Upgrade to Pro — share decks privately, control downloads, hide ads and more …

第6章 テキストデータのための素性

 第6章 テキストデータのための素性

本郷言語処理勉強会 #2 で用いた資料。
https://holy.connpass.com/event/122294/
「ゴールドバーグ, 『自然言語処理のための深層学習』, 共立出版, 2019」輪読会。

kmotohas

March 24, 2019
Tweet

More Decks by kmotohas

Other Decks in Technology

Transcript

  1. !LNPUPIBT
    ຊڷݴޠॲཧษڧձ
    ୈ6ষ ςΩετσʔλͷͨΊͷૉੑ

    View Slide

  2. ࣗݾ঺հ
    !3
    ຊڮ࿨و !LNPUPIBT

    ‣ ౦޻େجૅ෺ཧֶઐ߈1I%

    ‣ ιϑτόϯΫגࣜձࣾ

    ‣ %&&1$03&*OD

    ‣ εΧΠϚΠϯυגࣜձࣾ

    w ਂ૚ֶशϞσϧͷ։ൃɾӡ༻ϓϥοτϑΥʔϜ࡞ͬͯΔձࣾ
    +BWB4DBSBͰͰ͖ΔEFFQMFBSOJOHKͷ։ൃ
    ,FSBTOE UI
    DPOUSJCVUPST͕͍ͨΓ͢Δ
    w ๻͸ϓϦηʔϧεΤϯδχΞ

    View Slide

  3. ໨࣍
    !4
    ୈ̎ฤɹࣗવݴޠσʔλͷѻ͍
    ‣ ୈ̒ষɹςΩετσʔλͷͨΊͷૉੑ
    w ࣗવݴޠॲཧʹ͓͚Δ෼ྨ໰୊ͷλΠϓ෼͚
    w ࣗવݴޠॲཧ໰୊ͷͨΊͷૉੑ
    ௚઀؍࡯Ͱ͖Δૉੑ
    ਪଌ͞ΕΔݴޠతಛ௃
    ֩ͱͳΔૉੑͱ૊Έ߹Θͤૉੑ
    OάϥϜૉੑ
    ෼෍࿦తૉੑ

    View Slide

  4. ୈ̒ষͰѻ͏಺༰
    !5
    %//͸ૉੑΤϯδχΞϦϯάͷ͢΂ͯΛ୲ͬͯ͸͘Εͳ͍
    λεΫ͝ͱʹ֩ͱͳΔૉੑͷू߹Λద੾ʹܾఆ͢Δ
    ςΩετσʔλ ࣮਺஋ϕΫτϧ
    ૉੑؔ਺
    ɾૉੑநग़ GFBUVSFFYUSBDUJPO

    ɾૉੑදݱ GFBUVSFSFQSFTFOUBUJPO

    ͜ͷຊͰ͸lGFBUVSFzΛʮಛ௃ྔʯͰ͸ͳ͘ʮૉੑʯͱ༁͢໛༷

    View Slide

  5. ࣗવݴޠॲཧʹ͓͚Δ෼ྨ໰୊ͷλΠϓ෼͚
    !6
    ୯ޠ
    ‣ ੜ͖෺ʁɺͲͷݴޠʁɺҰൠతʁɺࣅͨ୯ޠ͸ʁɺλΠϙʁ
    ςΩετʢจॻ෼ྨEPDVNFOUDMBTTJpDBUJPOʣ
    ‣ εύϜʁɺ੓࣏εϙʔπʁɺൽ೑ʁɺߠఆతʁɺ୭ͷஶ࡞ʁɺ޷·ΕΔʁ
    ςΩετର
    ‣ ಉٛޠʁɺద੾ͳ຋༁ʁɺಉ͡ஶऀʁɺཁ໿ʁ
    จ຺ʹຒΊࠐ·Εͨ୯ޠ
    ‣ CPPL໊ࢺಈࢺܗ༰ࢺʁɺBQQMFاۀՌ෺ʁɺݻ༗දݱʁ
    ೋͭͷ୯ޠͷؔ܎
    ‣ ୯ޠ"͸ಈࢺ#ͷओޠʁɺ໨తޠʁ

    View Slide

  6. ίϥϜɿ୯ޠͱ͸Կ͔
    !7
    ୯ޠۭനͰ۠੾ΒΕͨจࣈͷܥྻʁ
    ‣ lEPH zlEPH zlEPHzlEPHz
    ୯ޠۭന͋Δ͍͸۟ಡ఺Ͱ۠੾ΒΕͨจࣈͷܥྻ UPLFO
    ʁ
    ‣ l*#.zl.SzlEPO`UzϔϒϥΠޠɺΞϥϏΞޠɺதࠃޠɺ೔ຊޠɺʜ
    ‣ lDBUzl$BUzl/FX:PSLzlJDFDSFBNzlJDFDSFBNzlLJDLUIFCVDLFUz
    ୯ޠ XPSE
    ͱτʔΫϯ UPLFO
    ͸۠ผ͞ΕΔ
    ‣ τʔΫϯτʔΫϯԽॲཧͷग़ྗ
    ‣ ୯ޠҙຯΛӡͿ୯Ґ

