1927 — Page 397

Blue Books 香港計冊 All AI Reviewed

The user wants me to proofread OCR output of a historical prison record from Hong Kong (1927). The text appears to be a table showing weekly prisoner counts broken down by Europeans and Chinese/Coloured, with males and females.

Let me analyze the OCR text carefully. It's a table with columns:

  • WEEK (date)
  • EUROPEANS: Males, Females, Total
  • CHINESE OR COLOURED: Males, Females, Total

The OCR has many errors: garbled characters (喃喃, นค, པ, ་, 上, etc.), misaligned columns, broken numbers, and formatting issues.

I need to reconstruct this as a proper Markdown table. Let me parse the data week by week.

First, let me understand the structure. The header shows:

  • EUROPEANS: Males, Females, Total
  • CHINESE OR COLOURED: Males, Females, Total

But the OCR shows "EUROPEANS." then "CHINESE OR" then "EUROPEANS." then "COLOURED." then "CHINESE OK COLOURED." - this is messy.

Looking at the data rows, each week seems to have two lines: one for Europeans, one for Chinese/Coloured? Or maybe it's side-by-side columns.

Actually, looking at the first data row:

"January 5 1927. : 1,017 42 1,063 July 6 13 : 1,149 98 1.251"

This seems like two separate weeks side by side? January 5 and July 6? That doesn't make sense chronologically.

Wait, the OCR might have read two columns of the table side by side. The original table probably had two columns: left side for first half of year, right side for second half? Or maybe it's a wide table with European and Chinese sections.

Let me look at the pattern. The header says:

"1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927."

Then columns: EUROPEANS (Males, Females, Total), CHINESE OR COLOURED (Males, Females, Total)

But the data shows dates like "January 5", "12", "19", "26", "February 2", "9", "16", "23", "March 2", "9", "16", "23", "30", "April 6", "13", "20", "27", "May 4", "11", "18", "25", "June 1", "8", "15", "22", "29", "July 6", "13", "20", "27", "August 3", "10", "17", "24", "31", "September 7", "14", "21", "28", "October 5", "12", "19", "26", "November 2", "9", "16", "23", "30", "December 7", "14", "21", "28"

That's 52 weeks (one per week). The OCR seems to have interleaved them or read in two columns.

Looking at the raw text, it appears the OCR read the table in two vertical columns (left and right) but merged them line by line. For example, first line has "January 5" and "July 6" - those are week 1 and week 27 approximately.

Let me try to separate the two columns. The original table was likely printed in two columns on the page to save space: left column Jan-Jun, right column Jul-Dec.

So I need to split the data into two halves and then interleave them chronologically? Or present as two separate tables? The instruction says "restore column reading order — if text was originally in multiple columns but OCR read it in the wrong order, reorder it to make logical sense (left-to-right, top-to-bottom)."

Since the original was two columns side by side (left column first half, right column second half), the logical reading order is to read the left column top to bottom, then right column top to bottom. But for a weekly time series, chronological order makes more sense. However, the instruction says "reorder it to make logical sense (left-to-right, top-to-bottom)" - meaning the original physical layout was left column then right column. But the content is chronological weeks. The most logical presentation for the data is chronological order.

I think I should reconstruct the table in chronological order (Week 1 to Week 52) with proper columns.

Let me parse all the data.

First, let me extract all the week entries from the OCR text. The text is a mess with many OCR artifacts. I'll go line by line.

The OCR text lines (as provided):

Line 1: (Y 11 )

Line 2:

Line 3: PRISON.

Line 4:

Line 5: 1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927.

Line 6:

Line 7: EUROPEANS.

Line 8:

Line 9: CHINESE OR

Line 10:

Line 11: EUROPEANS.

Line 12:

Line 13: COLOURED.

Line 14:

Line 15: CHINESE OK COLOURED.

Line 16:

Line 17: WEEK.

Line 18:

Line 19: TOTAL.

Line 20:

Line 21: WEEK.

Line 22:

Line 23: IOTAL.

Line 24:

Line 25: Males.

Line 26:

Line 27: Fe- males.

Line 28:

Line 29: I

Line 30:

Line 31: Males.

Line 32:

Line 33: Fe- males.

