Naver Corporation

About: Naver Corporation is a company organization based out in Seongnam-si, South Korea. It is known for research contribution in the topics: Terminal (electronics) & Computer science. The organization has 4038 authors who have published 4294 publications receiving 35045 citations. The organization is also known as: NAVER Corporation & NAVER.

Delving into VoxCeleb: environment invariant speaker recognition

Abstract: Research in speaker recognition has recently seen significant progress due to the application of neural network models and the availability of new large-scale datasets. There has been a plethora of work in search for more powerful architectures or loss functions suitable for the task, but these works do not consider what information is learnt by the models, apart from being able to predict the given labels. In this work, we introduce an environment adversarial training framework in which the network can effectively learn speaker-discriminative and environment-invariant embeddings without explicit domain shift during training. We achieve this by utilising the previously unused `video' information in the VoxCeleb dataset. The environment adversarial training allows the network to generalise better to unseen conditions. The method is evaluated on both speaker identification and verification tasks using the VoxCeleb dataset, on which we demonstrate significant performance improvements over baselines.

Method for offering a search-word advertisement and generating a search result list in response to the search-demand of a searcher and a system thereof

Abstract: The present invention relates to method and system for providing a search word advertisement which can compute a predetermined expected advertisement cost by predicting the maximum number of expected clicks at the future point with respect to particular search listing, and subtract a cost for execution of advertisement from the expected advertisement cost, wherein the cost for execution of advertisement is determined on the basis of the number of valid clicks that have actually occurred during a contracted advertising period. According to the present invention, in a model for providing an advertisement through a search word advertisement, there is an effect that it is possible to provide method and system for providing a search word advertisement which predict the maximum number of expected clicks by predetermined future point, and adopt a unit cost per click, thereby scientifically computing an expected advertisement cost and charging an advertiser a reasonable advertisement cost.

Method and system of determining vulnerability of web application

Abstract: A method of determining vulnerability of web application comprises: selecting fixed parameters from parameters of URL link extracted from a website; determining whether a process of determining vulnerability for the selected fixed parameter is completed or not; inserting an attack pattern for each attack type to an input value for the selected fixed parameter, when the process of determining vulnerability for the selected fixed parameter is not completed; and determining vulnerability of the selected fixed parameter by each attack type through an analysis of response to an input of URL link with the attack pattern inserted thereinto.

Dynamic Sensor Interrogation Using Wavelength-Swept Laser with a Polygon-Scanner-Based Wavelength Filter

Abstract: We report a high-speed (~2 kHz) dynamic multiplexed fiber Bragg grating (FBG) sensor interrogation using a wavelength-swept laser (WSL) with a polygon-scanner-based wavelength filter. The scanning frequency of the WSL is 18 kHz, and the 10 dB scanning bandwidth is more than 90 nm around a center wavelength of 1,540 nm. The output from the WSL is coupled into the multiplexed FBG array, which consists of five FBGs. The reflected Bragg wavelengths of the FBGs are 1,532.02 nm, 1,537.84 nm, 1,543.48 nm, 1,547.98 nm, and 1,553.06 nm, respectively. A dynamic periodic strain ranging from 500 Hz to 2 kHz is applied to one of the multiplexed FBGs, which is fixed on the stage of the piezoelectric transducer stack. Good dynamic performance of the FBGs and recording of their fast Fourier transform spectra have been successfully achieved with a measuring speed of 18 kHz. The signal-to-noise ratio and the bandwidth over the whole frequency span are determined to be more than 30 dB and around 10 Hz, respectively. We successfully obtained a real-time measurement of the abrupt change of the periodic strain. The dynamic FBG sensor interrogation system can be read out with a WSL for high-speed and high-sensitivity real-time measurement.

An ethanol extract derived from Bonnemaisonia hamifera scavenges ultraviolet B (UVB) radiation-induced reactive oxygen species and attenuates UVB-induced cell damage in human keratinocytes.

Abstract: The present study investigated the photoprotective properties of an ethanol extract derived from the red alga Bonnemaisonia hamifera against ultraviolet B (UVB)-induced cell damage in human HaCaT keratinocytes. The Bonnemaisonia hamifera ethanol extract (BHE) scavenged the superoxide anion generated by the xanthine/xanthine oxidase system and the hydroxyl radical generated by the Fenton reaction (FeSO4 + H2O2), both of which were detected by using electron spin resonance spectrometry. In addition, BHE exhibited scavenging activity against the 1,1-diphenyl-2-picrylhydrazyl radical and intracellular reactive oxygen species (ROS) that were induced by either hydrogen peroxide or UVB radiation. BHE reduced UVB-induced apoptosis, as shown by decreased apoptotic body formation and DNA fragmentation. BHE also attenuated DNA damage and the elevated levels of 8-isoprostane and protein carbonyls resulting from UVB-mediated oxidative stress. Furthermore, BHE absorbed electromagnetic radiation in the UVB range (280–320 nm). These results suggest that BHE protects human HaCaT keratinocytes against UVB-induced oxidative damage by scavenging ROS and absorbing UVB photons, thereby reducing injury to cellular components.