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
Neural Network Models for Paraphrase Identifica...
Search
onizuka laboratory
October 23, 2018
Research
79
1
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Neural Network Models for Paraphrase Identification, Semantic Textual Similarity, Natural Language Inference, and Question Answering
弊研究室で行なったCOLING2018読み会の発表資料です。
onizuka laboratory
October 23, 2018
More Decks by onizuka laboratory
See All by onizuka laboratory
Phrase-Based & Neural Unsupervised Machine Translation
onilab
0
120
Tell-and-Answer: Towards Explainable Visual Question Answering using Attributes and Captions
onilab
0
82
Card-660: A Reliable Evaluation Framework for Rare Word Representation Models
onilab
0
43
A Word-Complexity Lexicon and A Neural Readability Ranking Model for Lexical Simplification
onilab
0
140
Integrating Transformer and Paraphrase Rules for Sentence Simplification
onilab
0
66
An Auto-Encoder Matching Model for Learning Utterance-Level Semantic Dependency in Dialogue Generation
onilab
0
62
Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints
onilab
0
110
Modeling Multi-turn Conversation with Deep Utterance Aggregation
onilab
0
100
Learning Semantic Sentence Embeddings using Pair-wise Discriminator
onilab
0
130
Other Decks in Research
See All in Research
LiDAR点群の地表面分類手法の比較・検証
vegapunkhiroshi79
0
110
2026年度 生成AI を活用した論文執筆ガイド/ワークショップ / 2026 Academic Year Guide to Writing Papers Using Generative AI - Workshop
ks91
PRO
0
170
LLMアプリケーションの透明性について
fufufukakaka
0
230
はじまりの クエスチョンブック —余暇と豊かさにあふれた社会とは?
culturaltransition
PRO
0
510
進学校の生徒にはア行の苗字が多いのか
ozekinote
0
430
第66回コンピュータビジョン勉強会@関東 Epona: Autoregressive Diffusion World Model for Autonomous Driving
kentosasaki
0
620
英語教育 “研究” のあり方:学術知とアウトリーチの緊張関係
terasawat
1
990
R&Dチームを起ち上げる
shibuiwilliam
1
260
SAKURAONE:An Open Ethernet-based AI HPC System And Its Observed Workload Dynamicsin a Single-Tenant LLM Development Environment
yuukit
1
300
Ankylosing Spondylitis
ankh2054
0
170
SoftMatcha 2: 1兆語規模コーパスの超高速かつ柔らかい検索
e869120_sub
6
3.4k
Anthropic が提案する LLM の内部状態を自然言語で説明可能にした Natural Language Autoencoders / Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations
shunk031
0
120
Featured
See All Featured
Raft: Consensus for Rubyists
vanstee
141
7.5k
Sam Torres - BigQuery for SEOs
techseoconnect
PRO
0
280
We Analyzed 250 Million AI Search Results: Here's What I Found
joshbly
1
1.3k
New Earth Scene 8
popppiees
3
2.3k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
31
3.2k
A Modern Web Designer's Workflow
chriscoyier
698
190k
The SEO identity crisis: Don't let AI make you average
varn
0
480
AI Search: Implications for SEO and How to Move Forward - #ShenzhenSEOConference
aleyda
1
1.3k
Six Lessons from altMBA
skipperchong
29
4.3k
[SF Ruby Conf 2025] Rails X
palkan
2
1.1k
Digital Projects Gone Horribly Wrong (And the UX Pros Who Still Save the Day) - Dean Schuster
uxyall
0
1.6k
Chrome DevTools: State of the Union 2024 - Debugging React & Beyond
addyosmani
10
1.2k
Transcript
/FVSBM/FUXPSL.PEFMTGPS 1* 454 /-* BOE2" 8VXFJ -BOBOE8FJ9V *O1SPDPG$0-*/( QQ #FTU1BQFS"XBSE
1SFTFOUFECZ5PNPZVLJ,BKJXBSB
4FOUFODF1BJS.PEFMJOH จͱจͷؒͷҙຯతͳؔΛٻΊΔͭͷλεΫ • 1* 1BSBQISBTF*EFOUJGJDBUJPO ݴ͍͑ೝࣝ • 454 4FNBOUJD5FYUVBM4JNJMBSJUZ จͷྨࣅਪఆ
• /-* /BUVSBM-BOHVBHF*OGFSFODF ؚҙؔೝࣝ • 2" 2VFTUJPO"OTXFSJOH ࣭Ԡ 4FOUFODF 4FOUFODF 4FOUFODF1BJS.PEFM -BCFM
.PUJWBUJPOBOE3FTFBSDI2VFTUJPOT • ੲλεΫݻ༗ͷख๏Λ։ൃ͢Δ͔͠ͳ͔ͬͨ • ࠓڞ௨ͷϞσϧͰ1*454/-*2"શ෦ಉ͡Α͏ʹղ͚Δ • ͔͠͠ɺ֤छ405"ख๏dݸͷσʔλͰ͔͠ධՁ͍ͯ͠ͳ͍ Ø ͦͷϞσϧଞͷλεΫσʔληοτͰ༗༻Ͱ͔͢ʁ Ø
γεςϜઃܭύϥϝλ࠷దԽ͕ͲΕ͚ͩੑೳΛվળ͠·͔͢ʁ 4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" *OGFS4FOU 44& %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF 18*.
