• 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.26 no.10
• /
• pp.1-7
• /
• 2021

Recently, since most of the research on correcting speech recognition errors is based on English, there is not enough research on Korean speech recognition. Compared to English speech recognition, however, Korean speech recognition has many errors due to the linguistic characteristics of Korean language, such as Korean Fortis and Korean Liaison, thus research on Korean speech recognition is needed. Furthermore, earlier works primarily focused on editorial distance algorithms and syllable restoration rules, making it difficult to correct the error types of Korean Fortis and Korean Liaison. In this paper, we propose a context-sensitive post-processing model of speech recognition using a LSTM-based sequence-to-sequence model and Bahdanau attention mechanism to correct Korean speech recognition errors caused by the pronunciation. Experiments showed that by using the model, the speech recognition performance was improved from 64% to 77% for Fortis, 74% to 90% for Liaison, and from 69% to 84% for average recognition than before. Based on the results, it seems possible to apply the proposed model to real-world applications based on speech recognition.

• International journal of advanced smart convergence
• /
• v.5 no.3
• /
• pp.8-15
• /
• 2016

In recent years, one of deep learning models called Deep Belief Network (DBN) which formed by stacking restricted Boltzman machine in a greedy fashion has beed widely used for classification and recognition. With an ability to extracting features of high-level abstraction and deal with higher dimensional data structure, this model has ouperformed outstanding result on image and speech recognition. In this research, we assess the applicability of deep learning in dna classification level. Since the training phase of DBN is costly expensive, specially if deals with DNA sequence with thousand of variables, we introduce a new encoding method, using decimal-binary vector to represent the sequence as input to the model, thereafter compare with one-hot-vector encoding in two datasets. We evaluated our proposed model with different contrastive algorithms which achieved significant improvement for the training speed with comparable classification result. This result has shown a potential of using decimal-binary vector on DBN for DNA sequence to solve other sequence problem in bioinformatics.

• The Pure and Applied Mathematics
• /
• v.16 no.3
• /
• pp.255-258
• /
• 2009

It is well known that for a sequence a = ($a_0,\;a_1$,...) the general term of the dual sequence of a is $a_n\;=\;c_0\;^n_0\;+\;c_1\;^n_1\;+\;...\;+\;c_n\;^n_n$, where c = ($c_0,...c_n$ is the dual sequence of a. In this paper, we find the general term of the sequence ($c_0,\;c_1$,... ) and give another method for finding the inverse matrix of the Pascal matrix. And we find a simple proof of the fact that if the general term of a sequence a = ($a_0,\;a_1$,... ) is a polynomial of degree p in n, then ${\Delta}^{p+1}a\;=\;0$.

• KIISE Transactions on Computing Practices
• /
• v.22 no.11
• /
• pp.614-618
• /
• 2016

As information on the internet and the data from smart devices are growing, the amount of stream data is also increasing in the real world. The stream data, which is a potentially large data, requires online learnable models and algorithms. In this paper, we propose a novel class of models: predictive convolutional neural networks to be able to perform online learning. These models are designed to deal with longer patterns as the layers become higher due to layering convolutional operations: detection and max-pooling on the time axis. As a preliminary check of the concept, we chose two-month gathered GPS data sequence as an observation sequence. On learning them with the proposed method, we compared the original sequence and the regenerated sequence from the abstract information of the models. The result shows that the models can encode long-range patterns, and can generate a raw observation sequence within a low error.

• Journal of KIISE:Databases
• /
• v.36 no.3
• /
• pp.169-179
• /
• 2009

Communication log data consist of communication events such as sending and receiving e-mail or instance message and visiting web sites, etc. Many countries including USA and EU enforce the retention of these data on the communication service providers for the purpose of investigating or detecting criminals through the Internet. Because size of the retained data is very large, the efficient method for extracting valuable information from the data is needed for Law Enforcement Authorities to use the retained data. This paper defines the Interactive Communication Sequence Patterns(ICSPs) that is the important information when each communication event in communication log data consists of sender, receiver, and timestamp of this event. We also define a Mining(FDICSP) problem to discover such patterns and propose a method called Fast Discovering Interactive Communication Sequence Pattern(FDICSP) to solve this problem. FDICSP focuses on the characteristics of ICS to reduce the search space when it finds longer sequences by using shorter sequences. Thus, FDICSP can find Interactive Communication Sequence Patterns efficiently.

