InferSent를 활용한 오픈 도메인 기계독해

An open domain machine reading comprehension is a model that adds a function to search paragraphs as there are no paragraphs related to a given question. Document searches have an issue of lower performance with a lot of documents despite abundant research with word frequency based TF-IDF. Paragraph...

Full description

Saved in:
Bibliographic Details
Published in스마트미디어저널 Vol. 11; no. 10; pp. 89 - 96
Main Author 김정훈(Jeong-Hoon Kim), 심춘보(ChunBo Sim), 김준영(Jun-Yeong Kim), 박준(Jun Park), 박성욱(SungWook Park), 정세훈(Se Hoon Jung)
Format Journal Article
LanguageKorean
Published 한국스마트미디어학회 30.11.2022
Korean Institute of Smart Media
Subjects
Online AccessGet full text
ISSN2287-1322

Cover

Abstract An open domain machine reading comprehension is a model that adds a function to search paragraphs as there are no paragraphs related to a given question. Document searches have an issue of lower performance with a lot of documents despite abundant research with word frequency based TF-IDF. Paragraph selections also have an issue of not extracting paragraph contexts, including sentence characteristics accurately despite a lot of research with word-based embedding. Document reading comprehension has an issue of slow learning due to the growing number of parameters despite a lot of research on BERT. Trying to solve these three issues, this study used BM25 which considered even sentence length and InferSent to get sentence contexts, and proposed an open domain machine reading comprehension with ALBERT to reduce the number of parameters. An experiment was conducted with SQuAD1.1 datasets. BM25 recorded a higher performance of document research than TF-IDF by 3.2%. InferSent showed a higher performance in paragraph selection than Transformer by 0.9%. Finally, as the number of paragraphs increased in document comprehension, ALBERT was 0.4% higher in EM and 0.2% higher in F1. 오픈 도메인 기계독해는 질문과 연관된 단락이 존재하지 않아 단락을 검색하는 검색 기능을 추가한 모델이다. 문서 검색은 단어 빈도 기반인 TF-IDF로 많은 연구가 진행됐으나 문서의 양이 많아지면 낮은 성능을 보이는 문제가 있다. 아울러 단락 선별은 단어 기반 임베딩으로 많은 연구가 진행됐으나 문장의 특징을 가지는 단락의 문맥을 정확히 추출하지 못하는 문제가 있다. 그리고 문서 독해는 BERT로 많은 연구가 진행됐으나 방대한 파라미터로 느린 학습 문제를 보였다. 본 논문에서는 언급한 3가지 문제를 해결하기 위해 문서의 길이까지 고려한 BM25를 이용하며 문장 문맥을 얻기 위해 InferSent를 사용하고, 파라미터 수를 줄이기 위해 ALBERT를 이용한 오픈 도메인 기계독해를 제안한다. SQuAD1.1 데이터셋으로 실험을 진행했다. 문서 검색은 BM25의 성능이 TF-IDF보다 3.2% 높았다. 단락 선별은 InferSent가 Transformer보다 0.9% 높았다. 마지막으로 문서 독해에서 단락의 수가 증가하면 ALBERT가 EM에서 0.4%, F1에서 0.2% 더 높았다.
AbstractList An open domain machine reading comprehension is a model that adds a function to search paragraphs as there are no paragraphs related to a given question. Document searches have an issue of lower performance with a lot of documents despite abundant research with word frequency based TF-IDF. Paragraph selections also have an issue of not extracting paragraph contexts, including sentence characteristics accurately despite a lot of research with word-based embedding. Document reading comprehension