%PDF-1.3 1 0 obj << /Kids [ 3 0 R 4 0 R 5 0 R 6 0 R 7 0 R 8 0 R 9 0 R 10 0 R 11 0 R ] /Type /Pages /Count 9 >> endobj 2 0 obj << /Title (Aligning Books and Movies\072 Towards Story\055Like Visual Explanations by Watching Movies and Reading Books) /Producer (PyPDF2) /Author (Yukun Zhu\054 Ryan Kiros\054 Rich Zemel\054 Ruslan Salakhutdinov\054 Raquel Urtasun\054 Antonio Torralba\054 Sanja Fidler) /Subject (2015 IEEE International Conference on Computer Vision) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 13 0 R /Resources << /XObject << /x8 14 0 R /x6 17 0 R /x12 20 0 R /R42 23 0 R /x10 24 0 R >> /ExtGState << /s9 27 0 R /s11 30 0 R /a0 << /CA 1 /ca 1 >> /R23 33 0 R /s5 34 0 R /s7 37 0 R >> /Font << /R40 40 0 R /F2 44 0 R /R28 45 0 R /F1 49 0 R /R38 50 0 R /R36 54 0 R /R34 57 0 R /R32 62 0 R /R24 65 0 R /R30 68 0 R /R26 72 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 76 0 R 77 0 R 78 0 R 79 0 R 80 0 R 81 0 R 82 0 R 83 0 R 84 0 R 85 0 R 86 0 R 87 0 R 88 0 R 89 0 R 90 0 R ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 91 0 R /Resources << /ExtGState << /R23 33 0 R >> /Font << /R78 92 0 R /F2 96 0 R /F1 97 0 R /R38 50 0 R /R36 54 0 R /R24 65 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 98 0 R 99 0 R 100 0 R 101 0 R 102 0 R 103 0 R 104 0 R 105 0 R 106 0 R 107 0 R 108 0 R 109 0 R 110 0 R 111 0 R 112 0 R 113 0 R 114 0 R 115 0 R 116 0 R 117 0 R 118 0 R 119 0 R 120 0 R 121 0 R 122 0 R 123 0 R 124 0 R 125 0 R 126 0 R ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 127 0 R /Resources << /ExtGState << /R23 33 0 R >> /Font << /F2 128 0 R /R78 92 0 R /R100 129 0 R /R102 132 0 R /R98 136 0 R /F1 140 0 R /R38 50 0 R /R24 65 0 R /R96 141 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 145 0 R 146 0 R 147 0 R 148 0 R 149 0 R 150 0 R 151 0 R 152 0 R 153 0 R ] >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 154 0 R /Resources << /ExtGState << /R23 33 0 R >> /Font << /F2 155 0 R /R78 92 0 R /R100 129 0 R /F1 156 0 R /R102 132 0 R /R118 157 0 R /R40 40 0 R /R98 136 0 R /R126 161 0 R /R124 166 0 R /R122 169 0 R /R120 172 0 R /R34 57 0 R /R24 65 0 R /R96 141 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 175 0 R 176 0 R 177 0 R 178 0 R 179 0 R 180 0 R 181 0 R 182 0 R 183 0 R 184 0 R ] >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 185 0 R /Resources << /XObject << /R144 186 0 R >> /ExtGState << /R23 33 0 R >> /Font << /R78 92 0 R /R98 136 0 R /R126 161 0 R /R124 166 0 R /R138 187 0 R /R96 141 0 R /R145 190 0 R /F2 193 0 R /R147 194 0 R /R140 198 0 R /R142 201 0 R /R100 129 0 R /R102 132 0 R /R149 205 0 R /F1 208 0 R /R38 50 0 R /R24 65 0 R /R26 72 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 209 0 R 210 0 R 211 0 R 212 0 R 213 0 R ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Contents 214 0 R /Resources << /XObject << /R167 215 0 R /R166 216 0 R /R165 217 0 R /R164 218 0 R /R169 219 0 R /R168 220 0 R >> /ExtGState << /R23 33 0 R >> /Font << /F2 221 0 R /R78 92 0 R /R142 201 0 R /R100 129 0 R /F1 222 0 R /R102 132 0 R /R98 136 0 R /R126 161 0 R /R124 166 0 R /R38 50 0 R /R24 65 0 R /R96 141 0 R /R26 72 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /Group 223 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 224 0 R 225 0 R 226 0 R 227 0 R 228 0 R 229 0 R ] /Type /Page >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 230 0 R /Resources << /XObject << /R192 231 0 R /R193 232 0 R /R190 233 0 R /R191 234 0 R /R185 235 0 R /R187 236 0 R /R194 237 0 R /R186 238 0 R /R189 239 0 