논문 상세보기

대한언어학회> 언어학> Is c-command Machine-learnable?

KCI등재

Is c-command Machine-learnable?

Unsub Shin , Myung-kwan Park , Sanghoun Song
  • : 대한언어학회
  • : 언어학 29권1호
  • : 연속간행물
  • : 2021년 03월
  • : 183-204(22pages)
언어학

DOI


목차

1. Introduction
2. Background
3. Methods
4. Results
5. Discussion
6. Conclusion
References
Appendix A. Example items

키워드 보기


초록 보기

Many psycholinguistic studies have tested whether pronouns and polarity items elicit additional processing cost when they are not c-commanded. The previous studies claim that the c-command constraint regulates the distribution of relevant syntactic objects. As such, the syntactic effects of the c-command relation are greatly affected by the types of licensing (e.g. quantificational binding) and reading comprehension patterns of subjects (e.g. linguistic illusion). The present study investigates the reading behavior of the language model BERT when the syntactic processing of relational information (i.e. X c-commands Y) is required. Specifically, our two experiments contrasted the BERT comprehension of a c-commanding licensor versus a non-c-commanding licensor with reflexive anaphora and negative polarity items. The analysis based on the information-theoretic measure of surprisal suggests that violations of the c-command constraint are unexpected for BERT representations. We conclude that deep learning models like BERT can learn the syntactic c-command restriction at least with respect to reflexive anaphors and negative polarity items. At the same time, BERT appeared to have some limitations in its flexibility to apply compensatory pragmatic reasoning when a non-c-commanding licensor intruded in the dependency structure.

UCI(KEPA)

I410-ECN-0102-2022-700-000563772

간행물정보

  • : 어문학분야  > 언어학
  • : KCI등재
  • :
  • : 계간
  • : 1225-7141
  • : 2671-6283
  • : 학술지
  • : 연속간행물
  • : 1993-2022
  • : 1119


저작권 안내

한국학술정보㈜의 모든 학술 자료는 각 학회 및 기관과 저작권 계약을 통해 제공하고 있습니다.

이에 본 자료를 상업적 이용, 무단 배포 등 불법적으로 이용할 시에는 저작권법 및 관계법령에 따른 책임을 질 수 있습니다.

30권4호(2022년 12월) 수록논문
최근 권호 논문
| | | |

KCI등재

1조음과 청취에 근거한 수업을 통한 한국인 대학생의 영어모음 발음 변화

저자 : 강희조 ( Kang Hijo )

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 1-18 (18 pages)

다운로드

(기관인증 필요)

초록보기

This study tested whether the pronunciation of English vowels could improve through the explicit instruction based on the articulatory and auditory characteristics of English vowels, even during the pandemic, when oral practice was restricted. 30 university students who took an English pronunciation course participated in the study as the experimental group and they were taught English vowels with articulatory and auditory training. The 21 students in the control group listened to target sentences recorded by a native English speaker in an English listening and speaking course. At the beginning and at the end of the semester, the participants in both groups recorded 18 English sentences, four of which ([i], [ɪ], [ε], and [æ]) were acoustically analyzed. As Korean learners of English are known to have difficulties distinguishing those front vowels ([i] vs. [ɪ] and [ε] vs. [æ]), the F1 differences in each of those vowel pairs were statistically analyzed. The results showed that only female students belonging to the experimental group marked a marginally meaningful improvement in the vowel pair of [ε] and [æ]. The author concludes that it is possible to partially enhance the English vowel pronunciations of Korean students through implicit listening and explicit articulatory training. Teaching methods to improve students' pronunciation are suggested.

KCI등재

2한국어와 영어 수 단어의 빈도 분포 특성

저자 : 김선회 ( Kim Sun-hoi )

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 19-40 (22 pages)

다운로드

(기관인증 필요)

초록보기

The goal of this paper is to identify the distributional characteristics in the frequency of Korean and English number words through quantitatively analyzing their frequency data. The frequency distributions of number words were visualized and the correlation between the magnitude of numbers and the frequency of number words were measured through Spearman's and Kendall's correlation coefficients because the frequency distributions of number words did not follow the normal distribution. This paper shows that the main cross-linguistic characteristics in the frequency of number words, which were reported in Dehaene & Mehler (1992) and Jansen & Pollmann (2001), are also observed in Korean and English: the smaller the number, the more frequent the number words and the local increase effect of reference numbers on the frequency distributions. However, this paper additionally shows that the language-particular number system also affects the frequency distributions of number words.

