Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
why is academic writing important for us
Search
Sho Yokoi
PRO
October 26, 2017
Research
3
4.3k
why is academic writing important for us
2017-10-26, 研究室内勉強会資料
(1) なぜライティングスキルは重要なのか
(2) 論文投稿先に関する基礎知識
Sho Yokoi
PRO
October 26, 2017
Tweet
Share
More Decks by Sho Yokoi
See All by Sho Yokoi
Zipf 白色化:タイプとトークンの区別がもたらす良質な埋め込み空間と損失関数
eumesy
PRO
8
1.3k
Embers of Autoregression: Understanding Large Language Models Through the Problem They are Trained to Solve
eumesy
PRO
7
1.3k
「確率的なオウム」にできること、またそれがなぜできるのかについて
eumesy
PRO
8
3.3k
A Theory of Emergent In-Context Learning as Implicit Structure Induction
eumesy
PRO
5
1.4k
ChatGPT と自然言語処理 / 言語の意味の計算と最適輸送
eumesy
PRO
25
18k
Revisiting Over-smoothing in BERT from the Perspective of Graph
eumesy
PRO
0
1.2k
構造を持った言語データと最適輸送
eumesy
PRO
5
7.5k
最適輸送と自然言語処理
eumesy
PRO
19
13k
言葉の形を教えてくれる自然言語処理
eumesy
PRO
1
1.7k
Other Decks in Research
See All in Research
한국어 오픈소스 거대 언어 모델의 가능성: 새로운 시대의 언어 이해와 생성
inureyes
PRO
0
220
Weekly AI Agents News! 1月号 アーカイブ
masatoto
1
160
複数データセットを用いた動作認識
yuyay
0
110
Weekly AI Agents News! 10月号 論文のアーカイブ
masatoto
1
500
Weekly AI Agents News! 11月号 論文のアーカイブ
masatoto
0
290
地理空間情報と自然言語処理:「地球の歩き方旅行記データセット」の高付加価値化を通じて
hiroki13
1
190
リモートワークにおけるパッシブ疲労
matsumoto_r
PRO
6
4.9k
機械学習でヒトの行動を変える
hiromu1996
1