(FOFSBM'SBNFXPSLGPS4FOUFODF1BJS.PEFMJOH 4FOUFODF&ODPEJOHCBTFE 8PSE*OUFSBDUJPOCBTFE
(FOFSBM'SBNFXPSLGPS4FOUFODF1BJS.PEFMJOH *OQVU&NCFEEJOH-BZFS • ֶशࡁΈͷ୯ޠࢄදݱʢXPSEWFD(MP7FʣΛ͏ͷ͕Ұൠత • 4/4σʔλʹରͯ͠αϒϫʔυ୯Ґͷࢄදݱ͕༗ޮ 4FOUFODF&ODPEJOHCBTFE 8PSE*OUFSBDUJPOCBTFE
(FOFSBM'SBNFXPSLGPS4FOUFODF1BJS.PEFMJOH $POUFYU&ODPEJOH-BZFS • 3W//ͱ͔$//ͱ͔-45.ͱ͔Ͱ୯ޠ͔ΒจͷࢄදݱΛ࡞Δ • 4FOUFODF&ODPEJOHCBTFEϞσϧͷ߹ɺ͜ΕͰऴΘΓ 4FOUFODF&ODPEJOHCBTFE 8PSE*OUFSBDUJPOCBTFE
(FOFSBM'SBNFXPSLGPS4FOUFODF1BJS.PEFMJOH *OUFSBDUJPOBOE"UUFOUJPO-BZFS • ֤࣌ࠁͷӅΕΛநग़ͯ͠ɺ୯ޠΞϥΠϝϯτ෩ͷใΛಘΔ 4FOUFODF&ODPEJOHCBTFE 8PSE*OUFSBDUJPOCBTFE
4FOUFODF&ODPEJOHCBTFE.PEFMT *OGFS4FOU #J-45..BYQPPMJOH/FUXPSL 44& 4IPSUDVU4UBDLFE 4FOUFODF&ODPEFS.PEFM
8PSE*OUFSBDUJPOCBTFE.PEFMT • ͜͜ʹςΩετΛೖྗ 18*. 1BJSXJTF8PSE *OUFSBDUJPO.PEFM %FD"UU %FDPNQPTBCMF"UUFOUJPO.PEFM &4*.
&OIBODFE4FRVFOUJBM*OGFSFODF.PEFM
%BUBTFU 4J[F &YBNQMF -BCFM 4/-* 5SBJO %FW 5FTU
5XPNFOPOCJDZDMFTDPNQFUJOHJOBSBDF .FO BSFSJEJOHCJDZDMFTPOUIFTUSFFU FOUBJMNFOU OFVUSBM DPOUSBEJDU .VMUJ/-* 5SBJO %FW 5FTU 5IF0ME0OF BMXBZTDPNGPSUFE$B`EBBO FYDFQUUPEBZ $B`EBBO LOFXUIF0ME0OFWFSZXFMM FOUBJMNFOU OFVUSBM DPOUSBEJDU 2VPSB 5SBJO %FW 5FTU 8IBU TIPVME*EPUPBWPJETMFFQJOHJODMBTT )PXEP*OPUTMFFQJOBCPSJOHDMBTT QBSBQISBTF OPOQBSBQISBTF 5XJUUFS63- 5SBJO %FW 5FTU -FUUFSXBSOFE8FMMT'BSHPPGlXJEFTQSFBEzGSBVE JO -FUUFST TVHHFTU8FMMT'BSHPTDBOEBMTUBSUFEFBSMJFS QBSBQISBTF OPOQBSBQISBTF 1*5 5SBJO %FW 5FTU &[FLJFM "OTBI XUIF%TIBEFT1PQQFEPVUMFOT &[FLJFM"OTBI XBTXFBSJOHMFOTMFTT%HMBTTFT QBSBQISBTF OPOQBSBQISBTF 454 5SBJO %FW 5FTU 5IFOQFSIBQTXFDPVMEIBWFBWPJEFEBDBUBTUSPQIF 5IFOXFNJHIUIBWFCFFOBCMFUPBWPJEBEJTBTUFS TDPSF < > 8JLJ2" 5SBJO %FW 5FTU )PXNVDIJTUBCMFTQPPO PGXBUFS *O"VTUSBMJBPOFUBCMFTQPPO NFBTVSFNFOUVOJU JT N- USVF GBMTF 5SFD2" 5SBJO %FW 5FTU 8IPXBT-JODPMO`T4FDSFUBSZPG4UBUF 8JMMJBN 4FXBSE USVF GBMTF λεΫछྨͷσʔληοτʢؚҙؔೝࣝɺݴ͍͑ೝࣝɺจͷྨࣅਪఆɺ࣭Ԡʣ
