• Journal of KIISE
• /
• v.44 no.5
• /
• pp.503-509
• /
• 2017

In this paper, the copy mechanism and input feeding are applied to recurrent neural network(RNN)-search model in a Korean-document summarization in an end-to-end manner. In addition, the performances of the document summarizations are compared according to the model and the tokenization format; accordingly, the syllable-unit, morpheme-unit, and hybrid-unit tokenization formats are compared. For the experiments, Internet newspaper articles were collected to construct a Korean-document summary data set (train set: 30291 documents; development set: 3786 documents; test set: 3705 documents). When the format was tokenized as the morpheme-unit, the models with the input feeding and the copy mechanism showed the highest performances of ROUGE-1 35.92, ROUGE-2 15.37, and ROUGE-L 29.45.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.530-536
• /
• 2021

Recently, due to the development of deep learning, end-to-end speech recognition, which directly maps graphemes to speech signals, shows good performance. Especially, among the end-to-end models, conformer shows the best performance. However end-to-end models only focuses on the probability of which grapheme will appear at the time. The decoding process uses a greedy search or beam search. This decoding method is easily affected by the final probability output by the model. In addition, the end-to-end models cannot use external pronunciation and language information due to structual problem. Therefore, in this paper conformer with lexicon transducer is proposed. We compare phoneme-based model with lexicon transducer and grapheme-based model with beam search. Test set is consist of words that do not appear in training data. The grapheme-based conformer with beam search shows 3.8 % of CER. The phoneme-based conformer with lexicon transducer shows 3.4 % of CER.

• The Research Journal of the Costume Culture
• /
• v.20 no.2
• /
• pp.251-262
• /
• 2012

This research aimed to examine an advertising location according to consumer involvement about internet search advertising and advertising effect according to advertising forms. This research carried out a questionnaire survey from May 25, 2011 for seven days with the target of men and women at the age of 20s-30s who tried search advertising in Seoul and capital area, and 519 copies were used for final analysis. For analysis of collected materials, the reliability test, factor analysis, $t$-test, regression analysis and ${\chi}^2$ test were executed by using statistical package SPSS 16.0. The results are as follows. First, it appeared that consumer's clothing involvement, advertising involvement, advertising location and advertising forms had an influence on site attitude and clicking intention. Consumers of high clothing involvement, high advertising involvement and upper end advertising appeared to have high site attitude and clicking intention. Second, all consumers of the low and high clothing involvement and upper end advertising appeared to prefer the upper end advertising. It appeared that consumers of low advertising involvement prefer lower end advertising, and consumers of high advertising involvement prefer the upper end advertising at an advertising location according to clothing involvement. Accordingly, it is important that the search advertising exposes advertising in the upper end, and that should use an advertising phrase capable of giving trust to consumers. As the text advertising had high site attitude and clicking intention in the advertising forms, consumers should pay attention to the advertising phrase.

• Proceedings of the KIEE Conference
• /
• 2006.07d
• /
• pp.1830-1831
• /
• 2006

In this paper, the search is conducted along each of the coordinate directions for finding the minimum. If $e_i$ is the unit vector along the coordinate direction i, we determine the value a, minimizing f(a)= $f(x+ae_i)$, where a is a real number. A move is made to the new point $x+a_ie_i$ at the end of the search along the direction i. In an n dimensional problem, we define the search along all the directions as one stage. The function value at the end of the stage is compared to the value at the beginning of the stage in establishing the convergence. The gradient appears to be zero at point. We can safeguard this by introducing an acceleration step of one additional step along the pattern direction developed by moves along the coordinate directions.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.18 no.1
• /
• pp.217-222
• /
• 2014

In this paper, I propose an efficient data search system using data partitioning algorithm in Microsoft Excel. I propose searching algorithm to retrieve data quickly using VBA functioning in the Excel. This algorithm is to specify the sheet you are looking for. Once the sheet is specified, the algorithm searches the beginning and the end of the data in the sheet. The algorithm compares intermediate values and key words, from the starting position of the cell. In this way, it will search data to the end. This proposed algorithm was implemented and tested in the Excel system using VBA program. The experimental results showed that the performance was better than that of the conventional sequential search method.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2006.05a
• /
• pp.1609-1615
• /
• 2006

