• ETRI Journal
• /
• v.39 no.4
• /
• pp.443-454
• /
• 2017

In this paper, we propose a fine-grained mobile application clustering model using retrofitted document embedding. To automatically determine the clusters and their numbers with no predefined categories, the proposed model initializes the clusters based on title keywords and then merges similar clusters. For improved clustering performance, the proposed model distinguishes between an accurate clustering step with titles and an expansive clustering step with descriptions. During the accurate clustering step, an automatically tagged set is constructed as a result. This set is utilized to learn a high-performance document vector. During the expansive clustering step, more applications are then classified using this document vector. Experimental results showed that the purity of the proposed model increased by 0.19, and the entropy decreased by 1.18, compared with the K-means algorithm. In addition, the mean average precision improved by more than 0.09 in a comparison with a support vector machine classifier.

• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.986-1005
• /
• 2020

In this paper, we propose AnoVid, an automated video annotation tool based on deep neural networks, that automatically generates various meta data for each scene or shot in a long drama video containing rich elements. To this end, a novel meta data schema for drama video is designed. Based on this schema, the AnoVid video annotation tool has a total of six deep neural network models for object detection, place recognition, time zone recognition, person recognition, activity detection, and description generation. Using these models, the AnoVid can generate rich video annotation data. In addition, AnoVid provides not only the ability to automatically generate a JSON-type video annotation data file, but also provides various visualization facilities to check the video content analysis results. Through experiments using a real drama video, "Misaeing", we show the practical effectiveness and performance of the proposed video annotation tool, AnoVid.

• Phonetics and Speech Sciences
• /
• v.1 no.2
• /
• pp.105-110
• /
• 2009

This paper introduces the technique of semi-automatically identifying different types of noise in the frequency domain. Given the lower cutoff frequency of the frication noise, and a user-specified constant, the technique identifies the boundary between the frication and aspiration noise in a Korean lax fricative followed by the vowel /a/ by comparing the upper and lower sums of energy with respect to the cutoff frequency. The user-specified constant can be adjusted for different speakers. When the technique was applied to distinguish the two types of noise of Korean lax fricatives from the same speaker, the average and standard deviation of the difference between the manually inserted boundaries and the automatically inserted boundaries were 2.67ms and 1.80ms respectively.

• Proceedings of the IEEK Conference
• /
• 1998.10a
• /
• pp.527-530
• /
• 1998

In this paper, a transformation-based English part of speech tagging system is designed and implemented. The tagging system tags raw corpus at first and the transformation rule correct the errors. Apart from traditional rule based tagging system, this system makes rules automatically. Using 60,000 words of corpus as a training corpus, the transformation rules are generated automatically by iterative training. The idea how to calculate positive effect of transformation and select transformation rules is proposed to generate more effective and correct transformations. In this paper, part of the Brown corpus and English text is used for experimental data. And the performance of transformation based tagging system is demonstrated by the calculation of accuracy.

• 시스템엔지니어링워크숍
• /
• s.4
• /
• pp.182-187
• /
• 2004

Our research proposes how to, systematically, count function point from initial functional requirements based on natural language. Gradually, Function Point Analysis is used to overcome the limitation of LOC(Line Of Code) for estimating software size. Moreover, it plays an important role in cost management. Function point is derived from initial requirements and is determined by experts who have an education for function point. However, currently there are few researches to cout function point by systematic or automatic rules. Through extending our porposed method, we expect that function point is able to be counted automatically or semi-automatically. This would be our future research

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 1997.10a
• /
• pp.213-218
• /
• 1997

Thermal deformation of machine origin of machine tools due to internal and external heat sources has been the most important problem to fabricate products with higher accuracy and performance. In order to solve this problem, GMDH models were constructed to estimate thermal deformation of machine origin for a vertical machining ceneter through measurement of temperature data of specific points on the machine tool. These models are nonlinear equations with high-order polynomials and implemented in a multilayered perceptron type network structure. Input variables and orders are automatically selected by correlation and optimization procedure. Sensors with small influence are deleted automatically in this algorithm. It was shown that the points of temperature measurement can be reduced without sacrificing the estimation accuracy of $\pm$5${\mu}{\textrm}{m}$. From the experimental result, it was confirmed that GMDH methodology was superior to least square models to estimate the thermal behavior of machine tools.

• International Journal of Precision Engineering and Manufacturing
• /
• v.2 no.3
• /
• pp.27-33
• /
• 2001

The aim of the present work is to propose a model for Computer Aided programming of numerical Control lathe. This model is based on the concept by features. It has been developed in an Artificial Intelligence environment, that offers a rapidity as well as a precision for NC code elaboration. In this study a pre-processor has been elaborated to study the geometry of turning workpiece. This pre-processor is a hybrid system which combine a module of design by features and a module of features recognition for a piece provided from an other CAD software. Then, we have conceived a processor that is the heart of the CAD/CAM software. The main functions are to study the fixture of the workpiece, to choose automatically manufacturing cycles, to choose automatically cutting tools (the most relevant), to simulate tool path of manufacturing and calculate cutting conditions, end to elaborate a typical manufacturing process. Finally, the system generates the NC program from information delivered by the processor.

• Proceedings of the Korean Society of Soil and Groundwater Environment Conference
• /
• 2002.04a
• /
• pp.217-220
• /
• 2002

To investigate the conditions of groundwater resources In Jeonju, Wanju, and Goksung, a basic groundwater survey was performed. From the survey, various useful informations such as groundwater use, waterlevel distribution, water chemistry were obtained. This study focused on the analysis of the water levels, which were automatically monitored with pressure transducers or manually measured. The monitorings were conducted for both shallow wells completed in alluvial aquifers and deep wells in bedrock aquifers. The automatically monitored waterlevels for alluvial aquifer were also used for estimation of recharge in the study area. This study presents results of the investigation.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.290-293
• /
• 2005

This paper proposes and evaluates a new approach to detect ships as targets from Radarslit-l SAR (Synthetic Aperture Radar) imagery in the vicinity of Korean peninsula. To be more specific, a labeling technique and morphological filtering in conjunction with some other methods are employed to automatically detect the ships. From the test, the ships are revealed to be detected. For ground truth data, information from a radar system is used, which allows assessing accuracy of the approach. The results showed that the proposed approach has the high potential in automatically detecting the ships

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.1154-1159
• /
• 2004

Unexpected postoperative changes, such as growth in rib hump, has been occasionally reported after corrective surgery for scoliosis. However there has been experimental data for explanation of these changes, nor the suggestion of optimal correction method. This numerical study was designed to investigate the main correlating elements in operative kinematics with post-operative changes of vertebral rotation and rib cage deformation in the corrective surgery of scoliosis. To develop a scoliotic spine model automatically, a special program for converting normal spine model to scoliotic spine model was developed. A mathematical finite element model of normal spine including rib cage, sternum, both clavicles, and pelvis was developed with anatomical details. The skeletal deformity of scoliosis was reconstructed, by mapping the X-ray images of a scoliosis into this three dimensional normal spine and rib cage model. The geometric mapping was performed by translating and rotating the spinal colume with the amount analyzed from the digitized 12 built-in coordinate axes in each vertebral image. By utilizing this program, problems generated in mapping procedure such as facet joint overlapping, vertebral body deformity could be automatically resolved.