L1, L2 스페인어 접속법 사용에 대한 변이주의적 접근

황지희 서울대학교 라틴아메리카연구소 2023 이베로아메리카硏究 Vol.34 No.3

스페인어의 법(mood)은 크게 직설법과 접속법으로 나뉘며, 이에 관한 연구에서는 통사적이거나 의미적인 요소들이 두 법의 대조와 관련이 있다고 주장됐다. 제2언어(L2)로서의 스페인어 연구에서는 L2 학습자들이 스페인어 법 대조 습득에 많은 어려움을 겪으며, 이에 따라 법 대조는 스페인어 학습의 후반기에 습득되는 언어 구조라고 받아들여 왔다. 이러한 주장과 가정들은 주로 접속법의 형태, 통사, 의미, 또는 화용적인 자질에 초점을 맞춘 기존의 연구들을 통해 확인됐다. 그러나 최근의 사회언어학 연구들은 실제 언어 사용 환경에서 발견되는 스페인어의 법 대조 현상에는 이보다 더욱 복잡하고 다양한 요인들이 작용한다는 것을 보여주었다. 본 연구는 이러한 논의를 기반으로 하여 L1과 L2 스페인어 법 대조를 변이주의(variationist) 방법론을 사용하여 분석했다. 특히, 일부 선행 연구에서 상반되는 결과가 도출된 동사의 형태 규칙성에 대한 변수를 고려하여 분석을 진행했다. 스페인어 작문 코퍼스인 CEDEL2에서 추출한 L2 스페인어 학습자와 L1 스페인어 원어민 화자의 작문 데이터를 비교한 결과, 불규칙 변화 동사가 종속 명사절에서 접속법으로 더욱 자주 사용되며, 주절이 부정문으로 작성되거나 종속절 동사가 저빈도 동사 그룹에 속할 때 접속법이 더욱 자주 사용된다는 경향이 확인되었다. 본 연구는 변이주의 접근법을 통해 L2 스페인어 법 대조에 대한 새로운 통찰력을 제공하고자 시도했다는 점에서 의의가 있다. The Spanish mood is generally divided into the indicative and subjunctive, with previous research suggesting a close link between syntactic and/or semantic features and the distinction between these moods. In the study of Spanish as a second language (L2), it is believed that learners often struggle with acquiring the subjunctive, viewing it as a structure acquired in the later stages of learning. Previous studies have primarily supported these assumptions, focusing on morpho-syntactic, semantic, and/or pragmatic aspects of the subjunctive. However, recent sociolinguistic studies reveal a more complex array of factors influencing mood distinction observed in authentic Spanish usage. Building on these discussions, the present study employs a variationist methodology to analyze L1 and L2 Spanish mood distinction. Considering conflicting results in the literature regarding verb form regularity, this study also incorporates the analysis of this variable. Examining written production data from L2 Spanish speakers in the CEDEL2 corpus and comparing it with a control group of L1 Spanish speakers, the present study concludes that irregular verbs favor the subjunctive in subordinate noun clauses. Additionally, the negative polarity of the main clause and the low frequency of the verb in the subordinate clause condition the use of the subjunctive form. Given limited discussion on sociolinguistic methodology in previous L2 Spanish acquisition studies, the current research attempted to offer insight into the application and implications of a variationist approach to L2 Spanish mood distinction.

규칙적인 운동이 고혈압유발 흰쥐의 심근에서 Bcl-2발현 및 Apoptosis변화에 미치는 영향

이진 ( Jin Lee ),조형숙 ( Hyung Sook Cho ),김원규 ( Won Kyu Kim ) 한국스포츠정책과학원(구 한국스포츠개발원) 2006 체육과학연구 Vol.17 No.1

