논문 상세보기

서울대학교 인지과학연구소> Journal of Cognitive Science> Assisting Authors to Convert Raw Products into Polished Prose

KCI등재SCOUPUS

Assisting Authors to Convert Raw Products into Polished Prose

Takumi Ito , Tatsuki Kuribayashi , Hayato Kobayashi , Ana Brassard , Masato Hagiwara , Jun Suzuki , Kentaro Inui
  • : 서울대학교 인지과학연구소
  • : Journal of Cognitive Science 21권1호
  • : 연속간행물
  • : 2020년 03월
  • : 101-137(37pages)
Journal of Cognitive Science

DOI


목차

1. Introduction
2. The Sentence-level Revision task
3. The SMITH dataset
4. Analysis of the SMITH dataset
5. Experiments
6. Results
7. Related work
8. Conclusion and future work
Acknowledgements
References

키워드 보기


초록 보기

Being a notoriously complex problem, writing is generally decomposed into a series of subtasks: idea generation, expression, revision, etc. Given some goal, the author generates a set of ideas (brainstorming), which he integrates into some skeleton (outline, text plan, outline). This leads to a first draft which is submitted then for revision possibly yielding changes at various levels (content, structure, form). Having made a draft, authors usually revise, edit, and proofread their documents.
We confine ourselves here only to academic writing, focusing on sentence production. While there has been quite some work on this topic, most writing assistance has mainly dealt with grammatical errors, editing and proofreading, the goal being the correction of surface-level problems such as typography, spelling, or grammatical errors.
We broaden the scope by also including cases where the entire sentence needs to be rewritten in order to express properly all of the information planned. Hence, Sentence-level Revision (SentRev) becomes part of our writing assistance task.
Obviously, systems performing well in this task can be of considerable help for inexperienced authors by producing fluent, well-formed sentences based on the user’s drafts.
In order to evaluate our SentRev model, we have built a new, freely available crowdsourced evaluation dataset which consists of a set of incomplete sentences produced by nonnative writers paired with final version sentences extracted from published academic papers. We also used this dataset to establish baseline performance on SentRev.

UCI(KEPA)

I410-ECN-0102-2021-100-000944562

간행물정보

  • : 사회과학분야  > 심리학
  • : KCI등재
  • : SCOPUS
  • : 계간
  • : 1598-2327
  • : 1976-6939
  • : 학술지
  • : 연속간행물
  • : 2000-2022
  • : 321


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.

23권2호(2022년 06월) 수록논문
최근 권호 논문
| | | |

KCI등재 SCOPUS

1Word Associations and Depression In Adolescents

저자 : Hyeon Cho

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 23권 2호 발행 연도 : 2022 페이지 : pp. 123-149 (27 pages)

다운로드

(기관인증 필요)

초록보기

This study examines how words are associated in adolescents with depressive symptoms. For the main purpose of this paper two research questions were raised. The first question was whether and how a depressive disorder would be related to the words associated when being primed by a school-related stimulus item. The second question was whether age and gender would be related. In an attempt to respond to these two major concerns, a word association task was administered. Results showed that there was a similar ranking pattern in the most frequent words produced by all three groups of participants: the typical, potential, and risky groups. By contrast, the frequency ratios of a particular set of associated words were different at a statistically significant level across the three groups. As regards these main findings, this study proposed that a semantic relationship, a semantic network, and semantics of emotional words should be taken into consideration in interpreting the words associated most frequently in response to school-related stimuli in adolescents with depressive symptoms.

KCI등재 SCOPUS

2Heritage Speakers' Phonological Memory: Beyond Sub-Lexical Level of Sound Perception

저자 : Sun-young Lee

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 23권 2호 발행 연도 : 2022 페이지 : pp. 151-190 (40 pages)

다운로드

(기관인증 필요)

초록보기

This study investigated Korean heritage speakers' phonological knowledge compared with L2 learners and native controls. The results of a picture-word discrimination task using three alveolar stop sounds /t/, /th/ and /t'/ in Korean revealed radical differences between heritage speakers and L2 learners compared with native controls: heritage speakers' phonemic sensitivity to distinguish the three stop sounds was as robust as native speakers' even at the early stages of general oral proficiency whereas L2 learners' increased steadily as their general oral proficiency improved. In addition, heritage speakers' high performance was consistent regardless of conditions (e.g., tal for the image of 'the moon' in Match condition vs. thal or ttal in Mismatch condition) whereas L2 learners showed significantly lower accuracy rates in Mismatch condition, indicating they failed in rejecting mismatched stop sounds. Based on the results, we suggest that adult acquisition of phonology in heritage language is a process of retrieving and strengthening old phonological memories, which is different from encoding new sound memories in the case of typical adult second language learning.