KCI등재

3한국어 부사 부정극어 '통'과 '영'에 대한 연구: 한국어 교육의 관점에서

저자 : 김설연 ( Jin Xueyan )

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 41-62 (22 pages)

다운로드

(기관인증 필요)

초록보기

This paper aims to research Negative Polarity Items 'tong' and 'yeong' from the perspective of Korean language education. The analysis is done from three aspects: previous research, dictionary, and corpus. First of all, previous research revealed that different researchers have different opinions on the Syntagmatic Relation between 'tong' and 'yeong'. Besides, some problems occurred when vocabulary learning orders were set for Korean learners and no applicable scheme was put forward for Korean language education. Second, the analysis of dictionaries shows that 'tong' and 'yeong' are synonyms, which explains why Korean learners may find it difficult to distinguish them. Moreover the four dictionaries analyzed in this paper did not provide detailed word meanings, syntagmatic relation, and samples. Finally, actual usages of Korean native speakers are analyzed in an objective way and samples from a corpus are also explored. Generally speaking, 'yeong' is used more frequently, which indicates that native Korean speakers use 'yeong' more widely. Specifically speaking, 'tong' is often used in negative sentences while 'yeong' does not show much difference in its usage frequency in negative and positive sentences. This paper also shows that 'tong' and 'yeong' show a tendency of respective combination in sentences, which is explained in detail.

KCI등재

4비표준 종결어미 '-을까요?'와 '-실게요' 분석

저자 : 윤재학 ( Yoon Jae-hak ) , 김기국 ( Kim Gigook )

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 63-84 (22 pages)

다운로드

(기관인증 필요)

초록보기

Examining two non-standard sentence endings in Korean, we make an observation about the ulkkayo? form and make a case for a grammaticalization process for the silkeyyo form that provides a more intuitive explanation than in previous studies. An emerging non-standard use of the ulkkayo? form involves questions about something the listener must have direct access to, such as his or her own current state of mind or the number of his or her children, whereas the modality of the ulkkayo? form only describes a conjecture or surmise. It is illustrated that its anomaly derives from an incompatibility between questioned topics and the weak modal force of the form. The other non-standard sentence ending, the silkeyyo form, is an odd combination of the honorific marker si and ulkeyyo, the first person sentence ending, because rules prohibit self-honorification. It is hypothesized that this is the result of a grammaticalization process where a speaker utters the listener's expected answer, thereby shortening a conversational exchange in a hectic customer-service interaction. It is observed that the emergence of these non-standard forms reflects a society where politeness in the public sphere remains a high priority. Even when time efficiency forces a grammaticalization, it appears that a considerable effort is exerted to preserve a high level of politeness.

KCI등재

5한국어 보조사 '-도'의 초점영역확장에 대한 연구

저자 : 조기현 ( Joh Gi-hyun )

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 85-103 (19 pages)

다운로드

(기관인증 필요)

초록보기

This paper aims to investigate the focus domain expansion phenomenon in the Korean delimiter '-do' construction. For this, Chung(2021)'s oblique merge analysis is critically reviewed, the main idea of which is that thematic elements of a verb are formed as a complex focus cluster via the oblique merge operation in overt syntax. I point out that potential problems arise from the oblique merge operation itself, including binding, c-command, case licensing, and adverb interruption. I argue that the problems with Chung's oblique merge analysis are fundamentally caused by the systematic limit under the presupposed head-final structure for Korean. As an alternative, I propose that under the head-initial structure for Korean, the focus complex elements in question can be derived by raising the Del(imiter)P to the focus element, or a chain can be formed between the focus element and the DelP by the movement of the null operator of '-do,' in either case by QR at LF.

KCI등재

6영어 문제해결 은유와 사용

저자 : 조영순 ( Cho Youngsoon )

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 105-123 (19 pages)

다운로드

(기관인증 필요)

초록보기

The purpose of this study is to describe the conceptual metaphor of problem-solving in American English and to analyze its use based on the COCA corpus. Since the puzzle metaphor was explained as a characteristic problem-solving metaphor of American culture in Lakoff and Johnson (1980), no detailed linguistic research has followed. In this study, we reveal that there are two prevalent types of conceptualizations of PROBLEM and PROBLEMSOLVING in English: PROBLEM is a given task such as a puzzle, and PROBLEMSOLVING is fulfilling the task by finding the correct answer; PROBLEM is a danger, and PROBLEM-SOLVING is eliminating the danger with or without the correct answer. In the first type of metaphors, there emerges an entailment that the correct answer is in the problem itself, not in the outside environment. Analyzing the use of metaphors reveals that the machine breakdown metaphor and the disease metaphor are major alternative metaphors to the puzzle metaphor. The two metaphors ask us to find the correct answer before eliminating a danger, showing that, in American culture, finding the correct answer is crucial in problem-solving regardless of how a problem is conceived.

