한국인 영어학습자의 의미역 및 화용적 위반 문장 처

정원일(Chung, Wonil),박명관(Park, Myung-Kwan) 한국영어학학회 2011 영어학연구 Vol.17 No.3

In this paper we examined how Korean learners of English process presumably semantic relationships by employing the event-related potentials (ERP) paradigm. The participants read simple sentences presented word-by-word and made plausibility judgments on three sentence types: no violation, thematic-role animacy violation, and non-thematic-role pragmatic violation. ERPs were measured at critical verbs that assign a thematic role of AGENT to their subject NPs, where the latter are integrated with the former. In the previous studies on L1 processing, thematic-role animacy violation elicited P600 effects as an index of syntactic/grammatical violation, whereas non-thematic-role pragmatic violation elicited N400 effects as an index of semantic/pragmatic violation. In this study on L2 processing, non-thematic-role pragmatic violation (For breakfast the boys would only bury...) elicited significant N400 effects. In thematic-role animacy violation (For breakfast the eggs would only eat...), however, there were no effects of N400 or P600. In short, unlike L1 speakers, Korean L2 learners have great difficulty in syntactic/grammatical processing. We suppose that this difficulty arises owing to the following factors. First, L2 learners tend to use semantic rather than syntactic information in sentence processing. Second, L1 is acquired implicitly in children, but L2 is learned explicitly in formal classrooms. Finally, with maturational changes, late L2 learners use a more declarative memory system than a procedural one in L2 syntactic/grammatical processing.

언어 학습자로서의 신경망 언어모델: 영어 주어-동사 일치를 중심으로

정원일 ( Chung Wonil ),김유희 ( Kim Euhee ),박명관 ( Park Myung-kwan ) 서강대학교 언어정보연구소 2021 언어와 정보 사회 Vol.44 No.-

This paper assesses the language-processing ability of the neural language model (L2-GPT-2) trained on data sets of English textbooks published in Korea using the Generative Pre-trained Transformer (GPT)-2. Assuming that the language model (LM) is also an (artificial) language learner, we test it focusing on subject-verb agreement in English. It is a well-established fact that L1 speakers exhibit facilitatory interference effects in ungrammatical sentences with a plural subject and a singular form of verb. Unlike human native speakers, L2-GPT-2 as well as L1-GPT-2 display such effects in ungrammatical sentences either with a plural subject and a singular form of verb, or with a singular subject and a plural form of verb. Though there is a significant difference between human speakers and neural LMs in processing subject-verb agreement, the two LMs’ sensitivity to interference by a distractor NP points to the fact that they can attain a remarkably human-like linguistic generalization on subject-verb agreement.

유아 대상 언플러그드 컴퓨팅 교육 사례

정원일(Chung, Wonil),지정근(Ji, Jung Geun),고석주(Koh, Seok-Joo) 한국HCI학회 2019 한국HCI학회 학술대회 Vol.2019 No.2

최근 어린이를 대상으로 한 코딩 교육은 컴퓨터 없이 컴퓨팅 사고력을 키우기 위한 교육들이 여러 교육기관에서 진행되어 오고 있다. 그러나 유치원생을 포함한 초등 저학년 이하의 교육 사례에 대한 관심은 증가하고 있지만 특별한 대안이 없는 경우가 많았다. 경북대학교 소프트웨어교육센터는 유아들을 대상으로 한 소프트웨어 가치확산 프로그램을 연극의 형태로 개발하여 실시해 오고 있다. 본 논문에서는 경력단절여성과 협력하여 과학과 언플러그드 컴퓨팅 교육을 접목한 연극의 개발과 실행에 관한 사례를 소개하고자 한다.

