embedding and gated recurrent units (GRU) encoder. We evaluate the measurement model consist of experimentally optimized morpheme embedding and sentence encoding models. In the measurement of sentence similarity, the proposed model encoded using the pre-trained morpheme embedding improves the performance compared with the character-embedding model. In addition, it can be used effectively in the question and answering (Q&A) system.
For dialogue systems, it is critical to detect the out-of-domain (OOD) utterances in a conversati... more For dialogue systems, it is critical to detect the out-of-domain (OOD) utterances in a conversation. We detect OOD sentences occurring in a dialogue based on sentence distances. The sentence distances are measured by sentence embedding vectors using RNN(Recurrent Neural Network) encoders with the attention mechanism. Our approach improves the accuracy of the out-of-domain detection(OOD) method, and we apply this method to develop a chatbot system for customer services.
In this paper, we present two approaches to generate prognosis from general blood test results. T... more In this paper, we present two approaches to generate prognosis from general blood test results. The first approach is a knowledge-based approach using ripple-down rules (RDR). The knowledge-based approach with RDR converts knowledge of pathologists into a knowledge base with the minimum intervention of knowledge engineers. The second approach is a machine-learning(ML)-based approach using decision tree, random forest and deep neural network (DNN). The ML-based approach learns patterns of attributes from various cases of general blood test. Our experimental results show that there are indeed some important patterns of the attributes in general blood test results, and they are adequately encoded by the both approaches.
It is still a long way to communicate humans and machines emotionally. There are some tries to pr... more It is still a long way to communicate humans and machines emotionally. There are some tries to provide sentimental conversations among humans and machines. Computational humor is one of research topics in computational linguistics and artificial intelligence. We introduce a new method to generate jokes in a sentence related temporal and spatial contexts for continuous conversations with images. We propose a novel model based on a recurrent neural network with natural language processing (NLP) and understanding (NLU) methods. The method generates jokes in a sentence considering temporal and spatial context. The method can joke to trend sensitive users according to different points of humor that vary from region to region. Through this, the user can feel the interest of the conversational service with humorous responses or contents. We apply the method to some applications such as psychiatric counseling and stress management to enhance the applicability of conversational service.
2017 18th IEEE International Conference on Mobile Data Management (MDM)
There are early studies to attempt users for psychiatric counseling with chatbot. They lead to ch... more There are early studies to attempt users for psychiatric counseling with chatbot. They lead to changes in drinking habit based on intervention approach via chat bot. The application does not consider the user's psychiatric status through the conversations, continuous user monitoring, and ethical judgment in the intervention. We contend that more accurate and continuous emotion recognition gives better satisfaction to users who need mental health care. In addition, appropriate clinical psychological response based on ethical responses is as well. We suggest a conversational service for psychiatric counseling that is adapted methodologies to understand counseling contents based on of high-level natural language understanding (NLU), and emotion recognition based on multi-modal approach. The methodologies enable continuous observation of emotional changes sensitively. In addition, the case-based counseling response model that combines ethical judgment model provides a suitable response to clinical psychiatric counseling.
2017 IEEE International Conference on Big Data and Smart Computing (BigComp)
Early study tries to use chatbot for counseling services. They changed drinking habit of who bein... more Early study tries to use chatbot for counseling services. They changed drinking habit of who being consulted by leading them via intervene chatbot. However, the application did not concerned about psychiatric status through continuous conversation with user monitoring. Furthermore, they had no ethical judgment method that about the intervention of the chatbot. We argue that more reasonable and continuous emotion recognition will make better mental healthcare experiment. It will be more proper clinical psychiatric consolation in ethical view as well. This paper suggests a introduce a novel chatbot system for psychiatric counseling service. Our system understands content of conversation based on recent natural language processing (NLP) methods with emotion recognition. It senses emotional flow through the continuous observation of conversation. Also, we generate personalized counseling response from user input, to do this, we use additional constrains to generation model for the proper response generation which can detect conversational context, user emotion and expected reaction.
Studies in health technology and informatics, 2017
There are earlier studies for psychiatric counseling using chat bots. These studies have not cons... more There are earlier studies for psychiatric counseling using chat bots. These studies have not considered the user's emotional status and ethical judgment to provide interventions. This paper proposes an intelligent assistant for psychiatric counseling that understands dialogues using high-level features of natural language understanding, and multi-modal emotion recognition. A response generation model using machine leaning provides suitable responses for clinical psychiatric counseling.
2016 International Conference on Big Data and Smart Computing (BigComp), 2016
Recommending travel destinations on the basis of users' travel intentions is a research topic... more Recommending travel destinations on the basis of users' travel intentions is a research topic being studied recently in the field of intention analysis. This study considers travel intentions from a large number of travel-related reviews containing the reviewers' purpose for visiting the points of interest (POIs). We analyze travel intentions of 83,207 POIs using 6,791,427 reviews in www.TripAdvisor.com with domain-tailored word embedding model. Building an attraction network based on travel intentions helps to recommend travel destinations to travelers and reviewers. We present three prediction methods to recommend travel destinations with an attraction network and description logic. We also present the evaluation results of recommendations from some prediction scenarios. Consequently, the travel intention classification is commensurate with an analysis of intentions from textual data, and the attraction network is useful for recommending travel destinations on the basis of short-and long-term user preferences.
Uploads
Papers by KyoJoong Oh