KCI등재

7A Grammar of Contrastive Stripping Construction with Subordinating Conjunctions in English

저자 : Sae-youn Cho , Taeho Kim , Han-gyu Lee

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 125-144 (20 pages)

다운로드

(기관인증 필요)

초록보기

The paper seeks to provide a syntactic and pragmatic account for the Contrastive Stripping Construction (CSC) in English. The pragmatic account will give a full support for our syntactic analysis. The issue here is on how to generate the CSC with a subordinate conjunction and how to give a right contrastive interpretation. Pointing out the problems of the Movement-Ellipsis analysis, we propose a construction-based analysis of the CSC, which interacts the English CSC rule with the lexical information on the lexemes such as though, except and whereas. It gives us a simple explanation on how to produce the CSCs with a subordinating conjunction and assign an appropriate contrastive interpretation to them.

KCI등재

8Second Culture Acquisition and Language Development: Cultural Sensitivity as an Individual Difference Factor

저자 : Mun-hong Choe

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 145-167 (23 pages)

다운로드

(기관인증 필요)

초록보기

This paper constitutes an initial part of the research agenda that aims to explore the relationship between individuals' multicultural sensitivity and their acquisition of a second language (L2). In order to lay the groundwork for a systematic inquiry into this complex problem, the present study first reviews previous works on theoretical and empirical validation of the construct of cultural sensitivity. The affective, cognitive, and behavioral correlates of cultural sensitivity that have been identified and discussed in the literature are recapitulated from a perspective of L2 research. Secondly, drawing on a pool of extant instruments to measure individuals' cultural sensitivity, we develop an alternative scale that better fits the goal of subsequent research efforts to investigate the role of cultural sensitivity in L2 acquisition. Lastly, we put forward some research questions that can be tackled effectively by means of the instrument developed here in the context of learning and teaching English as a foreign language.

KCI등재

9Neural Network Language Models as Psycholinguistic Subjects: Focusing on Reflexive Dependency

저자 : Wonil Chung , Myung-kwan Park

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 169-190 (22 pages)

다운로드

(기관인증 필요)

초록보기

The purpose of this study is to investigate the reflexive-antecedent dependency resolution accompanying the wh-filler-gap dependency resolution in neural network language models (LMs)' sentence processing, comparing the processing result of LMs to the one of humans. To do so, we adopt the psycholinguistic methodology that Fraizer et al. (2015) used for humans. The neural-network language models employed in this study are four LMs: the Long Short-Term Memory (LSTM) trained on large datasets, the Generative Pre-trained Transformer-2 (GPT-2) trained on large datasets, an LSTM trained on small datasets (L2 datasets), and the GPT-2 trained on small datasets (L2 datasets). We found that only the LMs trained on large datasets were sensitive to the dependency between a reflexive and its antecedent matching in gender, but all of the four neural LMs failed to learn reflexive-antecedent dependency accompanying wh-filler-gap dependency. Furthermore, we also found that the neural LMs have a learning bias in gender mismatch.

KCI등재

10Authorial Reference and Rhetorical Functions in Research Articles of Mathematics and Linguistics

저자 : Hoyoon Eun

발행기관 : 대한언어학회 간행물 : 언어학 30권 4호 발행 연도 : 2022 페이지 : pp. 191-210 (20 pages)

다운로드

(기관인증 필요)

초록보기

This study explores how academic authors of different fields present themselves in the text and how they build up authorial identities with rhetorical functions of authorial referential expressions. The corpus for this study is comprised of 72 research articles (RAs) written in English in the areas of mathematics and linguistics. The frequency of authorial referential expressions was analyzed statistically and their rhetorical functions were also investigated qualitatively, based on Hyland's (2002) classifications. The results report the significant overuse of first-person plural pronouns in the corpus of mathematics although the corpus consists of all single-authored RAs. Meanwhile, various types of authorial referential markers are used in the corpus of linguistics. Also, authors of linguistics select more diverse rhetorical functions than the ones of mathematics. We can say that a discipline influences the choice of authorial referential markers more significantly than a language. The study ends with suggestions that EAP learners and RA authors should be more aware of authorial reference and their rhetorical functions.