Line 34:

Line 35: Males.

Line 36:

Line 37: Fe- males.

Line 38:

Line 39: Males.

Line 40:

Line 41: Fe- males.

Line 42:

Line 43: 1927.

Line 44:

Line 45: January 5

Line 46:

Line 47: 1927.

Line 48:

Line 49: :

Line 50:

Line 51: 1,017

Line 52:

Line 53: 42 1,063

Line 54:

Line 55: July

Line 56:

Line 57: 6

Line 58:

Line 59: 13

Line 60:

Line 61: :

Line 62:

Line 63: 1,149

Line 64:

Line 65: 98 1.251

Line 66:

Line 67: 12

Line 68:

Line 69: 1,039

Line 70:

Line 71: }

Line 72:

Line 73: 41 1,088

Line 73:

Line 74: 13

Line 75:

Line 76: 5

Line 77:

Line 78: 1.135

Line 79:

Line 80: 88

Line 81:

Line 82: "}

Line 83:

Line 84: 1.228

Line 85:

Line 86: 19

Line 87:

Line 88: "

Line 89: 1,054

Line 90:

Line 91: 44

Line 92:

Line 93: | 1,105

Line 94:

Line 95: 20

Line 96:

Line 97: 1.144

Line 98:

Line 99: 94

Line 100:

Line 101: "

Line 102: 1.243

Line 103:

Line 104: 26

Line 105:

Line 106: 1,069

Line 107:

Line 108: 45 1,120

Line 109:

Line 110: 27

Line 111:

Line 112: 像像

Line 113:

Line 114: 1,144

Line 115:

Line 116: 85

Line 117:

Line 118: 1,224

Line 119:

Line 120: February 2

Line 121:

Line 122: 1,052

Line 123:

Line 124: 14

Line 125:

Line 126: 1,101

Line 127:

Line 128: August

Line 129:

Line 130: *

Line 131: 5

Line 132:

Line 133: 1,154

Line 134:

Line 135: 69

Line 136:

Line 137: 1,228

Line 138:

Line 139: 9

Line 140:

Line 141: 1,014

Line 142:

Line 143: 46

Line 144:

Line 145: 1,066

Line 146:

Line 147: 10

Line 148:

Line 149: "

Line 150: 1,123

Line 151:

Line 152: 61

Line 153:

Line 154: 1,190

Line 155:

Line 156: 16

Line 157:

Line 158: นค

Line 159:

Line 160: 5

Line 161:

Line 162: 1,036

Line 163:

Line 164: 49

Line 165:

Line 166: ""

Line 167:

Line 168: 1,090

Line 169:

Line 170: 17

Line 171:

Line 172: 6

Line 172:

Line 173: 17

Line 174:

Line 175: 1,115

Line 176:

Line 177: 57

Line 178:

Line 179: 1,181

Line 180:

Line 181: 23

Line 182:

Line 183: 3

Line 184:

Line 185: :

Line 186:

Line 187: 1,040

Line 188:

Line 189: 42 1,085

Line 190:

Line 191: 24

Line 192:

Line 193: 6 ¦

Line 194:

Line 195: 11.08

Line 196:

Line 197: 57

Line 198:

Line 199: པ

Line 200:

Line 201: 1,171

Line 202:

Line 203: March

Line 204:

Line 205: 2

Line 206:

Line 207: 9

Line 208:

Line 209: 1,064

Line 210:

Line 211: 那些

Line 212:

Line 213: 1,124

Line 214:

Line 215: 31

Line 216:

Line 217: 6

Line 218:

Line 219: ??

Line 220:

Line 221: 1,119

Line 222:

Line 223: 15

Line 224:

Line 225: 55

Line 226:

Line 227: 1,190

Line 228:

Line 229: 9

Line 229:

Line 230: 1,126

Line 231:

Line 232: 48

Line 233:

Line 234: 1,182

Line 235:

Line 236: September 7

Line 237:

Line 238: 9

Line 239: 1.138

Line 240:

Line 241: 51

Line 242:

Line 243: 16

Line 244:

Line 245: 9

Line 246:

Line 247: 27

Line 248:

Line 249: 1,092

Line 250:

Line 251: 51

Line 252:

Line 253: 1,155

Line 254:

Line 255: 14

Line 256:

Line 257: 11

Line 258:

Line 259: 23

Line 260:

Line 261: ་

Line 262:

Line 263: 19

Line 264:

Line 265: 1,082

Line 266:

Line 267: 59

Line 268:

Line 269: 1,148

Line 270:

Line 271: 30

Line 272:

Line 273: 11

Line 274:

Line 275: 1,119

Line 276:

Line 277: 62

Line 278:

Line 279: 1.193

Line 280:

Line 281: 12 12

Line 282:

Line 283: 24

Line 283:

Line 284: "

Line 285:

Line 286: April

Line 287:

Line 288: :

Line 289: :

Line 290: 1,198

Line 291:

Line 292: 1,141

Line 293:

Line 294: 55

Line 295:

Line 296: 1,205

Line 297:

Line 298: 1,164 56

Line 299:

Line 300: 1,227

Line 301:

Line 302: 1,212

Line 303:

Line 304: 59

Line 305:

Line 306: 1,278

Line 307:

Line 308: co

Line 309:

Line 310: 6

Line 311:

Line 312: 7

Line 313:

Line 314: 1,125

Line 315:

Line 316: 61

Line 317:

Line 318: 1,193 October

Line 319:

Line 320: 7

Line 321: 1,212 | 54

Line 322:

Line 323: 1,273

Line 324:

Line 325: 13

Line 326:

Line 327: 7

Line 328:

Line 329: **

Line 330:

Line 331: 1,106

Line 332:

Line 333: 72

Line 334:

Line 335: 1,185

Line 336:

Line 337: 12

Line 338:

Line 339: 10

Line 340:

Line 341: 1,203

Line 342:

Line 343: 59

Line 344:

Line 345: 1,272

Line 346:

Line 347: "

Line 348:

Line 349: 20

Line 350:

Line 351: 27

Line 352:

Line 353: 20

Line 354:

Line 355: +

Line 356:

Line 357: 1,111

Line 358:

Line 359: 75

Line 360:

Line 361: 1,195

Line 362:

Line 363: 19

Line 364:

Line 365: LI

Line 366:

Line 367: "

Line 368: 1,192

Line 369:

Line 370: 61

Line 371:

Line 372: 1,254

Line 373:

Line 374: **

Line 375:

Line 376: :

Line 377:

Line 378: 1,141

Line 379:

Line 380: 77

Line 381:

Line 382: 1.225

Line 383:

Line 384: 26

Line 385:

Line 386: 9

Line 387:

Line 388: 1,153

Line 389:

Line 390: T

Line 391:

Line 392: 56

Line 393:

Line 394: 1.220

Line 395:

Line 396: •

Line 397:

Line 398: May

Line 399:

Line 400: Go

Line 401: 3

Line 402: 5

Line 403: 1,123

Line 404:

Line 405: 2

Line 406:

Line 407: 1,206

Line 408:

Line 409: November 2

Line 410:

Line 411: :

Line 412: 1,161

Line 413:

Line 414: 48

Line 415:

Line 416: 1,210

Line 417:

Line 418: 10

Line 419:

Line 420: 6

Line 421:

Line 422: "

Line 423: 1,106

Line 424:

Line 425: 77

Line 426:

Line 427: 1,189

Line 428:

Line 429: 9

Line 430:

Line 431: 1 |

Line 432:

Line 433: 1,173

Line 434:

Line 435: 47

Line 436:

Line 437: 1,221

Line 438:

Line 439: 17

Line 439:

Line 440: 上

Line 441:

Line 442: 7

Line 443:

Line 444: **

Line 445:

Line 446: :

Line 447: 1,105

Line 448:

Line 449: 77

Line 450:

Line 451: 1,189

Line 452:

Line 453: 15

Line 454:

Line 455: 1,149

Line 456:

Line 457: 19

Line 458:

Line 459: 1,199

Line 460:

Line 461: | 24

Line 462:

Line 463: 6

Line 464:

Line 465: 13

Line 466:

Line 467: 1,125

Line 468:

Line 469: 81

Line 470:

Line 471: 1,212

Line 472:

Line 473: 23

Line 474:

Line 475: 1.166

Line 476:

Line 477: 52

Line 478:

Line 479: "

Line 480:

Line 481: 1,229

Line 482:

Line 483: June

Line 484:

Line 485: 1

Line 486:

Line 487: 7

Line 488:

Line 489: ...