    View Slide

  7. ࣗવݴޠॲཧ໰୊ͷͨΊͷૉੑ
    !8
    ୯ޠ΍จࣈ͸཭ࢄత‎ࢦඪܭ਺ΛΑ͘༻͍Δ
    w ࢦඪૉੑ JOEJDBUPSGFBUVSF

    w ͋Δ৚݅Λຬͨͨ͠Β̍ɺ͞΋ͳ͚Ε͹̌
    ྫʣ୯ޠEPH͕จॻதʹҰ౓Ҏ্ݱΕͨΒ̍ɺͳ͚Ε͹̌
    ‣ ܭ਺ DPVOU

    w ͋Δग़དྷࣄ͕ੜͨ͡ճ਺
    ྫʣςΩετதʹ୯ޠEPH͕ݱΕͨճ਺
    ௚઀؍࡯Ͱ͖Δૉੑ
    ਪଌ͞ΕΔݴޠతಛ௃
    ֩ͱͳΔૉੑͱ૊Έ߹Θͤૉੑ
    OάϥϜૉੑ
    ෼෍࿦తૉੑ

    View Slide

  8. ௚઀؍࡯Ͱ͖Δૉੑ
    !9
    Ұͭͷ୯ޠͷૉੑ
    ςΩετͷૉੑ
    จ຺தͷ୯ޠͷૉੑ
    ୯ޠؔ܎ͷૉੑ

    View Slide

  9. Ұͭͷ୯ޠͷૉੑ
    !10
    ϨϯϚԽ MFNNBUJ[JOH

    ‣ ϨϯϚ MFNNB
    ࣙॻݟग़͠ޠ
    ‣ CPPLJOH CPPLFE CPPLTCPPL
    εςϛϯά TUFNNJOH

    ‣ ϨϯϚԽΑΓ΋ૈ͍ॲཧ
    ‣ QJDUVSF QJDUVSFT QJDUVSFEQJDUVS
    ݪจ "TUFNNFSNBQTTFRVFODFTPGXPSETUPTIPSUFSTFRVFODFT
    8PSE/FU-FNNBUJ[FS "TUFNNFSNBQTFRVFODFPGXPSEUPTIPSUFSTFRVFODF
    1PSUFS "TUFNNFSNBQTFRVFODPGXPSEUPTIPSUFSTFRVFOD
    -BODBTUFS BTUFNNBQTFRVPGXPSEUPTIPSUTFRV
    4UFNNJOHBOE-FNNBUJ[BUJPOXJUI1ZUIPO/-5,

    View Slide

  10. Ұͭͷ୯ޠͷૉੑ
    !11
    ޠኮࢿݯ MFYJDBMSFTPVSDFT

    ‣ 8PSE/FU>
    w ୯ޠʹ͍ͭͯͷ֓೦తҙຯత஌ࣝΛ֫ಘ͢Δ͜ͱΛ໨త
    w ୯ޠΛෳ਺ͷTZOTFU ಉٛޠू߹
    ʹଐ͢Δ
    w ਓखͰاըߏங
    IUUQXPSEOFUXFCQSJODFUPOFEVQFSMXFCXO
    *#.8BUTPO͸8PSE/FU %#1FEJB FUD
    ͳͲ͔Β஌ࣝମܥΛ֫ಘ
    ΫΠζ൪૊+FPQBSEZͰਓؒͷνϟϯϐϣϯΛഁͬͨ

    IUUQXXXNBUISZVLPLVBDKQdXXXRNBMBC*14+.(/QEG

    View Slide

  11. Ұͭͷ୯ޠͷૉੑ
    !12
    ޠኮࢿݯ MFYJDBMSFTPVSDFT

    ‣ 'SBNF/FUFUBM >
    ‣ 7FSC/FUFUBM >
    w ಈࢺͷपΓʹয఺Λ౰ͯͯؔ࿈͢Δ߲Λྻڍ
    ֩ͱͳΔ߲
    w ఏڙऀ %0/&3