KCI등재 SCOPUS

3A Longitudinal Study on the Effect of Length of Residence on Chinese Adults' Perception and Production of English /ɪ/-/i/

저자 : Ying Li

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 23권 2호 발행 연도 : 2022 페이지 : pp. 191-223 (33 pages)

다운로드

(기관인증 필요)

초록보기

It has been subject to debate regarding whether the increase of length of residence (LOR) in a second language (L2) community can facilitate the acquisition of L2 sounds or not (Flege & Fletcher, 1992; Jia et al., 2006). To shed further light on this issue, this study examined 34 Chinese speakers' acquisition of L2-English /I/-/i/ over 5 years, during which the participants varied in LORs in England. Their perception and production of /I/-/i/ were repeatedly tested with AXB discrimination tasks and read-aloud tasks in 2012 (Test1), 2013 (Test2), 2014 (Test3) and 2017 (Test4). After-test questions were asked to collect information on daily language use during the intervals of two tests. It was found that the participants' perception performance significantly improved during the first year of arrival but “fossilized” from then on. LOR was found to have played a significant effect on their perceptual improvement in Test2 but not in Test3 and Test4. However, findings from the production tests indicated that the “intensive” exposure to native English input during the first year of arrival did not significantly benefit participants' accuracies in producing /I/ in general; LOR was non-significant on their production of /I/ in any of the four tests. Overall, LOR was found to have significantly benefited participants' perception but not production at the first year of arrival, while it did not play a significant effect on perception or production from then on. Findings from the after-test questions suggested that LOR alone can hardly facilitate the acquisition of L2 sounds. Particularly, as hypothesized by SLM-r, the perceived L1-L2 distance and the quantity/quality of L2 input during the period of staying in an L2-speaking community are critical for the success of L2 sounds acquisition (Flege et al., 2021).

KCI등재 SCOPUS

4Lexicon and Grammar in Bilingual Individuals with Alzheimer's Disease and Parkinson's Disease: A Comparative Review

저자 : Miseon Lee , Eunkyu Lee

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 23권 2호 발행 연도 : 2022 페이지 : pp. 225-247 (23 pages)

다운로드

(기관인증 필요)

초록보기

Given the number of bilingual individuals worldwide, many of patients with Alzheimer's disease (AD) or Parkinson's disease (PD) are bilinguals. Yet little is known about the consequences of the two neurodegenerative diseases on the language of bilingual patients. In this paper, we review the available data in the bilingual literature, specifically the decline of language in patients with AD and PD at the lexical and grammatical levels. Our literature search of three electronic databases identified nine articles on bilingual patients with AD and five on those with PD that analyzed the lexicon and grammar of these patients. The findings of these studies are inconsistent but suggest that AD and PD should affect both languages in bilingual patients. They also show a trend that AD results in greater difficulties with lexicon and L2 grammar, while PD affects L1 grammar to a greater extent. This pattern is as expected by the declarative/procedural model that each disease affects a distinct memory system.

1
권호별 보기
같은 권호 다른 논문
| | | | 다운로드

KCI등재SCOUPUS

1Introduction to the Special Issue: AI at the Crossroads of NLP and Neurosciences

저자 : Michael Zock

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 21권 1호 발행 연도 : 2020 페이지 : pp. 1-13 (13 pages)

다운로드

(기관인증 필요)

키워드 보기
초록보기

KCI등재SCOUPUS

2Data-driven models and computational tools for neurolinguistics: a language technology perspective

저자 : Moscow Russia

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 21권 1호 발행 연도 : 2020 페이지 : pp. 15-52 (38 pages)

다운로드

(기관인증 필요)

초록보기

In this paper, our focus is the connection and influence of language technologies on the research in neurolinguistics. We present a review of brain imaging-based neurolinguistic studies with a focus on the natural language representations, such as word embeddings and pre-trained language models. Mutual enrichment of neurolinguistics and language technologies leads to development of brain-aware natural language representations. The importance of this research area is emphasized by medical applications.