신경망 언어 모델과 인간 언어 사용자의 주어와 목적어 관계절 처리 비교 연구

전수경(Sookyeong Jeon),정원일(Wonil Chung),박명관(Myung-Kwan Park) 한국응용언어학회 2022 응용 언어학 Vol.38 No.2

This paper is to investigate the distinct aspects of processing subject and object relative clauses (SRC and ORC) in neural language models (LMs) and humans using the materials, which are constructed by manipulating two RC types (SRC vs ORC) and two intervening PP types (locative vs temporal). Surprise values are collected from the transformer GPT-2 and BERT language models. Reading time data for humans are taken from Lowder and Gordon’s (2021) eye-tracking experiment. According to Lowder and Gordon (ibid.) that take as a critical region the matrix verb after the RC in the subject position, for humans the locative PP contained in the RC register longer reading times ORCs than SRCs, while the temporal PP does so for SRCs than ORCs. By contrast, for the two neural LMs, surprisals are higher for ORCs than SRCs regardless of whether the PP in question is locative or temporal. There was no statistically significant linear fit between human and the LMs’ responses. The result shows that to the extent that the distinction between locative and temporal PPs in RCs is syntactico-semantic, neither of the two neural language models is able to acquire human-like sensitivity to such a distinction.

한국인 영어학습자의 wh-의존성 처리

박명관(Park, Myung-Kwan),정원일(Chung, Wonil) 한국영어학회 2011 영어학 Vol.11 No.4

Neural Signatures of Intervention Effects in Korean way ‘why’ Questions: An ERP Study

Park, Myung-Kwan(박명관),Wonil Chung(정원일),Daeho Chung(정대호) 한국언어학회 2021 언어 Vol.46 No.4

In this paper we use event-related potentials (ERPs) to investigate the time course of comprehending the so-called intervention effect (IE) that negative polarity items (NPIs) induce on the wh-adverbial way ‘WHY’ in Korean. In keeping with Tomioka’s (2009) presupposition-based account for Ko’s (2005) paradigm of empirical data on the IE of NPIs on WHY, we construct three types of minimal pairs involving the WHY - NPI or the reverse orders. Type [1] with an NPI in the matrix clause and with WHY in the embedded clause elicited sustained positivity or P3b and P600. Type [2] with an NPI and WHY in the embedded clause evoked only P3b. Type [3] with negation on an embedded verb complex and Q-particle on a matrix one recorded sustained negativity. Following the lead by Jouralvlev et al. (2016), the P3b is taken as a neural marker of accommodating the presupposition or focus reduction on the NPI occurring before WHY, while the P300 is a neural signature of extending the domain of presupposition from embedded to matrix clause. On the other hand, the sustained negativity observed in Type [3] is a neural index of retaining the result from processing the embedded anomalous NPI-WHY order. Taken together, from experimental perspectives this paper demonstrates that presupposition accommodation interacting with word order and structural make-up plays an instrumental role in accounting for the various aspects of the IE in WHY questions.

Processing NPIs in Korean

Myung-Kwan Park(박명관),Euiyon Cho(조의연),Wonil Chung(정원일) 한국언어학회 2017 언어 Vol.42 No.4

This paper used the neuro-linguistic, event-related potential (ERP) experimental methodology to investigate the sentence processing of the two negative polarity items (NPIs) in Korean such as amwu+N+to ‘any + N’ and te isang ‘any longer/further.’Our starting point is the descriptive and theoretical analyses of the two NPIs at issue. They are established as NPIs that require a licensing element for their distribution within a sentence, but they differ in terms of the type of licensing elements. Manipulating the structural environments, we recorded the neural responses to the verbal complexes that can or cannot license the NPIs. We found that, first, N400 was evoked by both of the NPIs in the illegal environments like a positive clause. Since N400 is regarded as a neural index of incomplete semantic integration, it follows that the Korean NPIs’ licensee-licensor relation is resolved via semantic processes. Second, amwu+N+to–containing question and -ki cen-ey ‘before’clauses as well as positive clauses elicited N400, whereas te isang–containing question and -ki cen-ey ‘before’ clauses elicited anterior P600. We take the latter anterior P600 component to reflect not a violation of NPI licensing but a cognitive load of discourse/pragmatic processing due to the lexical meaning of te isang. Third, directly comparing the two NPIs in terms of neural profiles, we find that amwu+N+to undergoes a full gamut of semantic retrieval and integration, thus cognitively more demanding than te isang in the course of their processing.