    w डྖऀ 3&$*1*&/5

    w ର৅ 5)&.&

    ֩Ͱͳ͍߲
    w ࣌ 5*.&

    w ໨త 1&3104&

    w ৔ॴ 1-"$&

    w ༷ଶ ."//&3

    w ਓखͰاըߏங
    IUUQTGSBNFOFUJDTJCFSLFMFZFEVGO3FQPSUTEBUBGSBNF*OEFYYNM

    View Slide

  12. Ұͭͷ୯ޠͷૉੑ
    !13
    ޠኮࢿݯ MFYJDBMSFTPVSDFT

    ‣ 5IF1BSBQISBTF%BUBCBTF 11%#

    FUBM 1BWMJDLFUBM >
    w ࣗಈߏங͞Εͨݴ͍׵͑ͷσʔληοτ
    ෼෍࿦తૉੑ EJTUSJCVUJPOBMGFBUVSF

    ‣ ޙड़

    IUUQQBSBQISBTFPSH
    ޠኮࢿݯΛޮ཰తʹར༻͢Δํ๏͸λεΫʹґଘ
    ΤϯδχΞϦϯάతਚྗɾ૑ҙ޻෉͕ඞཁ
    //ϞσϧͰ͸සൟʹ׆༻͞Ε͍ͯͳ͍
    ʢঢ়گ͸มԽ͢Δ͔΋͠Εͳ͍ʣ

    View Slide

  13. ࢀߟʣ6OJ%JDCZࠃཱࠃޠݚڀॴ
    !14
    [email protected]
    ‏ܗଶૉղੳث.F$BC༻ͷղੳ༻ࣙॻ

    View Slide

  14. ௚઀؍࡯Ͱ͖Δૉੑ
    !15
    Ұͭͷ୯ޠͷૉੑ
    ςΩετͷૉੑ
    จ຺தͷ୯ޠͷૉੑ
    ୯ޠؔ܎ͷૉੑ

    View Slide

  15. ςΩετͷૉੑ
    !16
    ୯ޠόοά #BHPG8PSET#08

    ‣ ςΩετதͷ୯ޠͷग़ݱ਺ͷώετάϥϜ
    5'*%'ͷॏΈ෇͚

    View Slide

  16. ௚઀؍࡯Ͱ͖Δૉੑ
    !17
    Ұͭͷ୯ޠͷૉੑ
    ςΩετͷૉੑ
    จ຺தͷ୯ޠͷૉੑ
    ୯ޠؔ܎ͷૉੑ

    View Slide

  17. จ຺தͷ୯ޠͷૉੑ
    !18
    ૭ XJOEPX

    ‣ ͋Δ୯ޠͷ௚ۙͷจ຺ʹয఺
    w lର৅ͷ୯ޠͷपΓʹ୯ޠ9͕ݱΕͨz͜ͱ͕ૉੑ
    ର৅ͷۙ͘ʹ͋Δ୯ޠ͸ԕ͘཭Εͨ୯ޠΑΓ΋ର৅ͷ৘ใΛଟ͍࣋ͬͯ͘Δ͜ͱ͕ଟ͍
    UIFCSPXOGPYKVNQFEPWFSUIFMB[ZCPZ
    XJOEPXTJ[FL యܕతʹ͸

    \XPSECSPXO XPSEGPY XPSEPWFS XPSEUIF^
    \XPSECSPXO XPSEGPY XPSEPWFS XPSEUIF^
    Ґஔ෇͖૭ɿ૬ରతҐஔૉੑ SFMBUJWFQPTJUJPOBMGFBUVSFT^

    View Slide

  18. ௚઀؍࡯Ͱ͖Δૉੑ
    !19
    Ұͭͷ୯ޠͷૉੑ
    ςΩετͷૉੑ
    จ຺தͷ୯ޠͷૉੑ
    ୯ޠؔ܎ͷૉੑ
    ‣ จ຺தͷೋͭͷ୯ޠͷؒͷڑ཭ EJTUBODF