530
TransformerによるBEV Perception
hf149
1
690
AWS 音声基盤モデル トーク解析AI MiiTelの音声処理について
ken57
0
130
[輪講] Transformer Layers as Painters
nk35jk
4
650
Data-centric AI勉強会 「ロボットにおけるData-centric AI」
haraduka
0
430
Featured
See All Featured
VelocityConf: Rendering Performance Case Studies
addyosmani
328
24k
Build The Right Thing And Hit Your Dates
maggiecrowley
34
2.5k
Done Done
chrislema
182
16k
Six Lessons from altMBA
skipperchong
27
3.6k
Facilitating Awesome Meetings
lara
51
6.2k
Building a Modern Day E-commerce SEO Strategy
aleyda
38
7.1k
A designer walks into a library…
pauljervisheath
205
24k
Why Our Code Smells
bkeepers
PRO
336
57k
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
PRO
12
960
Product Roadmaps are Hard
iamctodd
PRO
50
11k
Bash Introduction
62gerente
611
210k
Docker and Python
trallard
44
3.3k
Transcript
Why is Writing Important ݚڀͱษڧͷϧʔϧͷҧ͍ɼ͓Αͼจߘઌʹؔ͢Δجૅࣝ Research Skills ษڧձ #1; October
26th, 2017 ౦େֶ סݚڀࣨ ԣҪ (D1) 1
1. ͳͥʮจͷॻ͖ํʯΠγϡʔͳͷ͔ 2
ษڧͱݚڀతධՁͷํ๏ҟͳΔ • ษڧɿطͷݟͷशಘ͕తɽࢼݧϨϙʔτΛ௨ͯ͠ɼֶ शऀͷशख़ΛධՁɾݕূɽ • ݚڀɿਓྨʹͱͬͯະͷࣄ࣮ͷൃݟ͕తɽࠪಡͱҾ༻Λ௨ ͯ͠ɼओுʢจʣͷଥੑॏཁੑΛධՁɾݕূɽ → ݚڀ׆ಈͷ࣮ફతˍظతͳඪɼݚڀ݁ՌΛจʹ·ͱ ΊͯɼࠪಡΛύε͢Δ͜ͱɽݚڀࣨଐ͔Β1ʙ2ͰͨͲΓண
͖͍ͨɽ 3
ͳͥࠪಡ͢Δͷ͔ɼͳͥҾ༻͢Δͷ͔ ਓྨશମͰֶΛલਐͤ͞Δํ๏ৗʹΞοϓσʔτ͞Ε͖ͯ ͨɽݱࡏࠪಡͱҾ༻ʹΑͬͯݚڀͷ࣭Λ୲อ͢Δํ๏͕ओྲྀɽ • Peer ReviewʢࠪಡʣɿݚڀՌʢจʣͷॏཁੑ৽نੑΛ ઐՈಉ࢜Ͱ૬ޓݕূʢࠪಡʣ͢ΔɽࠪಡΛύεͨ͠จ͕ग़ ൛͞Εɼଞऀ͔ΒࢀরͰ͖Δঢ়ଶʹͳΔɽˡ ࠓճͷείʔϓ •