본 연구의 목적은 고혈압흰쥐에 있어서 규칙적인 운동이 항고혈압 및 심근에서의 세포자멸사를 예방하는 효과를 알아보는 것이다. 이에 본 실험에서는 흰쥐에 L-NAME를 투여하여 고혈압을 유발시킨 후 규칙적인 운동(수영)을 시행하여 수축기와 이완기혈압을 측정하였고, 심근에서의 세포자멸사현상을 TUNEL방법으로, 항세포자멸사인자인 Bcl-2발현현상을 면역조직화학염색법으로 분석하였다. 연구대상은 Sprague-Dawley계 숫컷 흰쥐 40마리를 대조군(n=10), L-NAME를 투여한 L-투여고혈압군(n=10), L-NAME를 투여와 운동을 시행한 L-투여고혈압운동군(n=10) 및 운동군(n=10)의 4군으로 분류하였고, 고혈압유도는 L-NAME(50mg/ kg)를 주당 5회, 총 8주간 투여하였다. 약물 투여 1시간 후 1일 1회 30분씩 운동을 시행하였고, 주 1회 혈압을 측정하였다. 수축기혈압과 이완기혈압은 L-NAME 투여 1주 후부터 증가하여 실험기간동안, 고혈압(p<.001)을 보였으나, L-투여고혈압운동군에서는 수영운동 4주후부터 실험종료때까지 지속적으로 혈압이 감소(p<.05)하였다. L-투여고혈압군의 심근세포에서는 자멸세포의 수가 증가되었으나, L-투여고혈압운동군에서는 세포자멸사가 억제된 소견을 보였다. 특히 세포자멸사를 억제시키는 Bcl-2의 발현은 L-투여고혈압운동군에서 L-투여고혈압군에 비해 뚜렷하게 증가되었으며, 대조군과 비슷한 발현양상을 보였다. 이상의 결과를 종합하면 8주간의 규칙적인 운동은 고혈압 흰쥐의 혈압을 감소시켜 항고혈압효과를 나타내었고, Bcl-2의 발현으로 세포자멸사 증가가 억제함으로서 병적인 심장상태의 진행을 억제하는 효과가 있음을 알 수 있었다. 따라서 규칙적인 운동은 고혈압심장의 기능을 향상시켜 세포자멸사로부터 심장을 보호할 수 있는 것으로 생각된다. The present study was designed to examine the antihypertensive effects of regular exercise by suppressing apoptosis in myocardium of L-NAME induced hypertension rats. We measured systolic and diastolic blood pressure (BP), and performed Bcl-2 Immunohistochemistry and TUNEL for apoptosis in heart tissue. Sprague-Dawley (n=40) rats were divided into 4 groups as follows: (1) control group (n=10, Con), (2) hypertension group(n=10, L-Con) (3) hypertension+exercise group (n=10, L-Ex), (4) Exercise group (n=10, Ex). To induce hypertension we administrated L-NAME (50 mg/kg) for 5days per week during 8weeks. The regular exercise was performed in swim water bath for 30min after 1hr L-NAME administration. Both L-NAME treatment and swim were performed 5 days /week with duration of 30minutes for 8weeks. BP was measured once per week. Results showed that the BP was significantly increased (p<.001) from 1week after L-NAME treatment, but systolic BP of L-Ex(p<.001) and diastolic BP of Ex(p<.05) groups showed markedly decreased from 4weeks and 2weeks after exercise, respectively. Particulary, the Bcl-2 expression in L-Ex was markedly increased as compared with L-Con group, whereas L-Con group showed decreased Bcl-2 expression and increased apoptotic cell than other groups. These results suggest that regular exercise would induce anti-hypertensive effect by controling the BP and preventing the apoptosis through bcl-2 expression in myocardium of L-NAME treated rats. Thus, regular exercise may be helpful to improvement of hypertensive heart function.