    ‣ ؒʹݱΕΔ୯ޠͦͷ΋ͷ JEFOUJUZ

    View Slide

  19. ਪଌ͞ΕΔݴޠֶతಛ௃
    !20
    ౷ޠ TZOUBY

    ‣ ࣗવݴޠʹ͓͚Δจͷߏ଄Λࢧ഑͢Δنଇ
    UIFCPZXJUIUIFCMBDLTIJSUPQFOFEUIFEPPSXJUIBLFZ
    '&"563&4 '03 /-1 130#-&.4
    JOH EFHSFFT PG BDDVSBDZ 4QFDJBMJ[FE TZTUFNT FYJTU GPS UIF QSFEJDUJPO PG QBSUT PG TQFFDI TZOUBDUJD
    USFFT TFNBOUJD SPMFT EJTDPVSTF SFMBUJPOT BOE PUIFS MJOHVJTUJD QSPQFSUJFT XJUI WBSJPVT EFHSFFT PG
    BDDVSBDZ © BOE UIFTF QSFEJDUJPOT PGUFO TFSWF BT HPPE GFBUVSFT GPS GVSUIFS DMBTTJmDBUJPO QSPCMFNT
    -JOHVJTUJD "OOPUBUJPO -FUT FYQMPSF TPNF GPSNT PG MJOHVJTUJD BOOPUBUJPOT $POTJEFS UIF
    TFOUFODF UIF CPZ XJUI UIF CMBDL TIJSU PQFOFE UIF EPPS XJUI B LFZ 0OF MFWFM PG BOOPUBUJPO BTTJHOT
    UP FBDI XPSE JUT QBSU PG TQFFDI
    UIF CPZ XJUI UIF CMBDL TIJSU PQFOFE UIF EPPS XJUI B LFZ
    %Ėĥ /ĠĦğ 1ģĖġ %Ėĥ "ĕě /ĠĦğ 7Ėģē %Ėĥ /ĠĦğ 1ģĖġ %Ėĥ /ĠĦğ
    (PJOH GVSUIFS VQ UIF DIBJO XF NBSL TZOUBDUJD DIVOL CPVOEBSJFT JOEJDBUJOH UIF UIF CPZ
    JT B OPVO QISBTF
    <
    NP
    UIF CPZ > <
    PP
    XJUI > <
    NP
    UIF CMBDL TIJSU > <
    VP
    PQFOFE > <
    NP
    UIF EPPS > <
    PP
    XJUI > <
    NP
    B LFZ >
    /PUF UIBU UIF XPSE PQFOFE JT NBSLFE BT B WFSCBMDIVOL 71
    ɩJT NBZ OPU TFFN WFSZ
    VTFGVM CFDBVTF XF BMSFBEZ LOPX JUT B WFSC )PXFWFS 71 DIVOLT NBZ DPOUBJO NPSF FMFNFOUT
    DPWFSJOH BMTP DBTFT TVDI BT XJMM PQFOFE BOE EJE OPU PQFO
    ɩF DIVOLJOH JOGPSNBUJPO JT MPDBM " NPSF HMPCBM TZOUBDUJD TUSVDUVSF JT B DPOTUJUVFODZ
    USFF BMTP DBMMFE B QISBTFTUSVDUVSF USFF
    S
    NP
    DT NN NP
    PP PP
    VBD
    VP
    '&"563&4 '03 /-1 130#-&.4
    JOH EFHSFFT PG BDDVSBDZ 4QFDJBMJ[FE TZTUFNT FYJTU GPS UIF QSFEJDUJPO PG QBSUT PG TQFFDI TZOUBDUJD
    USFFT TFNBOUJD SPMFT EJTDPVSTF SFMBUJPOT BOE PUIFS MJOHVJTUJD QSPQFSUJFT XJUI WBSJPVT EFHSFFT PG
    BDDVSBDZ © BOE UIFTF QSFEJDUJPOT PGUFO TFSWF BT HPPE GFBUVSFT GPS GVSUIFS DMBTTJmDBUJPO QSPCMFNT
    -JOHVJTUJD "OOPUBUJPO -FUT FYQMPSF TPNF GPSNT PG MJOHVJTUJD BOOPUBUJPOT $POTJEFS UIF
    TFOUFODF UIF CPZ XJUI UIF CMBDL TIJSU PQFOFE UIF EPPS XJUI B LFZ 0OF MFWFM PG BOOPUBUJPO BTTJHOT
    UP FBDI XPSE JUT QBSU PG TQFFDI
    UIF CPZ XJUI UIF CMBDL TIJSU PQFOFE UIF EPPS XJUI B LFZ
    %Ėĥ /ĠĦğ 1ģĖġ %Ėĥ "ĕě /ĠĦğ 7Ėģē %Ėĥ /ĠĦğ 1ģĖġ %Ėĥ /ĠĦğ
    (PJOH GVSUIFS VQ UIF DIBJO XF NBSL TZOUBDUJD DIVOL CPVOEBSJFT JOEJDBUJOH UIF UIF CPZ
    JT B OPVO QISBTF
    <
    NP
    UIF CPZ > <
    PP
    XJUI > <
    NP
    UIF CMBDL TIJSU > <
    VP
    PQFOFE > <
    NP
    UIF EPPS > <
    PP
    XJUI > <
    NP
    B LFZ >
    /PUF UIBU UIF XPSE PQFOFE JT NBSLFE BT B WFSCBMDIVOL 71
    ɩJT NBZ OPU TFFN WFSZ
    VTFGVM CFDBVTF XF BMSFBEZ LOPX JUT B WFSC )PXFWFS 71 DIVOLT NBZ DPOUBJO NPSF FMFNFOUT
    DPWFSJOH BMTP DBTFT TVDI BT XJMM PQFOFE BOE EJE OPU PQFO
    ɩF DIVOLJOH JOGPSNBUJPO JT MPDBM " NPSF HMPCBM TZOUBDUJD TUSVDUVSF JT B DPOTUJUVFODZ
    USFF BMTP DBMMFE B QISBTFTUSVDUVSF USFF
    S
    NP VP
    '&"563&4 '03 /-1 130#-&.4
    JOH EFHSFFT PG BDDVSBDZ 4QFDJBMJ[FE TZTUFNT FYJTU GPS UIF QSFEJDUJPO PG QBSUT PG TQFFDI TZOUBDUJD
    USFFT TFNBOUJD SPMFT EJTDPVSTF SFMBUJPOT BOE PUIFS MJOHVJTUJD QSPQFSUJFT XJUI WBSJPVT EFHSFFT PG
    BDDVSBDZ © BOE UIFTF QSFEJDUJPOT PGUFO TFSWF BT HPPE GFBUVSFT GPS GVSUIFS DMBTTJmDBUJPO QSPCMFNT