CitationʢҾ༻ʣɿઌߦݚڀΛ౿·͑ɼݞʹΓʢҾ༻͠ʣɼ ݟΛ͞ΒʹਐΊΔɽ·ͨҾ༻ʹΑΓઌߦݚڀܟҙΛࣔ͢ɽ 4
ࠪಡͰνΣοΫ͞ΕΔ߲ • ݚڀͷ༰ʹؔΘΔ߲ Noveltyʢ৽نੑʣɼOriginalityʢಠੑʣɿ৽͠͞ SignificanceʢॏཁੑʣɼRelevanceʢؔ࿈ੑʣɿॏཁ͞ Correctnessʢਖ਼ੑʣɼSoundnessʢଥੑʣɿٞͷଥ͞ • จͷॻ͖ํʹؔΘΔ߲ ← ॻ͖ํۃΊͯॏཁ
ClarityɼPresentationɿهड़ͷ໌ղ͞ɼٞͷ͍͢͞ Repeatabilityɿ࠶ݱੑʢʹಡΈख͕ࢼՄೳ͔ʣ 5
·ͱΊɿͳͥʮจͷॻ͖ํʯΠγϡʔͳͷ͔ • ݚڀ׆ಈʢਓྨͷΛલਐͤ͞Δ׆ಈʣͷεϞʔϧΰʔϧݚ ڀՌΛࠪಡ͖จͱͯ͠ग़൛͢Δ͜ͱɽ • ࠪಡͰจͷॻ͖ํ͕νΣοΫ͞ΕΔʢʹΑ͘ॻ͚͍ͯΔ จʹՁ͕͋Δʣɽ • →ʮจͷॻ͖ํʯॏཁɽ •
·ͨจͷ໌ղ͞Λ্ͤ͞ΔաఔͰɼݚڀࣗମ͕લਐ͢Δɽ 6
2. จߘઌʹؔ͢Δجૅࣝ 7
ߘઌ จͷߘઌʹଟ͘ͷબࢶ͕͋Δɽ • ࠪಡɿࠪಡͷ༗ແ • ഔମɿจࢽɼձٞͷ༧ߘूɼϫʔΫγϣοϓͷ༧ߘू • ݴޠɿӳޠʢࠃࡍࢽɼձٞʣʀຊޠʢࠃࢽɼձٞʣ • Tierɿܝࡌจͷ࣭ɼࠪಡͷݫ͠͞
8
ࠪಡ • ࠪಡͷ༗ແɿجຊతʹࠪಡ͖จͷΈ͕Ҿ༻ͷରͱͳΔɽ ݴ͍͑Εɼࠪಡͳ͠ͷจʢྫ͑ࠃձٞͷ༧ߘʣҾ ༻ͷରͱͳΒͳ͍ɽ • ಗ໊ੑɿެਖ਼ੑͷͨΊɼDouble-blindʢೋॏݕʀஶऀͱࠪಡ ऀ͕͓ޓ͍ΛΒͳ͍ʣ Single-blindʢยଆݕʀஶऀଆͩ ͚ࠪಡऀΛΒͳ͍ʣͰࠪಡ͞ΕΔ͜ͱ͕ଟ͍ɽզʑ͕ߘ
͢Δจࢽࠃࡍձٞ΄ͱΜͲ double-blind peer reviewɽ 9
ഔମ • Journal Articleʢݪஶจʣɿ௨ৗจࢽʹ࠾͞Εͨจ ͕ݪஶจʢҰ࣍ࢿྉʣͱݟͳ͞ΕҾ༻ͷରͱͳΔɽ·ͨ ͬͱॏཁͳۀͱͳΔɽࠪಡϲ݄͔Βఔɽ • Proceedings Paperʢձٞ༧ߘʣɿଟ͘ͷʹ͓͍ͯձٞڝ ૪తͰͳ࣭͘୲อ͞Ε͓ͯΒͣۀʹͳΒͳ͍ɽ͔͠͠
ਝͳࠪಡΛॏΜ͡ΔܭࢉػՊֶͷҰ෦Ͱࠃࡍձٞڝ૪త ͔ͭ࠷ॏཁࢹ͞ΕΔɽNLPಛʹݦஶɽࠪಡ1ʙ2ϲ݄ఔɽ 10
ഔମ • Preprintɿग़൛લͷݪߘΛެ։͢ΔαʔϏεʢPreprint serverʀ యܕతʹ arXivʣ͕ۙΜʹΘΕ͍ͯΔɽૣΊͷެ։Ͱ ৽نੑΛओுͰ͖ɼ·ͨۀքશମͷݚڀαΠΫϧૣ·Δɽ ※ ࣭୲อ͞Εͣۄੴࠞ߹ɽʢҾ༻ʹΑΔ୲อՄೳʣ ※
Double-blind Ͱͷࠪಡ͕࣮࣭తʹෆՄೳʹͳΔ͋Δɽ ACLίϛϡχςΟɼߘ1ϲ݄લҎޙʹϓϨϓϦϯτΛެ։ࡁ ͷจΛෆ࠾ʹ͢Δࢫએݴɽ 11
ݴޠ • զʑͷۀքͰɼجຊతʹӳޠͰॻ͔ΕͨจͷΈ͕Ҿ༻ͷର ͱͳΔɽ • ͨͩ͠ࠃจࢽɾࠃձٞͷߘʹɼۀҎ֎ʹଟ͘ͷ Ձ͕͋Δɽ ✔ จͷܗʹ·ͱΊɼ·ͨଞେֶଞݚڀػؔͷݚڀऀ͔Βί ϝϯτΛΒ͏͜ͱͰɼݚڀΛਐΊΔྑ͍ػձʹͳΔɽ
✔ ࠃͷϓϨʔϠʔʢಛʹඇݚڀऀʣͷ༗༻ͳࢀরઌʹͳΔɽ 12