Design of Space Search-Optimized Polynomial Neural Networks with the Aid of Ranking Selection and L2-norm Regularization

Dan Wang,Sung-Kwun Oh,Eun-Hu Kim 대한전기학회 2018 Journal of Electrical Engineering & Technology Vol.13 No.4

The conventional polynomial neural network (PNN) is a classical flexible neural structure and self-organizing network, however it is not free from the limitation of overfitting problem. In this study, we propose a space search-optimized polynomial neural network (ssPNN) structure to alleviate this problem. Ranking selection is realized by means of ranking selection-based performance index (RS_PI) which is combined with conventional performance index (PI) and coefficients based performance index (CPI) (viz. the sum of squared coefficient). Unlike the conventional PNN, L2-norm regularization method for estimating the polynomial coefficients is also used when designing the ssPNN. Furthermore, space search optimization (SSO) is exploited here to optimize the parameters of ssPNN (viz. the number of input variables, which variables will be selected as input variables, and the type of polynomial). Experimental results show that the proposed ranking selection-based polynomial neural network gives rise to better performance in comparison with the neuron fuzzy models reported in the literatures.

전기 저항 단층촬영법에서의 조정기법 성능비교

강숙인(Suk-In Kang),김경연(Kyung-Youn Kim) 한국전기전자학회 2016 전기전자학회논문지 Vol.20 No.3

전기 저항 단층촬영법(ERT)은 대상체 내부 단면의 저항률 분포를 추정하고 이를 영상화하는 기술이다. ERT의 영상복원은 매우 비정치성이 강한 역문제의 일종으로 의미있는 영상을 얻기 위해서는 조정기법이 사용된다. 대표적으로 l2-norm 조정기법, l1-norm 조정기법, Total Variation 조정기법이 사용되며, 조정기법에 따라 ERT의 영상복원성능이 달라진다. 즉, 상황에 맞는 적절한 조정기법의 사용은 ERT 영상 복원을 개선할 수 있다. 따라서, 본 논문에서는 모의실험을 통하여 상황에 따른 세 가지 조정기법의 영상복원 성능을 비교하였다. Electrical resistance tomography (ERT) is an imaging technique where the internal resistivity distribution inside an object is reconstructed. The ERT image reconstruction is a highly nonlinear ill-posed problem, so regularization methods are used to achieve desired image. The reconstruction outcome is dependent on the type of regularization method employed such as l2-norm, l1-norm, and total variation regularization method. That is, use of an appropriate regularization method considering the flow characteristics is necessary to attain good reconstruction performance. Therefore, in this paper, regularization methods are tested through numerical simulations with different flow conditions and the performance is compared.

안정화된 딥 네트워크 구조를 위한 다항식 신경회로망의 연구

전필한(Pil-Han Jeon),김은후(Eun-Hu Kim),오성권(Sung-Kwun Oh) 대한전기학회 2017 전기학회논문지 Vol.66 No.12

In this study, the design methodology for alleviating the overfitting problem of Polynomial Neural Networks(PNN) is realized with the aid of two kinds techniques such as L₂ regularization and Sum of Squared Coefficients (SSC). The PNN is widely used as a kind of mathematical modeling methods such as the identification of linear system by input/output data and the regression analysis modeling method for prediction problem. PNN is an algorithm that obtains preferred network structure by generating consecutive layers as well as nodes by using a multivariate polynomial subexpression. It has much fewer nodes and more flexible adaptability than existing neural network algorithms. However, such algorithms lead to overfitting problems due to noise sensitivity as well as excessive trainning while generation of successive network layers. To alleviate such overfitting problem and also effectively design its ensuing deep network structure, two techniques are introduced. That is we use the two techniques of both SSC(Sum of Squared Coefficients) and L₂ regularization for consecutive generation of each layer"s nodes as well as each layer in order to construct the deep PNN structure. The technique of L₂ regularization is used for the minimum coefficient estimation by adding penalty term to cost function. L₂ regularization is a kind of representative methods of reducing the influence of noise by flattening the solution space and also lessening coefficient size. The technique for the SSC is implemented for the minimization of Sum of Squared Coefficients of polynomial instead of using the square of errors. In the sequel, the overfitting problem of the deep PNN structure is stabilized by the proposed method. This study leads to the possibility of deep network structure design as well as big data processing and also the superiority of the network performance through experiments is shown.

