• Phonetics and Speech Sciences
• /
• v.2 no.4
• /
• pp.223-228
• /
• 2010

This study investigated the closed quotient and other voice quality parameters using electroglottography (EGG) in sustaining the vowel /a/ and reading a sentence at the comfortable pitch and loudness in healthy Korean adults. Seventy two healthy adults (36 men, 36 women) aged 20~40 years were included in the study. The tasks were recorded and analyzed using Lx Speech Studio. In vowel sustaining task, closed quotient (Qx), fundamental frequency (Fx), sound pressure level (SPL), Jitter, and Shimmer were measured. In sentence reading task, closed quotient (DQx), fundamental frequency (DFx), and sound pressure level (DAx) were measured. The sex effects were observed on Qx, Fx, Shimmer, DQx, and DFx. Men had significantly higher Qx and DQx than women, but had significantly lower Shimmer than women. However, there was no sex effect on Jitter. The task effects on Qx and SPL as well as DQx and DAx were also assessed. Qx and SPL were significantly higher than DQx and DAx in both gender. This study showed that the closed quotients in both vowel sustaining and sentence reading tasks were significantly related to other voice quality parameters. Therefore, clinicians and researchers should describe the voice quality parameters like fundamental frequency, sound pressure level, Jitter, Shimmer, and so on when reporting closed quotients using EGG.

• English Language & Literature Teaching
• /
• v.18 no.2
• /
• pp.67-84
• /
• 2012

The present study investigated the effect of sentence translation as a production task on short-term and long-term retention of foreign vocabulary. 87 EFL university students at a beginning level, enrolled in reading class participated in the study. The study compared the performance of three groups on vocabulary recall: (1) Control group, (2) Translation group, and (3) Copy group. During the treatment sessions, translation group translated L1 sentences into English, while copy group simply copied given English sentences with each target word. Results of the immediate test were collected each week from week 2 to week 5 and analyzed by one-way ANOVA. Results revealed that regarding short-term vocabulary retention, participants in rote-copy condition outperformed those in translation group. Four weeks later a delayed test was administered to measure long-term vocabulary retention. In contrast, the results of two-way repeated measures ANOVA showed that long-term vocabulary retention of translation group was significantly greater than copy group. The findings suggest that although sentence translation is rather challenging to low-level learners, it may facilitate long-term retention of new vocabulary given the more elaborate and deeper processing the task entails.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.27 no.2
• /
• pp.167-176
• /
• 2024

Lip-reading is the task of inferring the speaker's utterance from silent video based on learning of lip movements. It is very challenging due to the inherent ambiguities present in the lip movement such as different characters that produce the same lip appearances. Recent advances in deep learning models such as Transformer and Temporal Convolutional Network have led to improve the performance of lip-reading. However, most previous works deal with English lip-reading which has limitations in directly applying to Korean lip-reading, and moreover, there is no a large scale Korean lip-reading dataset. In this paper, we introduce the first large-scale Korean lip-reading dataset with more than 120 k utterances collected from TV broadcasts containing news, documentary and drama. We also present a preprocessing method which uniformly extracts a facial region of interest and propose a transformer-based model based on grapheme unit for sentence-level Korean lip-reading. We demonstrate that our dataset and model are appropriate for Korean lip-reading through statistics of the dataset and experimental results.

• Language and Information
• /
• v.9 no.2
• /
• pp.127-139
• /
• 2005

In this paper I investigate whether Korean-speaking children know the basic meaning of reciprocal anaphors like each other. I further examine whether they have knowledge of subtle differences in the interpretations of such anaphors depending on the two types of verbs. Fiengo and Lasnik (1973) pointed out a contrast between reciprocal sentences with active verbs and stative verbs. For example, a sentence with an active verb like The men in the room are hitting each other, has both a strong reciprocal reading (i.e., everyone of them in the room is hitting every other one) and a weak reciprocal reading (i.e., certain pairs of men are not engaged in the action of hitting each other). In contrast, a sentence with a stative verb like The men in the room know each other allows only a strong reciprocal reading (i.e., everyone of them know every other one). 16 Korean children and 15 Korean adults were tested using the Truth Value Judgment Task methodology. The results of the present study show that like English children, Korean children know the meaning of reciprocal anaphor, and that they also know the semantic difference of reciprocal sentences with active and stative verbs. Therefore, the present study strongly supports the claim that the semantic distinction of reciprocal sentences with active and stative verbs may be universal, and that children's ability of this semantic distinction might be innately given.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.270-272
• /
• 2018

본 연구에서는 질의-본문 간 문장 비교 정보가 reading comprehension task 의 성능 향상에 도움이 되는지를 확인해 보았다. 기존의 reading comprehension 방법론이 질의-본문 간 의미 비교정보를 활용하지만, 본문 전체를 대상으로 한 비교이기 때문에 문장 단위의 정보가 활용되지 못하는 단점이 있었다. 실험에 사용한 데이터는 대표적인 RC 데이터 중 하나인 NewsQA[5] 를 이용하였으며, 질의-본문 문장 간 비교를 통한 성능 향상의 잠재력을 확인하였다.