형태-통사 및 의미 위반의 청각적 문장처리: 사건관련전위 기반 연구

박명관(Park, Myung-Kwan),조의연(Cho, Euyon),정원일(Chung, Wonil) 한국생성문법학회 2016 생성문법연구 Vol.26 No.4

The present study examined auditory event-related potential (ERP) responses to sentence-embedded morpho-syntactic (inflectional honorific marker) and semantic (selectional restriction (SR)) violations. Morpho-syntactic violations in the present study were found to evoke a left anterior negativity peaking at around 400 ms. SR-related semantic anomalies elicited an N400 as well as a late posteriorly distributed N600. The shorter latency and different distribution of the negativity observed in response to morpho-syntactic violations compared to SR-related semantic anomalies indicate that the observed negativity is functionally different, suggesting that morpho-syntactic information is processed prior to lexical-semantic information. The finding of an N600 in response to SR semantic anomalies suggests that the latter component reflects processes of so-called reanalysis that are based on both semantic and syntactic aspects of the sentence. The study in this paper thus adds to the accumulating evidence for finer-grained differentiation of the brain responses, distinguishing morpho-syntactic from semantic-syntactic processes in sentence comprehension.

한국인들과 영어원어민들의 영어 고모음 인지에 관한 사건관련 뇌전위 연구

윤영도(Yun, Yungdo),박명관(Myung-Kwan Park),정원일(Wonil Chung) 언어과학회 2018 언어과학연구 Vol.0 No.86

This study examined English and Korean speakers" brain responses to English high vowels. An ERP experiment was conducted using synthesized vowels to obtain brain waves and find their differences called mismatch negativity (MMN). The experiment results showed that both English and Korean speakers" brain waves had negative peaks around the same time after they heard the stimuli. However, the MMN peak of the English speakers was higher during longer periods than that of the Korean speakers. This is assumed to be a typical pattern for native speakers when they involuntarily hear phonemes of their mother tongue. There was no statistical difference between the Korean speakers" brain responses based on English proficiency levels.

Subject vs Object Asymmetry in Korean Syntax: Evidence from an ERP Study of Processing Stranded Numeral Classifiers

박명관(Myung-Kwan Park),오은정(Eunjeong Oh),정원일(Wonil Chung) 한국생성문법학회 2020 생성문법연구 Vol.30 No.3

This paper investigates the sentence-processing aspects of stranded numeral classifiers (SNCs) in Korean and explores their implications. Building on the previous theoretical and experimental studies of SNCs, particularly concentrating on the subject vs. object asymmetry in SNCs, we hypothesize that subject-related SNCs are not properly licensed syntactically, hence they associate with the adjacent illegitimate NPs. By contrast, object-related SNCs are properly licensed, thus they associate with the distant associate NPs across intervening adjacent NPs. Thus, the former are predicted to give rise to semantic anomaly, whereas the latter are predicted to display priming effects induced by the associate NPs. Furthermore, the former skip the second pass of syntactic reanalysis with the associate NPs after the first pass of semantic incongruence, but the latter do enter into such reanalysis after the first phase of prime effects. In this paper we use the temporal-resolution-wise excellent event-relate potential (ERP) paradigm to test these hypotheses. Subject-related SNCs recorded N400-like negativity at anterior regions, whereas object-related ones registered reduced N400-like negativity at left anterior regions. In addition, the latter additionally evoked P600 at left anterior regions. Taken together, the ERP results confirmed the predictions made above, showing that SNC processing involves the intricate interaction of both semantic and syntactic integration.