R /R188 240 0 R >> /ExtGState << /R23 33 0 R >> /Font << /F2 241 0 R /R78 92 0 R /R140 198 0 R /R100 129 0 R /R98 136 0 R /R126 161 0 R /F1 242 0 R /R38 50 0 R /R138 187 0 R /R24 65 0 R /R96 141 0 R /R26 72 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 243 0 R 244 0 R 245 0 R 246 0 R 247 0 R 248 0 R 249 0 R 250 0 R 251 0 R ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 252 0 R /Resources << /XObject << /R211 253 0 R /R210 254 0 R /R206 255 0 R /R207 256 0 R /R204 257 0 R /R205 258 0 R /R208 259 0 R /R209 260 0 R /R203 261 0 R /R202 262 0 R >> /ExtGState << /R23 33 0 R >> /Font << /R100 129 0 R /R102 132 0 R /F2 263 0 R /F1 264 0 R /R34 57 0 R /R24 65 0 R /R26 72 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 265 0 R 266 0 R ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 267 0 R /Resources << /ExtGState << /R23 33 0 R >> /Font << /F2 268 0 R /F1 269 0 R /R24 65 0 R /R38 50 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 270 0 R 271 0 R 272 0 R 273 0 R 274 0 R 275 0 R 276 0 R 277 0 R 278 0 R 279 0 R 280 0 R 281 0 R 282 0 R 283 0 R 284 0 R 285 0 R 286 0 R 287 0 R 288 0 R 289 0 R 290 0 R 291 0 R 292 0 R 293 0 R 294 0 R 295 0 R 296 0 R 297 0 R 298 0 R 299 0 R 300 0 R 301 0 R 302 0 R 303 0 R 304 0 R 305 0 R 306 0 R 307 0 R 308 0 R 309 0 R 310 0 R 311 0 R 312 0 R 313 0 R 314 0 R 315 0 R 316 0 R 317 0 R 318 0 R 319 0 R 320 0 R 321 0 R 322 0 R 323 0 R 324 0 R 325 0 R 326 0 R 327 0 R 328 0 R ] >> endobj 12 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 13 0 obj << /Length 21304 >> stream q q q 0.10000 0 0 0.10000 0 0 cm /R23 gs 0 g q 10 0 0 10 0 0 cm BT /R24 14.34620 Tf 1 0 0 1 78.31090 700.57900 Tm [ (Aligning) -250.01300 (Books) -250.01100 (and) -249.99100 (Mo) 10.01330 (vies\072) -310 (T) 91.98970 (o) 9.99625 (wards) -250.00300 (Story\055lik) 9.98944 (e) -250 (V) 37.01370 (isual) -250.00500 (Explanations) -249.99500 (by) ] TJ 105.05600 -17.93280 Td [ (W) 65.00200 (atching) -250.00800 (Mo) 10.01330 (vies) -250.01200 (and) -249.99100 (Reading) -249.99500 (Books) ] TJ /R26 11.95520 Tf -69.02420 -37.85820 Td [ (Y) 110.99500 (ukun) -250.00200 (Zhu) ] TJ /R28 7.97010 Tf 56.45200 4.33906 Td [ <03> -0.29866 ] TJ /R30 7.97010 Tf 3.98516 0 Td [ (\073) -0.09955 ] TJ /R32 7.97010 Tf 2.35195 0 Td [ (1) -0.29866 ] TJ /R26 11.95520 Tf 19.67700 -4.33906 Td [ (Ryan) -249.99500 (Kiros) ] TJ /R26 7.97010 Tf 54.78980 4.33906 Td (\052\0541) Tj /R26 11.95520 Tf 25.40430 -4.33906 Td [ (Richard) -250.00800 (Zemel) ] TJ /R32 7.97010 Tf 71.38400 4.33906 Td [ (1) -0.30019 ] TJ /R26 11.95520 Tf 19.67580 -4.33906 Td [ (Ruslan) -250.00400 (Salakhutdino) 14.99330 (v) ] TJ /R32 7.97010 Tf 105.09700 4.33906 Td [ (1) -0.30019 ] TJ /R26 11.95520 Tf -305.19200 -18.28590 Td [ (Raquel) -249.99400 (Urtasun) ] TJ /R32 7.97010 Tf 74.70700 4.33789 Td [ (1) -0.30019 ] TJ /R26 11.95520 Tf 19.67620 -4.33789 Td [ (Antonio) -250.01600 (T) 80.01470 (orralba) ] TJ /R32 7.97010 Tf 82.38280 4.33789 Td [ (2) -0.30019 ] TJ /R26 11.95520 Tf 19.67620 -4.33789 Td [ (Sanja) -249.98900 (Fidler) ] TJ /R32 7.97010 Tf 58.11370 4.33789 Td [ (1) -0.30019 ] TJ -280.58900 -15.36480 Td [ (1) -0.29866 ] TJ /R26 11.95520 Tf 4.73203 -4.33789 Td [ (Uni) 24.99460 (v) 14.98620 (ersity) -249.98900 (of) -250.01400 (T) 80.01470 (oronto) ] TJ /R32 7.97010 Tf 118.44000 4.33789 Td [ (2) -0.30019 ] TJ /R26 11.95520 Tf 4.73203 -4.33789 Td [ (Massachusetts) -249.99300 (Institute) -250.01600 (of) -250.01400 (T) 70.01640 (echnology) ] TJ /R34 8.96640 Tf -194.71700 -13.94800 Td [ (f) -0.89990 ] TJ /R36 8.96640 Tf 4.60703 0 Td (yukun\054rkiros\054zemel\054rsalakhu\054urtasun\054fidler) Tj /R34 8.96640 Tf 225.95300 0 Td [ (g) -0.90126 ] TJ /R36 8.96640 Tf 4.60703 0 Td [ (\100cs\056toronto\056edu\054) -600.01600 (torralba\100csail\056mit\056edu) ] TJ /R24 11.95520 Tf -164.29400 -41.04610 Td (Abstract) Tj /R38 9.96260 Tf -83.92770 -23.91090 Td [ (Books) -205.98900 (ar) 36.98520 (e) -206.00900 (a) -206 (ric) 14.98650 (h) -204.98000 (sour) 36.99630 (ce) -206.01300 (of) -205.99300 (both) -205.98300 <026e652d6772> 14.99260 (ained) -205.98800 (information\054) ] TJ -11.95510 -11.95510 Td [ (how) -270.99000 (a) -271.98900 (c) 15.01280 (har) 14.99380 (acter) 111.01100 (\054) -275.99100 (an) -271.00400 (object) -272.01700 (or) -271.00300 (a) -271.00900 (scene) -271.01600 (looks) -272.01300 (lik) 10.00810 (e) 9.99343 (\054) -275.99100 (as) -271.00100 (well) -272.01100 (as) ] TJ 11.95510 TL T* [ (high\055le) 15 (vel) -240 (semantics\054) -242.00500 (what) -240.00800 (someone) -239.98400 (is) -241.00500 (t) 0.98758 (h) -1.01454 (i) 0.98758 (nking) 10.01550 (\054) -243.01600 (feeling) -240.01600 (and) ] TJ T* [ (how) -256.01200 (these) -255.98700 (states) -256.98900 (e) 15.01220 (volve) -255.99800 (thr) 44.99130 (ough) -256.01600 (a) -255.99200 (story) 54.99080 (\056) -329.00100 (This) -256.00400 (paper) -256.01900 (aims) -256.99400 (to) ] TJ T* [ (align) -270.00400 (books) -270.99700 (to) -269.98200 (their) -270.01300 (mo) 10.00570 (vie) -269.99500 (r) 37.01960 (eleases) -271.00300 (in) -269.98200 (or) 36.98650 (der) -269.98700 (to) -269.98200 (pr) 44.98510 (o) 10.00320 (vide) -271.01100 (ric) 14.98770 (h) ] TJ T* [ (descriptive) -248.01100 (e) 19.99240 (xplanations) -247.98300 (for) -247 (visual) -248.01100 (content) -248.01700 (that) -248.01300 (go) -247.98800 (semanti\055) ] TJ 11.95630 TL T* [ (cally) -275.00300 (far) -274.99500 (be) 29.99560 (yond) -275.00200 (the) -273.98600 (captions) -275.01700 (available) -275.02000 (in) -275 (curr) 36.98650 (ent) -275.00600 (datasets\056) ] TJ 11.95470 TL T* [ (T) 91.99870 (o) -416.00400 (align) -415 (mo) 10.00630 (vies) -416.00400 (and) -415.99400 (books) -415.99200 (we) -414.98100 (e) 19.99180 (xploit) -415.98200 (a) -416.00400 (neur) 14.99010 (al) -415.99700 (sentence) ] TJ T* [ (embedding) -357.99200 (that) -357.01400 (is) -358.00500 (tr) 14.99140 (ained) -358 (in) -356.98700 (an) -358.00900 (unsupervised) -357.00200 (way) -358.00400 (fr) 44.98510 (om) -358.01100 (a) ] TJ T* [ (lar) 36.99260 (g) 10.00320 (e) -357.98400 (corpus) -357.01600 (of) -358.00600 (books\054) -385.01400 (as) -356.98800 (well) -357.99500 (as) -358.00800 (a) -356.99400 (video\055te) 20.00400 (xt) -358.01600 (neur) 14.99010 (al) -358.00600 (em\055) ] TJ T* [ (bedding) -222.99400 (for) -221.98400 (computing) -222.98400 (similarities) -222.01600 (between) -223.00400 (mo) 10.00570 (vie) -222.00500 (clips) -223 (and) ] TJ T* [ (sentences) -360.00600 (in) -360.00600 (the) -360.01100 (book\056) -639.99700 (W) 91.98590 (e) -361.00300 (pr) 44.98390 (opos) 1.00106 (e) -361.00300 (a) -360.01300 (conte) 20.00400 (xt\055awar) 37.00360 (e) -359.98400 (CNN) ] TJ 11.95590 TL T* [ (to) -298.99700 (combine) -298.99200 (information) -298.98600 (fr) 44.98640 (om) -299.00200 (multiple) -299.01100 (sour) 36.99630 (ces\056) -456.98300 (W) 91.98710 (e) -299.01400 (demon\055) ] TJ 11.95510 TL T* [ (str) 14.99750 (ate) -233.99200 (good) -235 (quantitative) -233.98700 (performance) -234.99000 (for) -233.98300 (mo) 10.00570 (vie\057book) -235.00200 (align\055) ] TJ T* [ (ment) -283.98100 (and) -284.01600 (show) -284.00200 (se) 15.01830 (ver) 15.01470 (al) -284.01900 (qualitative) -283.98300 (e) 19.99180 (xamples) -284.01400 (that) -284.00700 (showcase) ] TJ T* [ (the) -249.99000 (diver) 10.00810 (sity) -249.98200 (of) -249.98500 (tasks) -250.01700 (our) -249.98200 (model) -249.98200 (can) -249.99100 (be) -249.99600 (used) -249.98500 (for) 111.01800 (\056) ] TJ /R24 11.95520 Tf 24.45000 TL T* [ (1\056) -249.99000 (Intr) 18.01460 (oduction) ] TJ /R26 9.96260 Tf 11.95510 -18.92890 Td [ (A) -335 (truly) -334.98200 (intelligent) -334.99300 (machine) -334.98700 (needs) -334.99600 (to) -334.99100 (not) -336.00500 (only) -334.98100 (parse) -335.00300 (the) ] TJ -11.95510 -11.95630 Td [ (surrounding) -233.99800 (3D) -233.99300 (en) 39.99460 (vironment\054) -237.99700 (b) 20.00160 (ut) -233.98800 (also) -233.98700 (understand) -234.01200 (wh) 4.98691 (y) -235.01500 (peo\055) ] TJ 11.95510 TL T* [ (ple) -318.01700 (tak) 10.00630 (e) -319.01000 (cert) 1.01454 (ain) -318.99800 (actions\054) -335.01800 (what) -318.01400 (the) 14.98520 (y) -319 (will) -318 (do) -318.01600 (ne) 15.01710 (xt\054) -334.98800 (what) -318.99600 (the) 14.98520 (y) ] TJ T* [ (could) -284.01300 (possibly) -283.00400 (be) -283.99100 (thinking\054) -291.98100 (and) -283.00500 (e) 25.01050 (v) 14.98280 (en) -283.99100 (try) -283.98200 (to) -283 (empathize) -284.01100 (with) ] TJ T* [ (them\056) -770.00600 (In) -402.99000 (this) -403.98500 (quest\054) -441.98300 (language) -403.01600 (will) -403.00600 (play) -403.99900 (a) -402.99600 (crucial) -402.98900 (role) -404.00600 (in) ] TJ T* [ (grounding) -394.01300 (visual) -395.00600 (information) -394.00900 (to) -394.98100 (high\055le) 24.99830 (v) 14.98280 (el) -394.01000 (semantic) -395.00800 (con\055) ] TJ T* [ (cepts\056) -564.01100 (Only) -334.00200 (a) -335.00800 (fe) 25.00810 (w) -333.98100 (w) 10 (ords) -334.99100 (in) -334.01000 (a) -335.00800 (sentence) -334.99800 (may) -334.01000 (con) 39.99820 (v) 14.98280 (e) 15.01220 (y) -334.99800 (really) ] TJ 11.95590 TL T* [ (rich) -204.98100 (semantic) -205 (information\056) -295.00300 (Language) -205.01400 (also) -205.01000 (represents) -205.01500 (a) -204.99000 (natu\055) ] TJ 11.95510 TL T* [ (ral) -220.00300 (means) -221.00700 (of) -220 (interaction) -220.01500 (between) -221.01500 (a) -220.00700 (nai) 24.98360 (v) 14.98280 (e) -220.98800 (user) -220 (and) -219.99800 (our) -221.01500 (vision) ] TJ T* [ (algorithms\054) -265.00400 (which) -263 (is) -261.98200 (particularly) -262.01000 (important) -261.99300 (for) -263.01700 (applications) ] TJ T* [ (such) -249.98600 (as) -249.99600 (social) -249.98000 (robotics) -250.00900 (or) -249.99600 (assisti) 25.01180 (v) 14.98280 (e) -250.00200 (dri) 24.98480 (ving\056) ] TJ 11.95510 -11.95510 Td [ (Combining) -359.01200 (images) -359.01600 (or) -360.01700 (videos) -359.01400 (with) -359.01600 (language) -358.98400 (has) -360.01100 (gotten) ] TJ -11.95510 -11.95510 Td [ <7369676e690263616e74> -316.01700 (attention) -315.01000 (in) -316.01300 (the) -316.01800 (past) -316.01200 (year) 39.98600 (\054) -331.98100 (partly) -314.98900 (due) -315.98200 (to) -316.01300 (the) -316.01800 (cre\055) ] TJ 11.95590 TL T* [ (ation) -344 (of) -344.01900 (CoCo) -343.00400 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 115.17800 158.44700 Tm (20) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 125.14100 158.44700 Tm [ (\135\054) -366.99900 (Microsoft\047) 54.99820 (s) -344.01500 (lar) 17.99700 (ge\055scale) -343.98400 (captioned) -343.99900 (im\055) ] TJ -75.02850 -11.95510 Td [ (age) -236.98400 (dataset\056) -306.01300 (The) -237.01400 <02656c64> -236.99700 (has) -237.01200 (tackled) -237.00800 (a) -236.98500 (di) 24.98850 (v) 14.98280 (erse) -236.98500 (set) -237.00900 (of) -237.01700 (tasks) -237 (such) ] TJ 11.95510 TL T* [ (as) -331.00200 (captioning) -329.98500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 110.37600 134.53700 Tm (15) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 120.33800 134.53700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 126.12700 134.53700 Tm (13) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 136.08900 134.53700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 141.87700 134.53700 Tm (40) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 151.84000 134.53700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 157.61800 134.53700 Tm (39) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 167.58000 134.53700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 173.36900 134.53700 Tm (24) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 183.33100 134.53700 Tm [ (\135\054) -351.00100 (alignment) -329.98400 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 239.09200 134.53700 Tm (13) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 249.05400 134.53700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 254.84300 134.53700 Tm (17) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 264.80500 134.53700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 270.59300 134.53700 Tm (38) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 280.55600 134.53700 Tm (\135\054) Tj -230.44400 -11.95510 Td [ (Q\046A) -284.99300 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 78.40590 122.58200 Tm (22) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 88.36840 122.58200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 93.69840 122.58200 Tm (21) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 103.66100 122.58200 Tm [ (\135\054) -293.01100 (visual) -284.98600 (model) -284.98700 (learning) -284.99400 (from) -285.00100 (te) 14.98030 (xtual) -284.99200 (descrip\055) ] TJ -53.54880 -11.95510 Td [ (tions) -240.00800 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 75.19800 110.62700 Tm (9) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 80.17930 110.62700 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 85.06090 110.62700 Tm (29) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 95.02340 110.62700 Tm [ (\135\054) -242 (and) -239.01300 (semantic) -240.01400 (visual) -240.01300 (search) -240.00400 (with) -240.01600 (natural) -239.98700 (multi\055) ] TJ -44.91130 -11.95590 Td [ (sentence) -249.99100 (queries) -249.98600 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 121.48400 98.67110 Tm (19) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 131.44600 98.67110 Tm (\135\056) Tj ET Q 3.98000 w 0 G 501.12100 904.14800 m 1446.11000 904.14800 l S q 10 0 0 10 0 0 cm BT /R40 5.97760 Tf 1 0 0 1 60.14100 83.81290 Tm [ <03> -0.90058 ] TJ /R26 7.97010 Tf 4.31680 -2.81289 Td [ (Denotes) -250 (equal) -250.01300 (contrib) 19.99660 (ution) ] TJ ET Q q 3088.62000 5064.65000 2362.29000 738.99200 re W n q 2407.80000 0 0 886.87300 3088.62000 4916.77000 cm /R42 Do Q Q q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 308.86200 496.02200 Tm [ (Figure) -224.01700 (1\072) ] TJ /R26 8.96640 Tf 38.95080 0 Td [ (Shot) -223.98600 (from) -223.98900 (the) -223.05500 (mo) 