• Phonetics and Speech Sciences
• /
• v.12 no.4
• /
• pp.91-98
• /
• 2020

This study investigated the acoustic characteristics of sustained vowel /a/ and sentence utterance produced by patients with muscle tension dysphonia (MTD) using cepstrum-based acoustic variables. 36 women diagnosed with MTD and the same number of women with normal voice participated in the study and the data were recorded and measured by ADSVTM. The results demonstrated that cepstral peak prominence (CPP) and CPP_F0 among all of the variables were statistically significantly lower than those of control group. When it comes to the GRBAS scale, overall severity (G) was most prominent, and roughness (R), breathiness (B), and strain (S) indices followed in order in the voice quality of MTD patients. As these characteristics increased, a statistically significant negative correlation was observed in CPP. We tried to classify MTD and control group using CPP and CPP_F0 variables. As a result of statistic modeling with a Random Forest machine learning algorithm, much higher classification accuracy (100% in training data and 83.3% in test data) was found in the sentence reading task, with CPP being proved to be playing a more crucial role in both vowel and sentence reading tasks.

• Korean Journal of Cognitive Science
• /
• v.15 no.3
• /
• pp.25-33
• /
• 2004

Two experiments were conducted to investigate the effects of two syntactic cues in pronoun referential resolution: number cue (plural or singular) and gender cue (unambiguous or ambiguous). Using self-paced sentence reading task for pronoun sentences and lexical decision task for antecedents, Experiment 1 showed that the reading time of a plural pronoun ('they') was faster than a singular pronoun ('he' or 'she'), but the lexical decision time did not differ with a number cue and a Bender cue. In Experiment 2, using RSVP for pronoun sentences and lexical decision task for antecedents, the results showed that the lexical decision time differed for a gender cue only. These results suggested that the syntactic cues of a pronoun influenced strongly on referential resolution in discourse comprehension.

• Korean Journal of Cognitive Science
• /
• v.23 no.4
• /
• pp.487-515
• /
• 2012

The purpose of this study was to investigate the effect of the individual differences in cognitive processes on paragraph comprehension. To do so, the lexical decision task and the pattern comparison task were used to measure the low-level cognitive processes. Digit span task was used to test the phonological loop capacity. The individual differences of the central executive processing capacity were measured by operational span task. Reading span task was used to test the working memory capacity related with the sentence processing. Reading times and accuracies of the logically valid inferences and logically void inferences were tested to measure the high-level cognitive processes. Reading times and accuracies for the target sentences with and without prior explicit causal sentence were measured to test individuals' paragraph comprehension abilities. The results showed that the speed of the low-level cognitive processes was related with the speed of the high-level cognitive processes. Also, the accuracy of the low-level cognitive processes was related with the accuracy of the high-level cognitive processes while there was no significant correlation between the speed and the accuracy in any measures of the cognitive processes. Working memory capacity was related with the accuracy of the cognitive processes while it was not significantly correlated with the speed of the cognitive processes. Most importantly, the speed of low-level cognitive processes significantly affected the speed of the paragraph comprehension while the working memory capacity and the high-level cognitive processes had influences on the accuracies of the paragraph comprehension. The speed of the paragraph comprehension had no influence on the accuracies of the paragraph comprehension.

• ETRI Journal
• /
• v.41 no.6
• /
• pp.771-781
• /
• 2019

We propose the VS³-NET model to solve the task of question answering questions with machine-reading comprehension that searches for an appropriate answer in a given context. VS³-NET is a model that trains latent variables for each question using variational inferences based on a model of a simple recurrent unit-based sentences and self-matching networks. The types of questions vary, and the answers depend on the type of question. To perform efficient inference and learning, we introduce neural question-type models to approximate the prior and posterior distributions of the latent variables, and we use these approximated distributions to optimize a reparameterized variational lower bound. The context given in machine-reading comprehension usually comprises several sentences, leading to performance degradation caused by context length. Therefore, we model a hierarchical structure using sentence encoding, in which as the context becomes longer, the performance degrades. Experimental results show that the proposed VS³-NET model has an exact-match score of 76.8% and an F1 score of 84.5% on the SQuAD test set.

• Journal of the Korea Convergence Society
• /
• v.13 no.1
• /
• pp.81-88
• /
• 2022

The major educational goal of reading part, which occupies important portion in Korean language in Korean SAT, is to evaluated whether a given text can be fully understood. Therefore given questions in the exam must be able to solely solvable by given text. In this paper we developed a datatset based on Korean SAT's reading part in order to evaluate whether a deep learning language model can classify if the given question is true or false, which is a binary classification task in NLP. In result, by applying language model solely according to the passages in the dataset, we were able to acquire better performance than 59.2% in F1 score for human performance in most of language models, that KoELECTRA scored 62.49% in our experiment. Also we proved that structural limit of language models can be eased by adjusting data preprocess.