Tier • ࠪಡ͕ڝ૪తͰ࠾จͷ࣭͕ߴ͍ഔମͱͦ͏Ͱͳ͍ͷ͕͋ Δɽ׳ྫతʹڝ૪తͳॱʹTop (1st) Tier, 2nd Tier, ͱΑͿɽ •
Top Tier ͷจࢽɾձٞɼࠪಡऀͱͯ͠ۀܦݧͷ͋Δݚ ڀऀׂ͕ΓͯΒΕΔ͜ͱ͕ଟ͘ɼࠪಡίϝϯτ༗ӹɽ → ͳΔ͘ྑ͍ձٞʹग़͠·͠ΐ͏ɽ • ಡΈखͱͯ͠ Tier ͷߴ͍จࢽɾձ͔ٞΒαʔϕΠ͢Δͷ͕ ޮతɽ 13
ܭࢉػՊֶͷࠃࡍձٞͷྫ NLP AI ML; DM; ΄͔ 1st Tier पล͔Β ࢀর͞ΕΔ
ACL, EMNLP, NAACL AAAI, IJCAI NIPS, ICML; KDD, WSDM; WWW, SIGIR, CVPR, InterSpeech 2nd Tier ͔֘Β ࢀর͞ΕΔ EACL, COLING, IJCNLP, CoNLL UAI, ECAI AISTATS, ICLR; ICDM, ECMLPKDD, CIKM 14
Long Paper, Short Paper ࠃࡍձٞɼLong Paperʢ6ʙ8ϖʔδఔʣͱ Short Paperʢ4ʙ6ϖʔδఔʣΛબΔέʔε͕͋Δɽ • ҰൠʹɼLong
Paper ʹ࣮ݧߟͳͲ͕ेʹἧͬͨݚڀ ΛɼShort Paper ʹΞΠσΞҰൃ࣮ݧ͕ݶఆ͞ΕͨݚڀΛ ߘ͢Δɼͱ͞Ε͍ͯΔɽ • ҰൠʹɼLong Paper ͷํ͕ڝ૪తͰ࠾จͷߴ͍ɽ 15
Oral Presentation, Poster Presentation ࠃࡍձٞʹจ͕࠾͞ΕΔͱɼձٞͰݚڀͷ༰Λൃද͢Δػ ձ͕༩͑ΒΕΔɽൃදଟ͘ͷ߹ٛɽ • ൃදͷܗଶʹ Oral Presentationʢޱ಄ൃදʣͱ
Poster Presentationʢϙελʔൃදʣ͕͋ΔɽҰൠʹɼ࠾จͷൃ දͷܗଶओ࠵ऀଆ͔Βࢦࣔ͞ΕΔɽ • Ұൠʹɼจͷ࣭͕ߴ͘ଟ͘ͷௌऺ͕ظ͞ΕΔݚڀ͕ Oral Presentation ʹׂΓͯΒΕΔɽ 16
NLPʹ͓͚ΔΑ͋͘Δߘॱ 1. ࠃձٞɿຊޠɼࠪಡͳ͠ɽۀʹͳΒͳ͍ɽจԽͷػ ձɼଞݚڀऀͱٞ͢ΔػձʹɽݴޠॲཧֶձɼNLݚͳͲɽ ࠃࡍձٞซઃϫʔΫγϣοϓಉ༷ͷϝϦοτ͕͋Γɼਪɽ 2. ࠃࡍձٞɿӳޠɼࠪಡ͋Γɽ͕͜͜ओઓɽNLP12݄͔Β4 ݄ࠒ͕ߘγʔζϯɼ6݄͔Β9݄ࠒ͕ձٞγʔζϯɽ 3. จࢽɿΞΧσϛοΫͳจ຺ͰධՁΛड͚ΔࡍʹॏཁɽTACL
࠾͞ΕΔͱACL/EMNLP/NAACLͰൃදՄɽ 17
3. ࠓޙͷ Research Skills ษڧձ ΑΓΑ͘ॻ͚ΔΑ͏ʹͳΔͨΊʹ 18
ษڧձͷείʔϓ είʔϓ είʔϓ֎ Α͍ॻ͖͔ͨΛֶͿ Α͍ςʔϚઃఆΛֶͿ How to say What to
say ୡՄೳͳٕज़ ͓ؾ࣋ͪɼҙ 19
ษڧձͰѻ͏ςʔϚ • πʔϧͷ͍ํɿLaTeX ͷ Tips ؚΊͨࣥචڥɼϊϋ ɼKWIC ͷपลπʔϧ • ӳޠͷॻ͖ํɿจతɾ׳ྫతͳݴ͍ճ͠ɼΑ͋͘Δؒҧ͍
• Α͍ߏʹ͢ΔͨΊͷํ๏ɿoutline-driven writing • ΄͔ɿ༗ӹࢿྉͷڞ༗ɼ૬ޓࠪಡɼͳͲ 20