12
권호별 보기
같은 권호 다른 논문
| | | | 다운로드

KCI등재

1트위터에 나타난 접두어 '개-'의 사용 양상

저자 : 오효정 ( Oh Hyo-jung ) , 김종혁 ( Kim Chonghyuck )

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 1-23 (23 pages)

다운로드

(기관인증 필요)

초록보기

A few researchers have recently claimed that the prefix kay-, which literally means 'dog', has undergone a dramatic change over the past decade. Before the change, kaywas severely confined in terms of its syntax and semantics; it was restricted to combine with a handful of nouns, and semantically, it had the sole function of adding or boosting the negative sense of its host. Kay- is no longer used as such. It can virtually combine with a host of any category which has gradable meaning, as its semantics has acquired the function of an emphatic marker like very. A linguistic change is driven by the people who use it. As the number of people who adopt and use the new features of a linguistic item grows, the chance of its survival as a transformed linguistic entity increases. Thus, it is important to examine in a precise manner how kay- is used in order to track the course of change it has gone through in the past and to predict its survival in the future. No such effort, however, has been made so far. In this article, we show that we can obtain an extensive amount of data showing the real time usage of kay- over the past decade using techniques developed in informatics and that kay- has indeed undergone a dramatic change, as reported by other researchers. We also show, by analyzing the data extracted from tweets tweeted during the 48 hours, that kay- is much more frequently used in comtemporary Korean in the new pattern than in the old pattern, a concrete piece of evidence to back the claim that kay- is productively used.

KCI등재

2영한 표기법의 개선과 실제적 운용

저자 : 이종민 ( Lee Chongmin )

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 25-48 (24 pages)

다운로드

(기관인증 필요)

초록보기

This paper aims to review the given system of English-Korean transcription and to develop the revised version with the concepts of easiness and uniformity. Basically the vowel system should be simplified to reduce the syllable length in Korean, so the Korean people may use the imported words with uniformity. In the consonant system the positional variation of the same sound should be changed to be more consistent and recognizable among the Korean people. Often it has been assumed to maintain the idea that the foreign pronunciation should not be distorted. However, the proposed scheme has been devised to figure out how foreign words are transcribed more consistently. Though there are some exceptions existent in transcribed words, it can be concluded that the easiness and uniformity of English transcription can be systemized more effectively.

KCI등재

3COVID-19 영향으로 인한 한국 이공계 대학생의 영어 온라인수업에 대한 인식 연구

저자 : 차현지 ( Cha Hyun Ji ) , 김성희 ( Kim Sunghee )

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 49-69 (21 pages)

다운로드

(기관인증 필요)

초록보기

This study investigated how Korean college students majoring in science and engineering perceived their general English course held online for one semester during the COVID-19 pandemic. A survey in which 166 students participated was conducted at the end of the first semester of 2020. The results were as follows. First, most of the participants preferred pre-recorded online lectures delivered by their professors. The most common reason given was the flexibility of being able to watch them repeatedly without any concerns about time and place. Second, the preferred evaluation method was to submit only assignments without testing, since students believed this could prevent cheating. Third, the drawbacks of online classes were identified as difficulties in acquiring learning strategies from professors, problems with fair evaluation, overloading assignments, and difficulties in self-motivation due to the absence of peers. Fourth, while the majority of the students had negative opinions about online classes, there were a few positive answers. Thus, this study indicates that the participants had different preferences for and abilities to adapt to online classes. Therefore, students should reflect on their needs for online classes in the future. Discussing customized online classes tailored to each individual who takes part may be one requirement.