has an issue of slow learning due to the growing number of parameters despite a lot of research on BERT. Trying to solve these three issues, this study used BM25 which considered even sentence length and InferSent to get sentence contexts, and proposed an open domain machine reading comprehension with ALBERT to reduce the number of parameters. An experiment was conducted with SQuAD1.1 datasets. BM25 recorded a higher performance of document research than TF-IDF by 3.2%. InferSent showed a higher performance in paragraph selection than Transformer by 0.9%. Finally, as the number of paragraphs increased in document comprehension, ALBERT was 0.4% higher in EM and 0.2% higher in F1. 오픈 도메인 기계독해는 질문과 연관된 단락이 존재하지 않아 단락을 검색하는 검색 기능을 추가한 모델이다. 문서 검색은 단어 빈도 기반인 TF-IDF로 많은 연구가 진행됐으나 문서의 양이 많아지면 낮은 성능을 보이는 문제가 있다. 아울러 단락 선별은 단어 기반 임베딩으로 많은 연구가 진행됐으나 문장의 특징을 가지는 단락의 문맥을 정확히 추출하지 못하는 문제가 있다. 그리고 문서 독해는 BERT로 많은 연구가 진행됐으나 방대한 파라미터로 느린 학습 문제를 보였다. 본 논문에서는 언급한 3가지 문제를 해결하기 위해 문서의 길이까지 고려한 BM25를 이용하며 문장 문맥을 얻기 위해 InferSent를 사용하고, 파라미터 수를 줄이기 위해 ALBERT를 이용한 오픈 도메인 기계독해를 제안한다. SQuAD1.1 데이터셋으로 실험을 진행했다. 문서 검색은 BM25의 성능이 TF-IDF보다 3.2% 높았다. 단락 선별은 InferSent가 Transformer보다 0.9% 높았다. 마지막으로 문서 독해에서 단락의 수가 증가하면 ALBERT가 EM에서 0.4%, F1에서 0.2% 더 높았다.
Author 김정훈(Jeong-Hoon Kim), 심춘보(ChunBo Sim), 김준영(Jun-Yeong Kim), 박준(Jun Park), 박성욱(SungWook Park), 정세훈(Se Hoon Jung)
Author_xml – sequence: 1
  fullname: 김정훈(Jeong-Hoon Kim), 심춘보(ChunBo Sim), 김준영(Jun-Yeong Kim), 박준(Jun Park), 박성욱(SungWook Park), 정세훈(Se Hoon Jung)
BookMark eNpNjz9Lw0AYh2-oYK39DlkcA--9d3mvN5bin2qhg91DLrmDkJJA08W9gluXFlTqZodugg4d_ERN8h0s6OD0LA8Pv98Za-VFblusjdhTPheIp6xblqkBjlwSV9Rmcpg7O7u3-bzafnvNy6Z-3TXrjVc_vzerJ69aLqrdqn7be4f9x-FzUS0fm_XXOTtx0bS03T922OTqcjK48Ufj6-GgP_IzAvAjBCBBkltBmAgpJSFq7YxRQlgyzplAKAfqOC9Kgl6gSAOqIErQQhLHosMufrNZWs7TME_KaXjbvxsjIAKpo89BI__nPRSmCE1RZPHxkJ2FEjiAIK0g0CB-AGL2WD0
ContentType Journal Article
Copyright COPYRIGHT(C) KYOBO BOOK CENTRE ALL RIGHTS RESERVED
Copyright_xml – notice: COPYRIGHT(C) KYOBO BOOK CENTRE ALL RIGHTS RESERVED
DBID P5Y
SSSTE
JDI
DEWEY 302.23
657.84
DatabaseName Kyobo Scholar Journals
Scholar(스콜라)
KoreaScience
DatabaseTitleList
DeliveryMethod fulltext_linktorsrc
Discipline Anthropology
Social Sciences (General)
Business
DocumentTitleAlternate Open Domain Machine Reading Comprehension using InferSent
EndPage 96
ExternalDocumentID JAKO202206757610921
4010036970590
GroupedDBID .UV
P5Y
SSSTE
JDI
ID FETCH-LOGICAL-k600-a20063641e362d344462299fbb733e6bffb537f07228ad5857690275ad2e0dcc3
ISSN 2287-1322
IngestDate Fri Dec 22 12:03:21 EST 2023
Fri Aug 15 02:10:36 EDT 2025
IsOpenAccess true
IsPeerReviewed false
IsScholarly false
Issue 10
Keywords InferSent
Document Search
Document Reader
문서 독해
ALBERT
Open Domain Machine Reading Comprehension
문서 검색
오픈 도메인 기계독해
Language Korean
LinkModel OpenURL
MergedId FETCHMERGED-LOGICAL-k600-a20063641e362d344462299fbb733e6bffb537f07228ad5857690275ad2e0dcc3
Notes KISTI1.1003/JNL.JAKO202206757610921