영어 학습자들은 영어단어 재인 시에 철자와 소리 간 규칙성 정보를 사용하는가?

류재희,남기춘,김다희,백연지 한국인지및생물심리학회 2016 한국심리학회지 인지 및 생물 Vol.28 No.1

The present study examined whether Korean learners of English use grapheme-to-phoneme correspondence rule as in native English speakers during English word recognition. In visual word recognition, both word frequency and regularity play a major role in reading speed and reading accuracy. Hence, the purpose of this study was to investigate how spelling-to-sound regularity and word frequency influence performance on explicit reading and silent reading tasks among English second language learners. Previous word recognition studies with English monolinguals have reported word regularity effect, in which regular words(e.g., save) are recognized faster with lower error rates compared to irregular words(e.g., have). Word regularity effect has been widely used as the supporting evidence for dual route model in visual word recognition. In Experiment 1, an explicit reading task(English word naming task) was used to examine the presence of word regularity effect during read-aloud process. In Experiment 2, a silent reading task(lexical decision task) was administered to examine the influence of irregularities in spelling and sound when participants were not required to generate sound information from spellings during English word recognition. Results from Experiment 1 demonstrated a significant interaction between word frequency and regularity, where word frequency effect was significant in all experimental conditions. On the other hand, word regularity effect showed marginal significance in low frequency word condition only. Results from Experiment 2 only revealed a significant word frequency main effect. Overall, these results indicate that Korean English L2 learners also seem to be able to use spelling-to-sound information during English word recognition similar to those of English monolinguals. However, rather than actively making use of this information via phonological(indirect) route, they have the tendency to use lexical(direct) route which is more sensitive to spelling information. The current study re-evaluated the English word regularity effect among Korean learners of English in terms of their use of spelling-to-sound information during English word recognition. 본 연구는 한국인 영어 학습자들이 영어 모국어화자와 유사한 방식으로 철자와 소리 간 규칙성 정보를 사용하여 영어단어를 재인하는지 알아보기 위해 두 가지 읽기과제(단어 명명과제, 어휘판단과제)를 실시하였다. 영어단어 재인 시, 단어의 빈도와 규칙성 여부가 읽기의 속도와 정확도에 영향을 미치는데 이 과정에서 철자-소리 대응관계 규칙(grapheme-to-phoneme correspondence rule)이 어떠한 역할을 하는지 알아보고자 하였다. 영어 모국어화자의 경우, 철자-소리 대응관계 규칙에 따라 규칙적인 단어(예: save)가 불규칙적인 단어(예: have)에 비해 더 빠르게 재인되고 오류율도 낮다. 이러한 현상을 단어 규칙성 효과(regularity effect)라고 하는데 이는 시각적 단어재인에서 이중경로모형(dual-route model)을 지지하는 근거로 사용된다. 따라서 한국인 영어 학습자들도 영어단어의 철자-소리 대응규칙에 따른 규칙성 효과가 발생하는지 알아보기 위해 실험 1에서는 음독과정을 거치는 단어 명명과제(word naming task)를 사용하였고, 실험 2에서는 묵독과정을 거치는 어휘판단과제(lexical decision task)를 사용하였다. 실험 1의 결과, 단어 빈도와 규칙성 조건 간 상호작용 효과가 나타났으며, 단어 빈도 효과는 모든 조건에서 유의미했으나, 규칙성 효과는 저빈도 단어 조건에서만 유의미한 수준의 경향성을 보였다. 실험 2의 결과, 단어의 소리정보를 생성하지 않아도 되는 어휘판단과제에서는 단어 빈도 효과만 유의미하였고 규칙성 효과는 발생하지 않았다. 이러한 연구 결과로 미루어 볼 때, 한국인 영어 학습자들은 영어단어 재인 시, 철자-소리 대응규칙 정보를 사용할 수는 있으나 이를 적극적으로 활용하는 음운경로(간접경로)를 이용하기보다는 철자 유사성에 기초한 어휘경로(직접경로)를 주로 사용하는 것으로 보인다.