KCI등재

4코로나 19로 인한 비대면 온라인 강의 인식 조사: 대학생의 만족도와 불안도를 중심으로

저자 : 황요한 ( Hwang Yohan ) , 김창수 ( Kim Changsoo )

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 71-91 (21 pages)

다운로드

(기관인증 필요)

초록보기

The outbreak of the coronavirus has dramatically changed the way people live and inevitably transformed the practice of social distancing which has changed the forms and methods of education. The main purpose of this study is to examine the application and operation of non-face-to-face online classes in a university setting during COVID-19. In order to achieve this goal, this paper conducts an in-depth analysis of 255 university students' perceptions toward satisfaction and anxiety over online learning and new circumstances surrounding online education. The results and suggestions are as follows. The main factors that have a positive influence on students' satisfaction in an online classroom environment come from the easy access to course content and opportunities for review and the possibility of self-directed learning beyond time and space limitations. Second, in order to increase its effectiveness, both internal efforts to improve the quality of course contents and external support for student and class management have to be (re)considered. Based on the findings from crisis to opportunity, this study suggests the need for careful planning and implementation of new procedures and guidelines for inclusive and resilient online education systems in the aftermath of COVID-19.

KCI등재

5Resolution of Wh/Quantifier Scope Ambiguity in Native and Nonnative Language Processing

저자 : Youngjae An

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 93-113 (21 pages)

다운로드

(기관인증 필요)

초록보기

This study makes an attempt to identify how and when native (L1) and nonnative (L2) speakers of English make use of syntactic knowledge and nonsyntactic contextual information to resolve wh/quantifier scope ambiguity during online sentence processing. The question addressed in this research is investigated in the context of weak crossover phenomena (Chiercia, 1993). 19 native speakers of English and 32 Korean speakers of English participated in a self-paced reading experiment; the Korean speakers are further divided into advanced and intermediate group. The results show that L1 and L2 speakers are able to integrate the syntactic information with the contextual information in resolving wh/quantifier scope ambiguity online, suggesting that the ambiguity is influenced by the contextual information (Villalta, 2003). The findings also show that L2 proficiency is not a predictor of L2 speakers' use of the target syntactic knowledge online. More importantly, the results indicate that L1 and L2 speakers employ similar strategies in resolving wh/quantifier scope ambiguity online, supporting the view that L2 processing is not fundamentally different from L1 processing (An 2019; Dekydtspotter et al., 2006).

KCI등재

6Current Topics and Controversies in the Analysis of English L2 Learners' Verb Inflection Errors

저자 : Mun-hong Choe , Mi-ra Jung

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 115-137 (23 pages)

다운로드

(기관인증 필요)

초록보기

This article provides a synthetic review of current research on English L2 learners' acquisition of verb inflection, focusing specifically on four topic areas: tense and aspect, subject-verb agreement, lexical and auxiliary verbs, and individual differences. The primary goal is to address the question of what to consider when analyzing ungrammatical forms produced by learners. An additional goal is to gain further insights into such questions as what challenges interlanguage researchers are likely to face in the analysis of the sources of inaccurate forms and what kinds of errors can be regarded as (non-)systematic. This review results in several observations that merit further investigation. First, in the acquisition of tense-aspect morphemes, some general tendencies have been recognized, of which most notable is the fact that L2 learners' ability to use tense-aspect morphology tends to develop gradually while form acquisition normally precedes function acquisition. Second, L2 acquisition of subject-verb agreement has been a central issue in the debate over the role and representation of the innate language faculty. Third, regarding L2 learners' ability to discern lexical and auxiliary verbs, a prevalent UG-based argument is that the mental grammars of L2 learners are organised in the same way as those of L1 speakers and that they differ only in the nature of their lexemes for surface morphology. Finally, since verb inflectional morphology explicitly demonstrates language learners' grammatical competence, variations in its acquisition constitute the core of inquiry into age- and L1-related factors.

KCI등재

7A Correlation Study of Chinese English Learners' Writing Anxiety, Writing Strategies, and Writing Achievements

저자 : Ming-hao Jin , Yan Guo

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 139-160 (22 pages)

다운로드

(기관인증 필요)

초록보기

This study aims to investigate Chinese English learners' writing anxiety, and analyzes its correlation with their writing strategies and writing achievements. Two questionnaires, designed by referring to and re-editing Petric and Czarl's (2003) Writing Strategies Questionnaire and Cheng's (2004) Second Language Writing Anxiety Inventory, were completed by 90 Chinese senior high school students in Yanji, China. SPSS 24.0 was used to analyze the data and produce descriptive statistics, paired sample T-tests, and Pearson correlation coefficient analysis. The result shows that students have a moderate degree of anxiety. High anxiety students experience the highest cognitive anxiety, and low anxiety students experience the highest physical anxiety. English learners' writing anxiety shows a significant negative correlation with writing strategies. High anxiety students seldom use English writing strategies, while low anxiety students use them very often in the process of completing a writing task. English learners' writing anxiety shows a significant negative correlation with writing achievements as well. The higher English learners' writing anxiety is, the lower their writing achievements will be. Thus, English learners should be encouraged and trained to use English writing strategies well to reduce English writing anxiety and improve the level of writing achievements.