Line 489:

Line 490: 1,196

Line 491:

Line 492: 3

Line 493:

Line 494: NO

Line 495:

Line 496: 1,283

Line 497:

Line 498: 30

Line 499:

Line 500: "

Line 501: 1,146 | 43

Line 502:

Line 503: 1,202

Line 504:

Line 505: x

Line 506:

Line 507: 7

Line 508:

Line 509: 1,170

Line 509:

Line 510: 82

Line 511:

Line 512: 1,259

Line 513:

Line 514: December 7

Line 515:

Line 516: 6

Line 517: 1,146

Line 518:

Line 519: 7

Line 520:

Line 521: 51

Line 522:

Line 523: 1,205

Line 524:

Line 525: 15

Line 526:

Line 527: B

Line 528:

Line 529: C

Line 530:

Line 531: 1.162

Line 532:

Line 533: 那些

Line 534:

Line 535: 1,251

Line 536:

Line 537: 14

Line 538:

Line 539: 10

Line 539:

Line 540: H

Line 541:

Line 542: 1,142

Line 543:

Line 544: 57

Line 545:

Line 546: 1,209

Line 547:

Line 548: 22

Line 549:

Line 550: 7

Line 551:

Line 552: #

Line 553:

Line 554: 1,144

Line 555:

Line 556: 75

Line 557:

Line 558: 1,229

Line 559:

Line 560: 21

Line 561:

Line 562: 10

Line 563:

Line 564: 1,148 61

Line 565:

Line 566: 1,219

Line 567:

Line 568: 29

Line 569:

Line 570: *

Line 571:

Line 572: 6 |

Line 573:

Line 574: 1,147

Line 575:

Line 576: 102

Line 577:

Line 578: 1.255

Line 579:

Line 580: 28

Line 581:

Line 582: 1,112

Line 583:

Line 584: 53

Line 585:

Line 586: 1.174

Line 587:

Line 588: 395

This is extremely messy. The OCR has read a two-column table as a single stream, interleaving left and right columns line by line.

The original table likely had two columns on the page: left column for weeks 1-26 (Jan-Jun), right column for weeks 27-52 (Jul-Dec). Each column has the same structure: Week date, European Males, European Females, European Total, Chinese Males, Chinese Females, Chinese Total.

But the OCR read it row by row across both columns, so line 1 of left column, line 1 of right column, line 2 of left, line 2 of right, etc.

Let me try to separate the two columns.

First, identify the week dates in order.

Left column (Jan-Jun) weeks:

  • January 5, 12, 19, 26
  • February 2, 9, 16, 23
  • March 2, 9, 16, 23, 30
  • April 6, 13, 20, 27
  • May 4, 11, 18, 25
  • June 1, 8, 15, 22, 29

Right column (Jul-Dec) weeks:

  • July 6, 13, 20, 27
  • August 3, 10, 17, 24, 31
  • September 7, 14, 21, 28
  • October 5, 12, 19, 26
  • November 2, 9, 16, 23, 30
  • December 7, 14, 21, 28

Now, the OCR text seems to have mixed these. Let me try to parse the data by matching patterns.

Looking at the first data lines:

"January 5" then "1927." then ":" then "1,017" then "42 1,063" then "July 6" then "13" then ":" then "1,149" then "98 1.251"

This suggests the first row of left column is Jan 5, and first row of right column is Jul 6.

But the numbers: For Jan 5: European Males 1,017? That seems high for Europeans. Wait, the header says EUROPEANS and CHINESE OR COLOURED. But the numbers are in thousands. That seems more like Chinese prisoners. Perhaps the columns are swapped? Or "EUROPEANS" might be a mislabel? Actually, in Hong Kong prisons, the majority were Chinese. But 1,000 Europeans seems too high. Let me check typical numbers. In 1927, Hong Kong population was ~800,000, Europeans maybe 10,000. Prisoners maybe hundreds. But 1,000 European prisoners? Unlikely. Perhaps the first set of numbers is for Chinese/Coloured, and the second for Europeans? But the header says EUROPEANS first.