࣮ݧ݁Ռʢ੨ɿ࠷ߴɺɿ࠷ʣ جຊతʹΦϦδφϧͷใࠂΑΓ dϙΠϯτ͍ Ø ݸʑͷσʔληοτ͝ͱʹൣғͳνϡʔχϯά͍ͯ͠ͳ͍ Ø ϥϯμϜγʔυΛዞҙతʹબΜͰ͍ͳ͍ Ø શͯͷϞσϧΛ1Z5PSDIͰ࣮ͨ͠
4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" *OGFS4FOU 44& %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF 18*. ˞͋ͱͷϖʔδͰղઆ͢ΔͷͰɺ΅ʔͬͱݟͯ͘ΕͨΒ͍͍Ͱ͢
ଟ͘ͷλεΫͰ 8PSE*OUFSBDUJPOCBTFE.PEFM͕ڧ͍ ୯ޠؒͷؔΛཅʹߟྀͨ͠ํ͕ྑ͍
4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" 4FOUFODF&ODPEJOHCBTFE.PEFM 8PSE*OUFSBDUJPO.PEFM
-45.ʹؔ͢Δߟ %FD"UU 8PSE*OUFSBDUJPOCBTFE.PEFMXJUIPVU-45. &4*. %FD"UU XJUI #J 5SFF -45.&ODPEFST
4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF
-45.ʹΑΔจ຺ใͷΤϯίʔυ͕ॏཁ %FD"UU 8PSE*OUFSBDUJPOCBTFE.PEFMXJUIPVU-45. &4*. %FD"UU XJUI #J 5SFF -45.&ODPEFST
4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF
5SFF-45.ΑΓ #J-45.ͷํ͕ৗʹڧ͍ %FD"UU 8PSE*OUFSBDUJPOCBTFE.PEFMXJUIPVU-45. &4*. %FD"UU XJUI #J 5SFF -45.&ODPEFST
4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF
5SFF-45. 5XJUUFSͷσʔλͰߩݙ͢Δʢ63-ɺ1*5ʣ ྫ͑ lFWFSXPOEFSFE zͱ͍͏ϑϨʔζΛແࢹͰ͖ΔΑ͏ʹͳΔ FWFSXPOEFSFE XIZZPVSSFDPSEFEWPJDFTPVOETXFJSEUP ZPV BOEXIZEPPVSSFDPSEFEWPJDFTTPVOETPXFJSEUPVT
4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF Ξϊςʔλ͕࡞ͬͨଞͷσʔλͱҧ͍ɺ ͜ΕΒͷ5XJUUFSσʔλݱ࣮ͷςΩετ
܇࿅࣌ؒʢ4/-*ͷສจରʣ -45.ൈ͖ͷ%FD"UU͕࠷ɺ#J-45.ͷ*OGFS4FOU͕࣍ʹ͍
*OGFS4FOU 44& %FD"UU &4*.@TFR &4*.@USFF &4*.@TFR USFF 18*. <࣌ؒ>
୯ޠҰக͕͍΄Ͳղ͘ͷ͕͍͠
*OGFS4FOU 44& %FD"UU &4*.@TFR 18*. จؒͷ୯ޠҰக d
จ͕͍΄Ͳղ͘ͷ͕͍͠
*OGFS4FOU 44& %FD"UU &4*.@TFR 18*. จ UPLFOT dUPLFOT UPLFOT
ݴ͍͑ೝࣝʹ͓͚ΔసҠֶशʢ2VPSBສจରɺ1*5ສจରʣ
*OGFS4FOU 44& %FD"UU &4*.@TFR 18*. 2VPSBͰ܇࿅ˠ1*5ͰධՁ 1*5Ͱ܇࿅ˠ1*5ͰධՁ *OGFS4FOU 44& %FD"UU &4*.@TFR 18*. QBSBNFUFST . . , . . ύϥϝλ͕ଟ͍ͷͰେྔͷσʔλ͕ඞཁ PSυϝΠϯʹґଘ͠ͳ͍ҰൠతͳϞσϧ
IUUQTHJUIVCDPNMBOXVXFJ41.@UPPMLJU 4/-* ./-* 2VPSB 63- 1*5 454 8JLJ2" 5SFD2" *OGFS4FOU
44& %FD"UU &4*. 18*. 4FOUFODF&ODPEJOHCBTFE 8PSE*OUFSBDUJPOCBTFE