OpenAccessLink http://click.ndsl.kr/servlet/LinkingDetailView?cn=JAKO202206757610921&dbt=JAKO&org_code=O481&site_code=SS1481&service_code=01
PageCount 8
ParticipantIDs kisti_ndsl_JAKO202206757610921
kyobo_bookcenter_4010036970590
PublicationCentury 2000
PublicationDate 2022-11-30
PublicationDateYYYYMMDD 2022-11-30
PublicationDate_xml – month: 11
  year: 2022
  text: 2022-11-30
  day: 30
PublicationDecade 2020
PublicationTitle 스마트미디어저널
PublicationTitleAlternate Smart media journal
PublicationYear 2022
Publisher 한국스마트미디어학회
Korean Institute of Smart Media
Publisher_xml – name: 한국스마트미디어학회
– name: Korean Institute of Smart Media
SSID ssib012146176
ssib036278714
ssib022315842
ssib051117086
ssib044760798
Score 1.8138007
Snippet An open domain machine reading comprehension is a model that adds a function to search paragraphs as there are no paragraphs related to a given question....
SourceID kisti
kyobo
SourceType Open Access Repository
Publisher
StartPage 89
TableOfContents 서론 Ⅱ. 관련연구 Ⅲ. 제안 방법 Ⅳ. 실험 및 결과 Ⅴ. 결론 REFERENCES
Title InferSent를 활용한 오픈 도메인 기계독해
URI https://scholar.kyobobook.co.kr/article/detail/4010036970590
http://click.ndsl.kr/servlet/LinkingDetailView?cn=JAKO202206757610921&dbt=JAKO&org_code=O481&site_code=SS1481&service_code=01
Volume 11
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV29b9QwFLeODgghIThAlI8qA5ZAp6D44sTxmFxzKkWFgUPqdkoujlQV3UntdSgDU5HYurQSoLLB0A0Jhg4s_Du9u_-B93y-JKBIfCyJ9fz8LPtZz79n-9mE3M9zP8lFomw3UbnNlefZqczAGPJUtLMcQLPA4OSNp_7aC76-6W02Gj8qp5b2xumjwavauJL_0SrQQK8YJfsPmi2EAgHSoF_4gobh-1c6fozRes9xeS-OaOjRqNOi8SqVksoOjTtUhjSUmuIBpaVJAQ25JgGIDFpYMOjSgGsJEqnItEojzAvhRyNHJ1zDhNyekRnxKrbFkkGoxYMsAeKRLdBCkNI1CajDVOODBEyEDh64QOGQVeyf6GqlztEscl4tcAX1LNwkoI0BK1kibAIWNiy1OVBzxEyfRay6FgJutLmDcWEy2-D_2ehf_2LfWXUcOxVrPX-8yMz7suZG7t9myuL84nr45BnWj-6WwAvr8SqDCy5DQ7vxOl4YNKZfTS_xHkAxBoCvwJMAHsBeltvYnAvfEaX_C2CYCUe_Vlq0DJwo9Cy28L8_SqtoqHeVXDFujBXOx-Q10tgeNcnlyqsb-01ycRFU0STL8yhwy8wku9YDc935w-uEF6N48vm7NXt_Mv1wOjs-sabvPs2O3lqTw4PJ6dH045l1fvbl_OvB5PDN7PjbDdLrxr3Omm3e8rC3AVLbCa5cuT5nCtqcuZxzvw1AKE9T4brKT_M89VyROwLamWTgwgpf4oZ6krWVkw0G7k2yNBwN1S1iKT_jwht4LAX0D3Yl4SoRmZJeDh4YTPvLZEV3UH-Y7b7s12gKGbDn-ujG4hFotdPnDsObmaTAWOzbf5Jwh1wqR99dsjTe2VP3AJuO0xWt_5_f6Gwp
linkProvider ISSN International Centre
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=InferSent%EB%A5%BC+%ED%99%9C%EC%9A%A9%ED%95%9C+%EC%98%A4%ED%94%88+%EB%8F%84%EB%A9%94%EC%9D%B8+%EA%B8%B0%EA%B3%84%EB%8F%85%ED%95%B4&rft.jtitle=%EC%8A%A4%EB%A7%88%ED%8A%B8%EB%AF%B8%EB%94%94%EC%96%B4%EC%A0%80%EB%84%90&rft.au=%EA%B9%80%EC%A0%95%ED%9B%88&rft.au=%EA%B9%80%EC%A4%80%EC%98%81&rft.au=%EB%B0%95%EC%A4%80&rft.au=%EB%B0%95%EC%84%B1%EC%9A%B1&rft.date=2022-11-30&rft.issn=2287-1322&rft.volume=11&rft.issue=10&rft.spage=89&rft.epage=96&rft.externalDBID=n%2Fa&rft.externalDocID=JAKO202206757610921
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2287-1322&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2287-1322&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2287-1322&client=summon