14.99740 (vie) ] TJ /R38 8.96640 Tf 75.14020 0 Td [ (Gone) -223.99700 (Girl) ] TJ /R26 8.96640 Tf 36.37380 0 Td [ (\054) -229.00200 (along) -223.98100 (with) -224.00800 (the) -223.05500 (subtitle\054) ] TJ -150.46500 -10.95900 Td [ (aligned) -200.98900 (with) -202.00500 (the) -201.02200 (book\056) -294.00400 (W) 80.01060 (e) -201.98800 (reason) -201.01600 (about) -201.98000 (the) -201.02200 (visual) -202.00700 (and) -200.98900 (dialog) -201.01100 (\050te) 15.00830 (xt\051) ] TJ 10.95900 TL T* [ (alignment) -249.98100 (between) -249.99200 (the) -249.98900 (mo) 14.99470 (vie) -249.98900 (and) -250 (a) -249.99500 (book\056) ] TJ /R26 9.96260 Tf 11.95510 -22.16600 Td [ (Books) -244.00300 (pro) 14.98520 (vide) -243.00600 (us) -243.98600 (with) -242.99400 (v) 14.98280 (ery) -244.00300 (descripti) 25 (v) 14.98280 (e) -244.00300 (te) 14.98280 (xt) -243.00800 (that) -243.98400 (con) 39.99820 (v) 14.98280 (e) 15.01220 (ys) ] TJ -11.95510 -11.95470 Td [ (both) -399.00900 <026e652d677261696e6564> -399.99000 (visual) -399.00400 (details) -399.00700 (\050ho) 24.98110 (w) -400.00700 (things) -398.99700 (look) -399.00700 (lik) 10.01790 (e\051) -400.02100 (as) ] TJ 11.95510 TL T* [ (well) -338.99000 (as) -339 (high\055le) 24.99830 (v) 14.98280 (el) -339 (semantics) -339.99200 (\050what) -338.99200 (people) -338.98800 (think\054) -361.00300 (feel\054) -361.97900 (and) ] TJ T* [ (ho) 24.98600 (w) -402.98900 (their) -403.99900 (states) -403.01400 (e) 25.01050 (v) 20.00160 (olv) 14.99500 (e) -402.99600 (through) -403.98200 (a) -402.99600 (story\051\056) -770.00900 (This) -403.00600 (source) -404.00600 (of) ] TJ T* [ (kno) 24.99090 (wledge\054) -284.99700 (ho) 24.98600 (we) 25.01300 (v) 14.98280 (er) 39.98600 (\054) -285.00900 (does) -277.00300 (not) -278.01200 (come) -277.98300 (with) -278.00800 (associated) -278.01200 (visual) ] TJ T* [ (information) -309.00200 (that) -308.99300 (w) 10 (ould) -308.98500 (enable) -309.00200 (us) -308.99700 (to) -308.99700 (ground) -308.98800 (it) -308.99300 (with) -308.98300 (natural) ] TJ 11.95590 TL T* [ (language\056) -310.01000 (Grounding) -250 (descr) 1.01944 (iptions) -249.99300 (in) -249.98800 (books) -248.99300 (to) -249.98800 (vision) -250.00200 (w) 10.00320 (ould) ] TJ 11.95510 TL T* [ (allo) 24.99090 (w) -283.00900 (us) -283 (to) -283 (get) -283.00400 (te) 14.98030 (xtual) ] TJ /R38 9.96260 Tf 91.73910 0 Td [ (e) 19.99180 (xplanations) ] TJ /R26 9.96260 Tf 53.53980 0 Td [ (or) -283.01200 (stories) -282.98700 (about) -282.99200 (the) -283.00200 (vi\055) ] TJ -145.27900 -11.95470 Td [ (sual) -395.01700 (w) 10 (orld) -395.01700 (rather) -396.01500 (than) -395.02000 (short) ] TJ /R38 9.96260 Tf 119.19300 0 Td (captions) Tj /R26 9.96260 Tf 37.70080 0 Td [ (a) 19.99180 (v) 24.98110 (ailable) -395.01700 (in) -394.98300 (current) ] TJ -156.89400 -11.95510 Td [ (datasets\056) -409.98800 (It) -283.00700 (could) -284.01400 (also) -282.99700 (pro) 14.98280 (vide) -282.99700 (us) -283.00200 (with) -284.00700 (a) -283.01700 (v) 14.98280 (ery) -283.01700 (lar) 17.99700 (ge) -283.99200 (amount) ] TJ T* [ (of) -249.99500 (data) -249.99500 (\050with) -250.01700 (tens) -249.98300 (of) -249.99500 (thousands) -249.99700 (books) -250.01200 (a) 19.99180 (v) 24.98110 (ailable) -249.98300 (online\051\056) ] TJ 11.95510 -12.11990 Td [ (In) -319.98300 (this) -320 (paper) 39.99090 (\054) -337 (we) -319.98800 (e) 15.01220 (xploit) -319.99100 (the) -318.99600 (f) 9.99343 (act) -319.98600 (that) -320.01000 (man) 14.99260 (y) -319.98100 (books) -320 (ha) 19.99670 (v) 14.98280 (e) ] TJ -11.95510 -11.95510 Td [ (been) -354.99700 (turned) -354.98500 (into) -353.99500 (mo) 14.99500 (vies\056) -625.01800 (Books) -353.98500 (and) -354.99500 (their) -354.99000 (mo) 14.99750 (vie) -354.99000 (releases) ] TJ T* [ (ha) 19.99670 (v) 14.98280 (e) -299.01400 (a) -299.99400 (lot) -298.98900 (of) -299.98700 (common) -299.01400 (kno) 24.99090 (wledge) -300.01900 (as) -299.00900 (well) -298.99400 (as) -299.98900 (the) 14.98770 (y) -299.00400 (are) -299.99900 (com\055) ] TJ T* [ (plementary) -276.00800 (in) -275.98100 (man) 14.99010 (y) -275.98800 (w) 10 (ays\056) -388.98400 (F) 14.99260 (or) -275.99300 (instance\054) -282.01200 (books) -276.98800 (pro) 14.98280 (vide) -276.01800 (de\055) ] TJ T* [ (tailed) -256.01900 (descriptions) -255.99400 (about) -256.01900 (the) -255.98700 (intentions) -255.98700 (and) -257.01100 (mental) -256.01100 (states) -256.01600 (of) ] TJ T* [ (the) -336.01500 (characters\054) -357.00700 (while) -334.98600 (mo) 14.99500 (vies) -336.00800 (are) -335.99300 (better) -334.99300 (at) -336.01700 (capturing) -335.99800 (visual) ] TJ 11.95590 TL T* [ (aspects) -249.98500 (of) -249.99500 (the) -249.99000 (settings\056) ] TJ 11.95510 -12.11910 Td [ (The) -281.98700 <02727374> -280.98700 (challenge) -282.01900 (we) -281.99200 (need) -281.01200 (to) -282.02200 (address\054) -290.01500 (and) -281.00700 (the) -281.98200 (focus) -281.98200 (of) ] TJ -11.95510 -11.95510 Td [ (this) -256.01100 (paper) 39.99090 (\054) -257.99400 (is) -255.98400 (to) -255.98400 (align) -256.01600 (books) -256.01100 (with) -256.01100 (their) -255.98700 (mo) 14.99750 (vie) -255.98700 (releases) -257.01600 (in) -255.98700 (or) 19.99670 (\055) ] TJ 11.95510 TL T* [ (der) -357.00200 (to) -358.00600 (obtain) -357.01400 (rich) -358.01400 (descriptions) -356.99400 (for) -357.97900 (the) -356.98900 (visual) -357.01400 (content\056) -633.00700 (W) 79.98660 (e) ] TJ T* [ (aim) -251.00200 (to) -251.98500 (align) -250.99700 (the) -251.01000 (tw) 10.00810 (o) -251.01200 (sources) -251.98700 (with) -250.99200 (tw) 10.00810 (o) -251.01200 (types) -250.99700 (of) -251.99700 (information\072) ] TJ /R38 9.96260 Tf 11.95630 TL (visual) ' /R26 9.96260 Tf 23.80000 0 Td [ (\054) -223.01900 (where) -216.00800 (the) -215.99600 (goal) -215.99100 (is) -215.99100 (to) -215.99300 (link) -216.01800 (a) -216.00800 (mo) 14.99750 (vie) -215.99300 (shot) -216.01800 (to) -215.99300 (a) -216.00800 (book) -215.98400 (para\055) ] TJ -23.80000 -11.95510 Td [ (graph\054) -310.01700 (and) ] TJ /R38 9.96260 Tf 45.61290 0 Td [ (dialo) 9.98853 (g) ] TJ /R26 9.96260 Tf 25.36410 0 Td [ (\054) -309.98500 (where) -297.01400 (we) -297.99000 (w) 10.00320 (ant) -297.98000 (to) -298.01900 <026e64> -296.98500 (correspondences) ] TJ -70.97700 -11.95470 Td [ (between) -299.98200 (sentences) -300.01600 (in) -300.01600 (the) -299.00200 (mo) 14.99750 (vie\047) 55.01530 (s) -299.98400 (subtitle) -299.98900 (and) -299.98400 (sentences) -300.01900 (in) ] TJ 11.95510 TL T* [ (the) -267.00700 (book) -266.01400 (\050Fig\056) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 368.03000 152.73100 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R26 9.96260 Tf 1 0 0 1 373.01100 152.73100 Tm [ (\051\056) -360.01800 (W) 79.98660 (e) -267.01900 (introduce) -266.99000 (a) -266 (no) 14.98770 (v) 14.98280 (el) -267.00900 (sentence) -267.00900 (similarity) ] TJ -64.14920 -11.95510 Td [ (measure) -312.01900 (based) -312.01900 (on) -312.99700 (a) -311.99200 (neural) -311.98200 (sentence) -311.98200 (embedding) -311.99200 (trained) -312.99200 (on) ] TJ 11.95510 TL T* [ (millions) -272.01600 (of) -271.01100 (sentences) -271.98200 (from) -271.00400 (a) -271.99900 (lar) 17.99700 (ge) -271.01400 (corpus) -272.01800 (of) -271.01400 (books\056) -375.00600 (On) -271.98400 (the) ] TJ 11.95590 TL T* [ (visual) -381.98800 (side\054) -415.01200 (we) -382.01700 (e) 15.01220 (xtend) -381.99800 (the) -382.00500 (neural) -382.01000 (image\055sentence) -381.99500 (embed\055) ] TJ 11.95510 TL T* [ (dings) -280.98300 (to) -281 (the) -279.98500 (video) -280.99500 (domain) -280.98300 (and) -281.00700 (train) -281.00200 (the) -279.98300 (model) -280.98700 (on) -281.00200 (D) 39.99820 (VS) -280.99200 (de\055) ] TJ T* [ (scriptions) -259.99300 (of) -259.01300 (mo) 14.99500 (vie) -259.98800 (clips\056) -339.01700 (Our) -259.01100 (approach) -259.99100 (combines) -260.00600 (dif) 24.98600 (ferent) ] TJ T* [ (similarity) -227.98900 (measures) -227.98900 (and) -227.99600 (tak) 10.00810 (es) -228.00100 (into) -228.01600 (account) -227.99100 (conte) 14.98770 (xtual) -227.98200 (infor) 20.00650 (\055) ] TJ -13.74100 -29.88790 Td (1) Tj ET Q Q Q q q 1 1 1 rg /a0 gs 48.40600 786.42200 515.18800 -52.69900 re f q /s5 gs /x6 Do Q q /s7 gs /x8 Do Q q /s9 gs /x10 Do Q q /s11 gs /x12 Do Q Q Q Q q 1 0 0 1 0 0 cm BT /F1 12 Tf 14.40000 TL ET 1 1 1 rg n 270 47 72 14 re f* 0.50000 0.50000 0.50000 rg BT /F2 9 Tf 10.80000 TL ET BT 1 0 0 1 301.50000 50 Tm (19) Tj T* ET Q endstream endobj 14 0 obj << /Filter /FlateDecode /Resources << /ExtGState << /a0 << /CA 1 /ca 1 >> >> /XObject << /x18 15 0 R >> >> /Length 28 /Group << /Type /Group /S /Transparency /CS /DeviceRGB /I true >> /BBox [ 78 746 96 765 ] /Type /XObject /Subtype /Form >> stream x+O4PH/VЯ0Pp 0 endstream endobj 15 0 obj << /Filter /FlateDecode /Resources 16 0 R /Length 107 /Type /XObject /BBox [ 78 746 96 765 ] /Subtype /Form >> stream xe AC̬wʠ =p,?]%+H-
Jc "82w8VSnGW;"
endstream
endobj
16 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
17 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x15 18 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 67 752 84 775 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ04Up
0
endstream
endobj
18 0 obj
<<
/Filter /FlateDecode
/Resources 19 0 R
/Length 228
/Type /XObject
/BBox [ 67 752 84 775 ]
/Subtype /Form
>>
stream
xeQKn!s ?FPav6R٪TS.
b];15YyR
{7QL.\:Rv/x9l+L7h%1!}i/AI(kz"U&,YO![R hg{3}4/GyYF:!w}Gn+'xJcO9i뽼_-:`
endstream
endobj
19 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
20 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x24 21 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 133 751 479 772 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Qp
0
endstream
endobj
21 0 obj
<<
/Filter /FlateDecode
/Resources 22 0 R
/Length 53008
/Type /XObject
/BBox [ 133 751 479 772 ]
/Subtype /Form
>>
stream
xlK8,8?DKs9mavd
{f-8*2Y@H >חk}y}uf`v)_s}1z#*Gw_gX jow\o'1c|Z^G<
A̺X}ay?IT|y~L.[ {Ȟb\3-3]_'X\竵0{+_۾oY-wj+ B;)Aa=/
1~a(>}m_K' >*Q1:?OPnGsQ_`?mzN
?WBB!lU n'S 1Ipzχ1xF_09,}?e){:̹~Gu7g {0mD`pdS)>;ys`Pµx'O
+혝 oI~!p7̵&m=dC+>5ALV"I1"S*CKha#03L4|@p灟vMRQ==*/m"r>@uqo9ͻ;Vz 'ȳ;IBzWR~ihwa8PlT>]fZ8y52)r빃W{B7G/q)}M]elS(k=5^A] сڸc'w!Ll'M=Ȥ뎓z;#կN>\q6r>|8}}Lr$Mo5:>syls+Ǵ`C?/9{Y||jD u#i^\rJ8a|oY+Rxl<{[؏s迨nYaյ[?/^u;+ ,bhL}wrj:ɴdM8^q|m'aQ6;0k_ar|E8HɃpN('&r`V;Rs볩Nk\&W=DDG#5]~M{VyCSRH9ܜXˏt7߾i9ADگ
USڝ|{(hbIRZXg @6Hɹ0Dq;si\Tv<1?mUŏV7bF
p7Y9sƨ*.uz_LբřsB݃ɹ^'_Yae91f+n%A̶d
Ċc#q
CF=Mt'\&ξ)z;L-4:n}41j:mӶ_)\݇ |T@uAƓxAN6Ql;}Ɲ{~_y"g%]nCi']I~,xa?feoɌr0u?V=phldx.Y;~nZL603> \`ęܓ