KCI등재

8Quotative be like in Ohio English

저자 : Seung Han Lee

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 161-182 (22 pages)

다운로드

(기관인증 필요)

초록보기

This study aims (1) to provide the frequency of occurrence of be like as quantitatively measured; (2) to discover a systematic sociolinguistic distribution of be like by specifically comparing use by gender, age, education, and employment; and (3) to shed light on the classification of speech introduced by be like under discourse accounts. Accordingly, we identified a total of 485 tokens from the Ohio Buckeye Corpus of spontaneous speech. As for the distribution, the results show first and foremost that speakers prefer tense agreement between be like and its quote, favouring the present tense to trigger dramatic effect in narratives. Be like leads to expansion into the third person, second person, and first person. In Ohio English, no gender bias is identified in the use of be like. Also, younger speakers use be like more frequently. Speakers' education and employment status are not in relationship to the occurrence of be like. In what follows, the speech be like introduces is classified into four types: internal speech, direct speech, external speech, and hypothetical speech. The most frequently occurring internal speech is again subcategorized into internal judgement, internal surprise, internal volition, and internal situation; speakers' own evaluation, surprise, volition, or objective description of the preceding context is verbally unuttered. Hypothetical speech is of special interest in that speakers deliver the quote of be like under assumption, in keeping with the given context. Last, expletive it described here must include the consideration of the preceding context in order to classify the quotes of be like properly.

KCI등재

9Is c-command Machine-learnable?

저자 : Unsub Shin , Myung-kwan Park , Sanghoun Song

발행기관 : 대한언어학회 간행물 : 언어학 29권 1호 발행 연도 : 2021 페이지 : pp. 183-204 (22 pages)

다운로드

(기관인증 필요)

초록보기

Many psycholinguistic studies have tested whether pronouns and polarity items elicit additional processing cost when they are not c-commanded. The previous studies claim that the c-command constraint regulates the distribution of relevant syntactic objects. As such, the syntactic effects of the c-command relation are greatly affected by the types of licensing (e.g. quantificational binding) and reading comprehension patterns of subjects (e.g. linguistic illusion). The present study investigates the reading behavior of the language model BERT when the syntactic processing of relational information (i.e. X c-commands Y) is required. Specifically, our two experiments contrasted the BERT comprehension of a c-commanding licensor versus a non-c-commanding licensor with reflexive anaphora and negative polarity items. The analysis based on the information-theoretic measure of surprisal suggests that violations of the c-command constraint are unexpected for BERT representations. We conclude that deep learning models like BERT can learn the syntactic c-command restriction at least with respect to reflexive anaphors and negative polarity items. At the same time, BERT appeared to have some limitations in its flexibility to apply compensatory pragmatic reasoning when a non-c-commanding licensor intruded in the dependency structure.

1
주제별 간행물
간행물명 최신권호

KCI등재

언어학
30권 4호

공공언어학
8권 0호

중한언어문화연구
23권 0호

KCI등재

국어교육학연구
56권 5호

KCI등재

국어교육학연구
57권 5호

KCI등재

언어와 정보
26권 2호

KCI등재 SCOUPUS

언어연구
39권 3호

KCI등재

국어교육학연구
57권 4호

KCI등재

기호학 연구
72권 0호

KCI등재

한국언어문화
79권 0호

KCI등재

Journal of Pan-Pacific Association of Applied Linguistics (Journal of PAAL )
26권 2호

KCI등재

일본어문학
95권 0호

KCI등재

텍스트언어학
53권 0호

KCI등재

이중언어학
90권 0호

KCI등재

언어와 문화
18권 4호

KCI등재

언어와 정보사회
47권 0호

KCI등재

중국학연구
102권 0호

KCI등재

외국어교육연구
36권 4호

KCI등재

언어와 언어학
98권 0호

KCI등재

언어연구
38권 3호
발행기관 최신논문
자료제공: 네이버학술정보
발행기관 최신논문
자료제공: 네이버학술정보

내가 찾은 최근 검색어

최근 열람 자료

맞춤 논문

보관함

내 보관함
공유한 보관함

1:1문의

닫기