    -JOHVJTUJD "OOPUBUJPO -FUT FYQMPSF TPNF GPSNT PG MJOHVJTUJD BOOPUBUJPOT $POTJEFS UIF
    TFOUFODF UIF CPZ XJUI UIF CMBDL TIJSU PQFOFE UIF EPPS XJUI B LFZ 0OF MFWFM PG BOOPUBUJPO BTTJHOT
    UP FBDI XPSE JUT QBSU PG TQFFDI
    UIF CPZ XJUI UIF CMBDL TIJSU PQFOFE UIF EPPS XJUI B LFZ
    %Ėĥ /ĠĦğ 1ģĖġ %Ėĥ "ĕě /ĠĦğ 7Ėģē %Ėĥ /ĠĦğ 1ģĖġ %Ėĥ /ĠĦğ
    (PJOH GVSUIFS VQ UIF DIBJO XF NBSL TZOUBDUJD DIVOL CPVOEBSJFT JOEJDBUJOH UIF UIF CPZ
    JT B OPVO QISBTF
    <
    NP
    UIF CPZ > <
    PP
    XJUI > <
    NP
    UIF CMBDL TIJSU > <
    VP
    PQFOFE > <
    NP
    UIF EPPS > <
    PP
    XJUI > <
    NP
    B LFZ >
    /PUF UIBU UIF XPSE PQFOFE JT NBSLFE BT B WFSCBMDIVOL 71
    ɩJT NBZ OPU TFFN WFSZ
    VTFGVM CFDBVTF XF BMSFBEZ LOPX JUT B WFSC )PXFWFS 71 DIVOLT NBZ DPOUBJO NPSF FMFNFOUT
    DPWFSJOH BMTP DBTFT TVDI BT XJMM PQFOFE BOE EJE OPU PQFO
    ɩF DIVOLJOH JOGPSNBUJPO JT MPDBM " NPSF HMPCBM TZOUBDUJD TUSVDUVSF JT B DPOTUJUVFODZ
    USFF BMTP DBMMFE B QISBTFTUSVDUVSF USFF
    S
    NP
    DT
    the boy opened
    IN IN
    NN
    NP NP
    DT
    DT
    with with
    the
    the black shirt a key
    door
    JJ NN NN
    DT
    NN NP
    PP PP
    VBD
    VP
    $POTUJUVFODZ USFFT BSF OFTUFE MBCFMFE CSBDLFUJOH PWFS UIF TFOUFODF JOEJDBUJOH UIF IJ
    FSBSDIZ PG TZOUBDUJD VOJUT UIF OPVO QISBTF UIF CPZ XJUI UIF CMBDL TIJSU JT NBEF PG UIF OPVO
    '&"563&4 '03 5&956"- %"5"
    QISBTF UIF CPZ BOE UIF QSFQPTJUJPO QISBTF 11
    XJUI UIF CMBDL TIJSU ɩF MBUUFS JUTFMG DPOUB
    UIF OPVO QISBTF UIF CMBDL TIJSU )BWJOH XJUI B LFZ OFTUFE VOEFS UIF 71 BOE OPU VOEFS
    /1 UIF EPPS TJHOBMT UIBU XJUI B LFZ NPEJmFT UIF WFSC PQFOFE PQFOFE XJUI B LFZ
    SBUIFS U
    UIF /1 B EPPS XJUI B LFZ

    " EJĊFSFOU LJOE PG TZOUBDUJD BOOPUBUJPO JT B EFQFOEFODZ USFF 6OEFS EFQFOEF
    TZOUBY FBDI XPSE JO UIF TFOUFODF JT B NPEJmFS PG BOPUIFS XPSE XIJDI JT DBMMFE JUT IF
    &BDI XPSE JO UIF TFOUFODF JT IFBEFE CZ BOPUIFS TFOUFODF XPSE FYDFQU GPS UIF NBJO XP
    VTVBMMZ B WFSC XIJDI JT UIF SPPU PG UIF TFOUFODF BOE JT IFBEFE CZ B TQFDJBM iSPPUw OPEF
    the boy with the black shirt opened the door with a key
    det prep
    prep
    amod
    det
    pobj
    nsubj
    root
    dobj pobj
    det det
    8IJMF DPOTUJUVFODZ USFFT NBLF FYQMJDJU UIF HSPVQJOH PG XPSET JOUP QISBTFT EFQFOEF
    USFFT NBLF FYQMJDJU UIF NPEJmDBUJPO SFMBUJPOT BOE DPOOFDUJPOT CFUXFFO XPSET 8PSET UIBU
    ඼ࢺ 1BSUPG4QFFDI104