Overlay multicast technique is an effective way as an alternative to IP multicast. Traditional IP multicast is not widely deployed because of the complexity of IP multicast technology and lack of application. But overlay multicast can be easily deployed by effectively reducing complexity of network routers. Because overlay multicast resides on top of densely connected IP network, In case of multimedia streaming service over overlay multicast tree, real-time data is sensitive to end-to-end delay. Therefore, moderate algorithm's development to this network environment is very important. In this paper, we are interested in minimizing maximum end-to-end delay in overlay multicast tree. The problem is formulated as a degree-bounded minimum delay spanning tree, which is a problem well-known as NP-hard. We develop tabu search heuristic with intensification and diversification strategies. Robust experimental results show that is comparable to the optimal solution and applicable in real time

• Journal of the Korean BIBLIA Society for library and Information Science
• /
• v.7 no.1
• /
• pp.171-192
• /
• 1994

End-user Computing(EUC) is one of the significant developments of the 1980s with the explosive supply of personal computers and application software related to it. EUC can be defined as the direct assumption of system development and data processing tasks by the user of the service for his own direct benefit. This is in contrast to the traditional approach to computing in which user requirement are identified and defined and then turned over to professional system designers and programmers for implementation, with the end-user a relatively passive participant in the process. The traditional approach to system development has two obvious drawbacks. First, there is the problem that something may be lost in translation of the problem as originally formulated by the requestor and as it is implemented by the computer specialists. Second, there is the problem that it can take a longtime as growth of significant workloads within system department staff. EUC is an attempt to compensate for this drawbacks associated with the traditional system department process. It permits the end-user to interact directly with an information system, utilizing user friendly software support tools. End-user Searching(EUS) can be defined as accessing online databases and performing search operations for the purpose of finding information to be used by that same person rather than another, in contract with intermediary searcher. The concept of EUS is related to the concept of EUS. That is, two concept are similarly to development background, effects and functions. Therefore, as EUC bring on a change the traditional data processing environment, EUS also bring on a change the traditional intermediary search environment.

• The Journal of Korea Robotics Society
• /
• v.14 no.1
• /
• pp.40-49
• /
• 2019

Reinforcement learning has been applied to various problems in robotics. However, it was still hard to train complex robotic manipulation tasks since there is a few models which can be applicable to general tasks. Such general models require a lot of training episodes. In these reasons, deep neural networks which have shown to be good function approximators have not been actively used for robot manipulation task. Recently, some of these challenges are solved by a set of methods, such as Guided Policy Search, which guide or limit search directions while training of a deep neural network based policy model. These frameworks are already applied to a humanoid robot, PR2. However, in robotics, it is not trivial to adjust existing algorithms designed for one robot to another robot. In this paper, we present our implementation of Guided Policy Search to the robotic arms of the Baxter Research Robot. To meet the goals and needs of the project, we build on an existing implementation of Baxter Agent class for the Guided Policy Search algorithm code using the built-in Python interface. This work is expected to play an important role in popularizing robot manipulation reinforcement learning methods on cost-effective robot platforms.

• Proceedings of the Korean Society for Information Management Conference
• /
• 1995.08a
• /
• pp.45-48
• /
• 1995

Since the early 1970s online searching has provided end-users with commercial and governmental bibliographic databases by trained intermediaries through the developed Vendor/Agency Systems. Recently, the situation has begun to change. The era of end-user searching is coming. End Users are the “information consumers” as the final user of an IT product or set of information. This study on end-user online searching is presented with the following aspects: development, characterization, training, end user online search services in university libraries, role of the intermediaries and librarians, the future.

• Asia Marketing Journal
• /
• v.16 no.3
• /
• pp.23-33
• /
• 2014

The selection of keywords for bidding is a critical component of paid search advertising. When the number of possible keywords is enormous, it becomes difficult to choose the best keywords for advertising and then subsequently to assess their effect. To this end, we propose an ultrahigh dimensional keyword selection approach that not only reduces the dimension for selections, but also generates the top listed keywords for profits. An empirical analysis using a unique panel dataset from a large online clothes retailer that advertises on the largest search engine in China (i.e., Baidu) is presented to illustrate the usefulness of our approach.