• Proceedings of the KOSOMBE Conference
• /
• v.1998 no.11
• /
• pp.55-56
• /
• 1998

The 3-D Fast Gradient Echo (Turbo FLASH, Turbo Fast Low Angle Shot) sequence is optimized to achieve a good T1 contrast using variable excitation flip angles. In Turbo FLASH sequence, depending on the contrast preparation scheme, various types of image contrast can be established. While proton density contrast is obtained when using a short repetition time with a short echo time and small flip angles, T1 or T2 weighting can be obtained with proper contrast preparation sequences applied before the above proton density Turbo FLASH sequence. To maximize the contrast to noise ratio while retaining a sharp impulse response (smooth frequency domain response), the excitation flip-angle pattern is optimized through simulation and experiments. The TI (the delay after the preparation sequence which is a 180 degree inversion RF pulse in the IR T1 weighted imaging case), TD (the delay time between the Turbo FLASH sequence and the next preparation), and TR are also optimized fur the best image quality. The proposed 3-D Turbo FLASH provides $1mm\times1mm\times1.5mm$ high resolution images within a reasonable 5-8 minutes of imaging time. The proposed imaging sequence has been implemented in a Medison's Magnum 1.0T system and verified through simulations as well as human volunteer imaging. The experimental results show the utility of the proposed method.

• Proceedings of the IEEK Conference
• /
• 2004.08c
• /
• pp.486-490
• /
• 2004

Programmable logic controller (PLC) is widely used in manufacturing system or process control. This paper presents the design of a 32-bit microprocessor for a sequence control using an Application Specification Integrated Circuit (ASIC). The 32-bit microprocessor was designed by a VHDL with top down method; the program memory was separated from the data memory for high speed execution of 274 specified sequence instructions. Therefore it was possible that sequence instructions could be operated at the same time during the instruction fetch cycle. And in order to reduce the instruction decoding time and the interface time of the data memory interface, an instruction code size was implemented by 32-bits. And the real time debugging as single step run, break point run was implemented. Pulse instruction, step controller, master controllers, BIN and BCD type arithmetic instructions, barrel shit instructions were implemented for many used in PLC system. The designed microprocessor was synthesized by the S1L50000 series which contains 70,000 gates with 0.65um technology of SEIKO EPSON. Finally, the benchmark was performed to show that designed 32-bit microprocessor has better performance than Q4A PLC of Mitsubishi Corporation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.2
• /
• pp.538-561
• /
• 2020

Recommender Systems (RecSys) have a major role in e-commerce for recommending products, which they may like for every user and thus improve their business aspects. Although many types of RecSyss are there in the research field, the state of the art RecSys has focused on finding the user similarity based on sequence (e.g. purchase history, movie-watching history) analyzing and prediction techniques like Recurrent Neural Network in Deep learning. That is RecSys has considered as a sequence prediction problem. However, evaluation of similarities among the customers is challenging while considering temporal aspects, context and multi-component ratings of the item-records in the customer sequences. For addressing this issue, we are proposing a Deep Learning based model which learns customer similarity directly from the sequence to sequence similarity as well as item to item similarity by considering all features of the item, contexts, and rating components using Dynamic Temporal Warping(DTW) distance measure for dynamic temporal matching and 2D-GRU (Two Dimensional-Gated Recurrent Unit) architecture. This will overcome the limitation of non-linearity in the time dimension while measuring the similarity, and the find patterns more accurately and speedily from temporal and spatial contexts. Experiment on the real world movie data set LDOS-CoMoDa demonstrates the efficacy and promising utility of the proposed personalized RecSys architecture.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.3 no.5
• /
• pp.492-506
• /
• 2009

Index structures that are based on sequence matching for XPath processing such as ViST, PRIX and LCS-TRIM have recently been proposed to reduce the search time of XML documents. However, ViST can cause a lot of unnecessary computation and I/O when processing structural joint queries because its numbering scheme is not optimized. PRIX and LCS-TRIM require much processing time for matching XML data trees and queries. In this paper, we propose a novel index structure that solves the problems of ViST and improves the performance of PRIX and LCS-TRIM. Our index structure provides the minimum sequence matching scheme to efficiently process structural queries. Finally, to verify the superiority of the proposed index structure with the minimum sequence matching scheme, we compare our index structure with ViST, PRIX and LCS-TRIM in terms of query processing of a single path or of a branching path including wild-cards ('*' and '//' ).