KCI등재SCOUPUS

3A Unified Hierarchy for AI and Natural Intelligence through Auto-Programming for General Purposes

저자 : Juyang Weng

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 21권 1호 발행 연도 : 2020 페이지 : pp. 53-99 (47 pages)

다운로드

(기관인증 필요)

초록보기

Despite the current great public interest in neural network based artificial intelligence, there exists a huge gap between artificial intelligence (AI) and natural intelligence (NI). Furthermore, there is such a lack of united hierarchy of intelligence that intelligence is widely regarded piecemeal. This status quo results in highly brittle AI systems. This paper proposes how an autonomous agent, natural or artificial, develops a united intelligence hierarchy in the brain. The term "united" means not only for AI and NI both, but also for all practical sensory modalities and motor modalities, including perception, representation/reasoning, learning, societal activities and politics. This line of work has been supported by rigorous mathematical proofs and initial experimental verications.
However, this paper minimizes mathematical material so that the new in- formation here can reach a wide audience. We should ask a new and general question: "How can a machine, natural or artificial, Autonomously Program For General Purposes (APFGP) from the real physical world?"
We have given this question a solution, theoretically, experimentally, and mathematically. A clear but powerful learning engine|Developmental Network (DN) enables APFGP. Hopefully, understanding APFGP for both AI and NI not only fully automate constructions of AI systems but also improve human development, individually and societally.

KCI등재SCOUPUS

4Assisting Authors to Convert Raw Products into Polished Prose

저자 : Takumi Ito , Tatsuki Kuribayashi , Hayato Kobayashi , Ana Brassard , Masato Hagiwara , Jun Suzuki , Kentaro Inui

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 21권 1호 발행 연도 : 2020 페이지 : pp. 101-137 (37 pages)

다운로드

(기관인증 필요)

초록보기

Being a notoriously complex problem, writing is generally decomposed into a series of subtasks: idea generation, expression, revision, etc. Given some goal, the author generates a set of ideas (brainstorming), which he integrates into some skeleton (outline, text plan, outline). This leads to a first draft which is submitted then for revision possibly yielding changes at various levels (content, structure, form). Having made a draft, authors usually revise, edit, and proofread their documents.
We confine ourselves here only to academic writing, focusing on sentence production. While there has been quite some work on this topic, most writing assistance has mainly dealt with grammatical errors, editing and proofreading, the goal being the correction of surface-level problems such as typography, spelling, or grammatical errors.
We broaden the scope by also including cases where the entire sentence needs to be rewritten in order to express properly all of the information planned. Hence, Sentence-level Revision (SentRev) becomes part of our writing assistance task.
Obviously, systems performing well in this task can be of considerable help for inexperienced authors by producing fluent, well-formed sentences based on the user's drafts.
In order to evaluate our SentRev model, we have built a new, freely available crowdsourced evaluation dataset which consists of a set of incomplete sentences produced by nonnative writers paired with final version sentences extracted from published academic papers. We also used this dataset to establish baseline performance on SentRev.

KCI등재SCOUPUS

5Computational Representation of Chinese Characters: Comparison Between Singular Value Decomposition and Variational Autoencoder

저자 : Yu-hsiang Tseng , Shu-kai Hsieh

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 21권 1호 발행 연도 : 2020 페이지 : pp. 139-158 (20 pages)

다운로드

(기관인증 필요)

초록보기

Being a notoriously complex problem, writing is generally decomposed into a series of subtasks: idea generation, expression, revision, etc. Given some goal, the author generates a set of ideas (brainstorming), which he integrates into some skeleton (outline, text plan, outline). This leads to a first draft which is submitted then for revision possibly yielding changes at various levels (content, structure, form). Having made a draft, authors usually revise, edit, and proofread their documents.
We confine ourselves here only to academic writing, focusing on sentence production. While there has been quite some work on this topic, most writing assistance has mainly dealt with grammatical errors, editing and proofreading, the goal being the correction of surface-level problems such as typography, spelling, or grammatical errors.
We broaden the scope by also including cases where the entire sentence needs to be rewritten in order to express properly all of the information planned. Hence, Sentence-level Revision (SentRev) becomes part of our writing assistance task. Obviously, systems performing well in this task can be of considerable help for inexperienced authors by producing fluent, well-formed sentences based on the user's drafts.
In order to evaluate our SentRev model, we have built a new, freely available crowdsourced evaluation dataset which consists of a set of incomplete sentences produced by nonnative writers paired with final version sentences extracted from published academic papers. We also used this dataset to establish baseline performance on SentRev.

KCI등재SCOUPUS

6NLP's Clever Hans Moment has Arrived

저자 : Benjamin Heinzerling

발행기관 : 서울대학교 인지과학연구소 간행물 : Journal of Cognitive Science 21권 1호 발행 연도 : 2020 페이지 : pp. 159-167 (9 pages)

다운로드

(기관인증 필요)

초록보기

Large, pretrained language models have led to a flurry of new state-of-the-art results being reported in many areas of natural language processing. However, recent work has also shown that such models tend to solve language tasks by relying on superficial cues found in benchmark datasets, instead of acquiring the capabilities envisioned by the task designers. In this short opinion piece, I review a report by Niven & Kao (2019) of this so-called Clever Hans effect on an argument reasoning task and discuss possible solutions for its prevention.

1
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