An Adaptation Method in Noise Mismatch Conditions for DNN-based Speech Enhancement

( Xu Si-ying ),( Niu Tong ),( Qu Dan ),( Long Xing-yan ) 한국인터넷정보학회 2018 KSII Transactions on Internet and Information Syst Vol.12 No.10

The deep learning based speech enhancement has shown considerable success. However, it still suffers performance degradation under mismatch conditions. In this paper, an adaptation method is proposed to improve the performance under noise mismatch conditions. Firstly, we advise a noise aware training by supplying identity vectors (i-vectors) as parallel input features to adapt deep neural network (DNN) acoustic models with the target noise. Secondly, given a small amount of adaptation data, the noise-dependent DNN is obtained by using L2 regularization from a noise-independent DNN, and forcing the estimated masks to be close to the unadapted condition. Finally, experiments were carried out on different noise and SNR conditions, and the proposed method has achieved significantly 0.1%-9.6% benefits of STOI, and provided consistent improvement in PESQ and segSNR against the baseline systems.

통합 베이즈 총변이 정규화 방법과 영상복원에 대한 응용

류재흥 한국전자통신학회 2022 한국전자통신학회 논문지 Vol.17 No.1

본 논문은 통합 베이즈 티코노프 정규화 방법을 총변이 정규화에 대한 해법으로 제시한다. 통합된 방법은 총변이 항을 가중된 티코노프 정규화 항으로 변형하여 정규화 모수를 구하는 공식을 제시한다. 정규화 모수를 구하고 이를 바탕으로 새로운 가중인수를 구하는 것을 복원된 영상이 수렴하기까지 반복한다. 실험결과는 영상 복원 문제에 대하여 제안하는 방법의 효능을 보여준다.

Statistical Consistencies in the Spelling of English Sounds in English Textbooks

Yongeun Lee 한국중원언어학회 2008 언어학연구 Vol.- No.12

It is well-known that English spelling is highly inconsistent in terms of sound-to-spelling (e.g., /?/ →salt, caught, lawn, etc.) and spelling-to-sound (e.g., a→spa, change, bald, etc.) correspondences. Kessler and Treiman (2001), however, report that the correspondence becomes more consistent when context is taken into account. For example, the spelling i is more likely to be pronounced /?/ before t as in bit than before nd or ld as in mind or wild. In the current study, I ask how extensive such context effect is in high-frequency English words that Korean learners of English encounter. This is done by performing statistical analyses on consistencies in the spelling of consonants and vowels in English words consisting of an onset, a vowel and a coda which appea in seven middle-school English textbooks. Consistent with previous studies, the results indicate that vowel-spelling consistency is the least regular but the consistency of vowel spelling increases when the onset and coda spelling is considered. Similarly, onset- and coda-spelling consistency increase when the pronunciation of the vowel is taken into account. Implications of the findings are discussed in terms of acquisition of statistical regularities in L2 languages.

Robust and Fast Tracking via Joint Collaborative Representation

Fei Zhou,Guizong Zhang,Xinyue Fan,Dandan Yi 보안공학연구지원센터 2016 International Journal of Hybrid Information Techno Vol.9 No.4

In this paper, we present a robust and fast tracking method based on joint collaborative representation. Traditional sparse coding based tracking methods code the candidates as a sparse linear combination of a series of object and trivial templates and perform time consuming L1 regularizations. In contrast to these methods, this paper adopts the L2-regularized least square models to reduce the computational complexity. The tracked object can be represented by the linear combination of a series of object templates, and also can be represented by candidate samples in the current frame. We propose a joint objective function to handle the tracking process. In addition, we introduce an effective update scheme to deal with the change of target appearance over time. Experiments on several challenging image sequences show that our proposed tracking method is robust and efficient.