    ౷ޠతνϟϯΫ
    TZOUBDUJDDIVOL

    ߏ੒ཁૉ໦
    DPOTUJUVFODZUSFF

    ۟ߏ଄໦
    QISBTFTUSVDUVSFUSFF

    ґଘߏ଄໦
    EFQFOEFODZUSFF

    View Slide

  20. ਪଌ͞ΕΔݴޠֶతಛ௃
    !21
    '&"563&4 '03 5&956"- %"5"


    QISBTF UIF CPZ BOE UIF QSFQPTJUJPO QISBTF 11
    XJUI UIF CMBDL TIJSU ɩF MBUUFS JUTFMG DPOUBJOT
    UIF OPVO QISBTF UIF CMBDL TIJSU )BWJOH XJUI B LFZ OFTUFE VOEFS UIF 71 BOE OPU VOEFS UIF
    /1 UIF EPPS TJHOBMT UIBU XJUI B LFZ NPEJmFT UIF WFSC PQFOFE PQFOFE XJUI B LFZ
    SBUIFS UIBO
    UIF /1 B EPPS XJUI B LFZ

    " EJĊFSFOU LJOE PG TZOUBDUJD BOOPUBUJPO JT B EFQFOEFODZ USFF 6OEFS EFQFOEFODZ
    TZOUBY FBDI XPSE JO UIF TFOUFODF JT B NPEJmFS PG BOPUIFS XPSE XIJDI JT DBMMFE JUT IFBE
    &BDI XPSE JO UIF TFOUFODF JT IFBEFE CZ BOPUIFS TFOUFODF XPSE FYDFQU GPS UIF NBJO XPSE
    VTVBMMZ B WFSC XIJDI JT UIF SPPU PG UIF TFOUFODF BOE JT IFBEFE CZ B TQFDJBM iSPPUw OPEF
    the boy with the black shirt opened the door with a key
    det prep
    prep
    amod
    det
    pobj
    nsubj
    root
    dobj pobj
    det det
    8IJMF DPOTUJUVFODZ USFFT NBLF FYQMJDJU UIF HSPVQJOH PG XPSET JOUP QISBTFT EFQFOEFODZ
    USFFT NBLF FYQMJDJU UIF NPEJmDBUJPO SFMBUJPOT BOE DPOOFDUJPOT CFUXFFO XPSET 8PSET UIBU BSF
    GBS BQBSU JO UIF TVSGBDF GPSN PG UIF TFOUFODF NBZ CF DMPTF JO JUT EFQFOEFODZ USFF 'PS FYBNQMF
    CPZ BOE PQFOFE IBWF GPVS XPSET CFUXFFO UIFN JO UIF TVSGBDF GPSN CVU IBWF B EJSFDU OTVCK
    FEHF DPOOFDUJOH UIFN JO UIF EFQFOEFODZ USFF
    ɩF EFQFOEFODZ SFMBUJPOT BSF TZOUBDUJD UIFZ BSF DPODFSOFE XJUI UIF TUSVDUVSF PG UIF
    TFOUFODF 0UIFS LJOET PG SFMBUJPOT BSF NPSF TFNBOUJD 'PS FYBNQMF DPOTJEFS UIF NPEJmFST PG
    UIF WFSC PQFO BMTP DBMMFE UIF BSHVNFOUT PG UIF WFSC ɩF TZOUBDUJD USFF DMFBSMZ NBSLT UIF CPZ
    XJUI UIF CMBDL TIJSU
    UIF EPPS BOE XJUI B LFZ BT BSHVNFOUT BOE BMTP UFMMT VT UIBU XJUI B LFZ
    JT BO BSHVNFOU PG PQFO SBUIFS UIBO B NPEJmFS PG EPPS *U EPFT OPU UFMM VT IPXFWFS XIBU BSF
    UIF TFNBOUJDSPMFT PG UIF BSHVNFOUT XJUI SFTQFDU UP UIF WFSC JF UIBU UIF CPZ JT UIF "ĘĖğĥ
    QFSGPSNJOH UIF BDUJPO BOE UIBU B LFZ JT BO *ğĤĥģĦĞĖğĥ DPNQBSF UIBU UP UIF CPZ PQFOFE UIF
    EPPS XJUI B TNJMF )FSF UIF TFOUFODF XJMM IBWF UIF TBNF TZOUBDUJD TUSVDUVSF CVU VOMFTT XF
    BSF JO B NBHJDBMXPSME B TNJMF JT B .ĒğğĖģ SBUIFS UIBO BO *ğĤĥģĦĞĖğĥ ɩF TFNBOUJD SPMF
    MBCFMJOH BOOPUBUJPOT SFWFBM UIFTF TUSVDUVSFT
    ґଘߏ଄໦
    EFQFOEFODZUSFF

