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.

QTL Analysis of Yield Components in Rice Using a Cheongcheong/Nagdong Doubled Haploid Genetic Map

Abstract: In this study, only two of 12 quantitative trait loci (QTLs) affecting yield and yield components were identified in a single year, indicating that individual QTLs are probably sensitive to the environment. A rice growth survey of “Cheongcheong” and “Nag dong” in a doubled haploid population in 2012 revealed that yield capacity was influenced by climate change. Analysis of yield and yield components indicated that five average traits are high in “Cheongcheong”. Frequency distribution tables indicated that panicles per plant (PPP), spike lets per panicle (SPP), and 1000-grain weight (TGW) were normally distributed. The strongest relationship was identified between SPP and seed set percentage (SSP) among phenotypic correlations related to yield and yield components found on chromosomes 2, 3, 6, 8 in 2012. SPP and SSP was a very relevant requisite about quantity. Analysis of QTL about quantity was total 9. In the present study, a doubled haploid population was used to analyze the epistatic effects on yield and yield components in rice. Although other epistatic QTLs were not included in any of the main-effect QTLs, they significantly influenced the traits. These results indicated that epistatic interaction plays an important role in controlling the expression of complex traits. Thus, the utilization of marker-assisted selection in rice breeding programs should take epistatic effects into consideration. Hence, the QTLs responsible for major effects are more suitable for marker-assisted selection programs to improve yield and related traits across different environments.

Context-Aware Answer Extraction in Question Answering

Abstract: Extractive QA models have shown very promising performance in predicting the correct answer to a question for a given passage. However, they sometimes result in predicting the correct answer text but in a context irrelevant to the given question. This discrepancy becomes especially important as the number of occurrences of the answer text in a passage increases. To resolve this issue, we propose \textbf{BLANC} (\textbf{BL}ock \textbf{A}ttentio\textbf{N} for \textbf{C}ontext prediction) based on two main ideas: context prediction as an auxiliary task in multi-task learning manner, and a block attention method that learns the context prediction task. With experiments on reading comprehension, we show that BLANC outperforms the state-of-the-art QA models, and the performance gap increases as the number of answer text occurrences increases. We also conduct an experiment of training the models using SQuAD and predicting the supporting facts on HotpotQA and show that BLANC outperforms all baseline models in this zero-shot setting.

Method and system for classifying data consisting of multiple attribues represented by sequences of text words or symbols using deep learning

Abstract: Disclosed are a method and system for automatically classifying data represented by multiple factors having sequences of text words and symbols as values by using deep learning. A method implemented by a computer comprises the following steps of: representing a word vector including sequence information of multiple factors through sequence learning of a word corresponding to the factors with respect to each of the factors constituting the data in a first model using data represented by the factors as an input; calculating a score for each category for category classification of the data using the word vector including the sequence information of the factors in a second model using an output of the first model as an input; and determining at least one category with respect to the data by using the score for the each category.

The Modified Electrode by PEDOP with MWCNTs-Palladium Nanoparticles for the Determination of hydroquinone and Catechol

Abstract: Poly-ethylenedioxypyrrole (PEDOP) coated thiolated multiwall carbon nanotubes palladium nanoparticles (MWCNTs-Pd) modified glassy carbon electrode (GCE) [PEDOP/MWCNTs-Pd/GCE] for the determination of hydroquinone (HQ) and it’s isomer catechol (CA) were synthesized and compared with bare GCE and thiolated multiwall carbon nanotubes (MWCNTs-SH/GCE). The modification could be made by simple processes on a GCE with MWCNTs-Pd covered by PEDOP in a 0.05 M tetrabutylammonium perchlorate (TBAP)/MeCN solution system. A well-defined peak potential evaluation of the oxidation of hydroquinone to quinone at 0.05 V (vs. Ag/AgCl), and electrochemical reduction back to hydroquinone were found by cyclic voltammetry (CV) in phosphate buffered saline (PBS) at pH 7.4. Peak current values increased linearly with increasing hydroquinone contents. The peak separation between the anodic and cathodic peaks at the PEDOP/MWCNTs-Pd/GCE was = 40 mV for HQ and = 70 mV for CA, resulting in a higher electron transfer rate. Moreover, good reproducibility, excellent storage stability, a wide linear range (0.1 - 5 mM for HQ and 0.01 - 6 mM for CA), and low detection limits ( M for HQ and M for CA; S/N = 3) were determined using differential pulse voltammetry (DPV) and amperometric responses; this makes it a promising candidate as a sensor for determination of HQ and CA.

Improved Parallel Wavegan Vocoder with Perceptually Weighted Spectrogram Loss

Abstract: This paper proposes a spectral-domain perceptual weighting technique for Parallel WaveGAN-based text-to-speech (TTS) systems. The recently proposed Parallel WaveGAN vocoder successfully generates waveform sequences using a fast non-autoregressive WaveNet model. By employing multi-resolution short-time Fourier transform (MR-STFT) criteria with a generative adversarial network, the light-weight convolutional networks can be effectively trained without any distillation process. To further improve the vocoding performance, we propose the application of frequency-dependent weighting to the MR-STFT loss function. The proposed method penalizes perceptually-sensitive errors in the frequency domain; thus, the model is optimized toward reducing auditory noise in the synthesized speech. Subjective listening test results demonstrate that our proposed method achieves 4.21 and 4.26 TTS mean opinion scores for female and male Korean speakers, respectively.