    '&"563&4 '03 /-1 130#-&.4


    the boy with the black shirt opened the door with a smile
    the boy with the black shirt opened the door with a key
    Agent
    Instrument
    Patient
    Manner
    Patient
    Agent
    #FTJEFT UIF PCTFSWBCMF QSPQFSUJFT MFUUFST XPSET DPVOUT MFOHUIT MJOFBS EJTUBODFT GSFRVFO
    DJFT FUD
    XF DBO BMTP MPPL TVDI JOGFSSFE MJOHVJTUJD QSPQFSUJFT PG XPSET TFOUFODFT BOE EPDVNFOUT
    'PS FYBNQMF XF DPVME MPPL BU UIF QBSUPGTQFFDI UBH 104
    PG B XPSE XJUIJO B EPDVNFOU *T JU B
    OPVO B WFSC BEKFDUJWF PS B EFUFSNJOFS
    UIF TZOUBDUJD SPMF PG B XPSE %PFT JU TFSWF BT B TVCKFDU PS
    ҙຯ໾ׂϥϕϧ෇͚
    TFNBOUJDSPMFMBCFMJOH

    View Slide

  21. ਪଌ͞ΕΔݴޠֶతಛ௃
    !22
    ඼ࢺ 1BSUPG4QFFDIUBH104

    ‣ ໊ࢺɺಈࢺɺܗ༰ࢺɺݶఆࢺ
    ౷ޠత໾ׂ TZOUBDUJDSPMF

    ‣ ಈࢺͷओޠɺ໨తޠɺओಈࢺɺ෭ࢺతम০ޠ
    ҙຯ໾ׂ TFNBOUJDSPMF

    ‣ UIFLFZ ಓ۩
    PQFOUIFEPPSUIFCPZ ಈ࡞ओ
    PQFOTUIFEPPS
    ౷ޠతґଘߏ଄໦ TZOUBDUJDEFQFOEFODZUSFF

    ‣ ଟ͘ͷ୯ޠͰִͯΒΕΔೋͭͷ୯ޠ͕౷ޠߏ଄ʹ͓͍ͯ͸ྡ઀ͯ͠Δ͜ͱ΋
    ஊ࿩ؔ܎ EJTDPVSTFSFMBUJPOT

    ‣ จͱจͱͷؔ܎ɻৄࡉԽ NPSFPWFS
    ɺ൓࿦ IPXFWFS
    ɺҼՌ CFDBVTF
    ͳͲʢ઀ଓࢺͳ͍৔߹΋ʣ
    রԠ BOBQIPSB

    ‣ UIFCPZPQFOFEUIFEPPSXJUIBLFZ*UXBTO`UMPDLFEBOEIFFOUFSFEUIFSPPN)FTBXBNBO)FXBTTNJMJOH

    View Slide

  22. ݴޠֶత֓೦͸ඞཁ͔
    !23
    ݴޠֶతಛ௃͸ඞཁͳ͘ɺ//͸தؒදݱΛࣗ෼Ͱֶश͢Δ
    ʢਂ૚ֶऀࢧ࣋ऀͷҰ෦ʣ
    े෼ͳྔͷσʔλͱਖ਼͍͠ํ޲෇͚͕͋Ε͹Մೳʢචऀʣ
    ‣ े෼ͳσʔλ͕ͳ͍৔߹͸໌ࣔతʹҰൠతͳ֓೦Λ//ʹఏڙ͢Δͷ͸
    ༗ӹ
    ‣ ݴޠֶతಛ௃Λೖྗૉੑʹར༻͠ͳ͍৔߹Ͱ΋ɺϚϧνλεΫֶशʹ
    ͓͚Δ௥Ճతͳڭࢣͱͯ͠༻͍Δ͜ͱ͕Ͱ͖Δ
    %FFQ-FBSOJOH"$SJUJDBM"QQSBJTBM.BSDVT>

    View Slide

  23. ֩ͱͳΔૉੑͱ૊Έ߹Θͤૉੑ
    !24
    ֩ͱͳΔૉੑ DPSFGFBUVSFT

    ૊Έ߹Θͤૉੑ DPNCJOBUJPOGFBUVSFT

    ‣ ৘ใྔ͕͋Γɺ͔ͭແବ͕ൺֱతগͳ͍Α͏ͳૉੑΛ࡞Δͷ͸೉͍͠
    ɾ ઢܗϞσϧͰ͸ɺ૊Έ߹Θͤʢ9㱸:ʣ΋9 :ͦΕͧΕͷ߹ܭҎ্ͷείΞʹͳΒͳ͍
    ‣ χϡʔϥϧωοτϫʔΫͷΑ͏ͳඇઢܗϞσϧͰ͜ͷ໰୊Λճආ
    ɾ ֩ૉੑͷू߹͚ͩఆٛ͠ɺॏཁͳ૊Έ߹Θͤૉੑ͸ωοτϫʔΫͷ܇࿅Ͱ֫ಘ
    \XJOEPXOPU XJOEPX9 XJOEPX: XJOEPX;^
    \XJOEPX9` XJOEPXMJLF XJOEPX:` XJOEPX;`^
    \XJOEPXOPU XJOEPXMJLF XJOEPX9`` XJOEPX:``^

    1BSJT‎ཱྀߦΧςΰϦ
    )JMUPO‎ཱྀߦΧςΰϦ
    1BSJT)JMUPO‎༗໊ਓPSΰγοϓΧςΰϦ

    View Slide

  24. OάϥϜૉੑ
    !25
    OάϥϜ OHSBN

    ‣ /FX:PSLOPUHPPE1BSJT)JMUPOͷΑ͏ͳߏ଄Λ֫ಘͰ͖Δ
    w PGUIFPOBUIFCPZͳͲͷόΠάϥϜ͸Ұൠత
    Ϟσϧͷௐ੔Ͱখ͍͞ॏΈΛׂΓ౰ͯΔ
    ‣ .-1͸9㱸:ͱ͍͏૊Έ߹Θͤ͸ֶशͰ͖Δ͕9: CJHSBN
    ͸ೝࣝͰ͖ͳ͍
    w ৞ΈࠐΈωοτϫʔΫ͸OHSBNGFBUVSFΛݟ͚ͭग़͢Α͏ʹઃܭ͞Ε͍ͯΔ
    w CJEJSFDUJPOBM3//͸Մม௕OHSBN΍ɺؒʹΪϟοϓΛؚΉOHSBNΛײ஌

    View Slide

  25. ෼෍࿦తૉੑ
    !26
    ݴޠͷ෼෍Ծઆ EJTUSJCVUJPOBMIZQPUIFTJT
    )BSSJT>
    ‣ 'JSUIl:PVTIBMMLOPXBXPSECZUIFDPNQBOZJULFFQTz
    w ୯ޠͷҙຯ͸ͦΕ͕༻͍ΒΕΔจ຺͔Βਪ࿦Ͱ͖Δ CZ8JUUHFOTUFJO

    Firth, J.R. (1957). A synopsis of linguistic theory 1930-1955
    ɾQJ[[B
    DIBJSɾ ɾCVSHFS
    ɾQJ[[B
    DIBJSɾ
    ɾCVSHFS
    ݪจΊͪΌͪ͘ΌԼ඼ͳྫͰস͏

    View Slide

  26. จ຺͔Β୯ޠͷҰൠԽΛֶश
    !27
    ΫϥελϦϯάʹجͮ͘ख๏ DMVTUFSJOHCBTFENFUIPET

    ‣ ྨࣅͨ͠୯ޠΛಉ͡୯ޠΛಉ͡ΫϥελʹׂΓ౰ͯΔ
    ‣ Ϋϥελʹॴଐ͢Δ౓߹͍ʹΑͬͯͦΕͧΕͷ୯ޠΛදݱ
    ຒΊࠐΈʹجͮ͘ख๏ FNCFEEJOHCBTFENFUIPET

    ‣ ͦΕͧΕͷ୯ޠΛϕΫτϧͰදݱ
    ‣ ྨࣅͨ͠୯ޠ͸ྨࣅͨ͠ϕΫτϧͰදݱ͞ΕΔ
    ɾQJ[[B
    DIBJSɾ ɾCVSHFS
    ɾQJ[[B
    DIBJSɾ
    ɾCVSHFS
    ˎ-POEPO#FSMJO͸ৗʹྨࣅͨ͠୯ޠͱͯ͠ѻ͏΂͖ʁʢߤۭνέοτ༧໿ɺจॻ຋༁ͳͲͰ͸۠ผ͍ͨ͠ʣ

    View Slide

  27. ·ͱΊ
    !28
    ςΩετσʔλͷૉੑͱͯ͠࢖ΘΕ͍ͯΔ΋ͷΛྻڍ
    ‣ ෼ྨର৅͸୯ޠɺςΩετɺςΩετରɺจষதͷ୯ޠɺ୯ޠͷؔ܎
    ‣ ௚઀؍࡯Ͱ͖Δૉੑ
    w ϨϯϚɺεςϛϯάɺޠኮࢿݯɺ5'*%'ɺ8JOEPX
    ‣ ਪଌ͞ΕΔݴޠతಛ௃
    w จ๏
    ‣ ֩ͱͳΔૉੑͱ૊Έ߹Θͤૉੑ
    w //Λ༻͍ͯ૊Έ߹ΘͤૉੑΛ֫ಘ
    ‣ OάϥϜૉੑ
    w l/FX:PSLzͳͲෳ਺ͷ୯ޠ͔Βߏ੒͞ΕΔݴ༿Λଊ͑Δ
    ‣ ෼෍࿦తૉੑ
    w l:PVTIBMMLOPXBXPSECZUIFDPNQBOZJULFFQTz

    View Slide