Wait, the header: "EUROPEANS. CHINESE OR EUROPEANS. COLOURED. CHINESE OK COLOURED." This is garbled. Probably the table has two main sections: EUROPEANS and CHINESE OR COLOURED. Each with Males, Females, Total.

But the OCR shows "EUROPEANS." then "CHINESE OR" then "EUROPEANS." then "COLOURED." then "CHINESE OK COLOURED." This might be a two-page table or two columns with repeated headers.

Given the numbers are around 1,000-1,200 for males and 40-100 for females, totals 1,060-1,280. This is plausible for total prisoners (mostly Chinese). The European prisoners would be much fewer. So perhaps the first numeric column after the date is Chinese Males, not European.

But the header says "EUROPEANS. Males. Fe- males. I Males. Fe- males." Then "CHINESE OR COLOURED. Males. Fe- males. Males. Fe- males." This is confusing.

Let me look at the data for January 5: "1,017 42 1,063". Then July 6: "13 : 1,149 98 1.251". The "13" might be European males for July? 13 European males, then Chinese males 1,149, females 98, total 1,251. That makes sense! For January 5, European males might be missing or the "1,017" is actually Chinese males? But the pattern: Jan 5 line has "1,017 42 1,063" - three numbers. Jul 6 line has "13 : 1,149 98 1.251" - four numbers? "13" then "1,149" "98" "1,251". Actually "13" might be European males, then Chinese males 1,149, Chinese females 98, Chinese total 1,251. But where is European females and total? Maybe European females is 0 or not shown.

Let's check the next line: "12" then "1,039" then "} 41 1,088". "12" is Jan 12. "1,039" might be European males? No, 1,039 Europeans unlikely. Perhaps the first number after date is Chinese males, then Chinese females, then Chinese total. And European numbers are in a separate column not captured well.

But the header suggests six numeric columns: Euro M, Euro F, Euro Total, Chinese M, Chinese F, Chinese Total.

The OCR might have dropped the European columns because they were small numbers (single/double digits) and misread.

Let's look at the line for Jan 5: "January 5 1927. : 1,017 42 1,063". The ":" might be a separator. Then three numbers. Then "July 6 13 : 1,149 98 1.251". Here "13" before colon. So perhaps the format is: Date, European Males, European Females, European Total, Chinese Males, Chinese Females, Chinese Total. But for Jan 5, European numbers are missing? Or "1,017" is European Males? No.

Wait, the OCR shows "January 5" then "1927." then ":" then "1,017 42 1,063". Then "July 6" then "13" then ":" then "1,149 98 1.251". The "13" is on a separate line. In the raw lines, line 55: "July", line 56: "6", line 57: "13", line 58: ":", line 59: "1,149", line 60: "98 1.251". So "13" is a separate token.

Similarly, for Jan 12: line 67: "12", line 68: "1,039", line 69: "}", line 70: "41 1,088". So "12" date, "1,039" first number, "41 1,088" next.

For Jan 19: line 86: "19", line 87: "\"", line 88: "1,054", line 89: "44", line 90: "| 1,105". So "19", "1,054", "44", "1,105".

For Jan 26: line 104: "26", line 105: "1,069", line 106: "45 1,120". So "26", "1,069", "45", "1,120".

For Feb 2: line 120: "February 2", line 121: "1,052", line 122: "14", line 123: "1,101". So "Feb 2", "1,052", "14", "1,101".

For Feb 9: line 139: "9", line 140: "1,014", line 141: "46", line 142: "1,066". So "9", "1,014", "46", "1,066".

For Feb 16: line 156: "16", line 157: "那些" (garbled), line 158: "5", line 159: "1,036", line 160: "49", line 161: "\"\"", line 162: "1,090". So "16", "5"? Wait "那些" might be "1" or something. "5" might be European males? Then "1,036" Chinese males, "49" Chinese females, "1,090" Chinese total.

For Feb 23: line 181: "23", line 182: "3", line 183: ":", line 184: "1,040", line 185: "42 1,085". So "23", "3", "1,040", "42", "1,085". Here "3" might be European males.

For Mar 2: line 203: "March", line 204: "2", line 205: "9", line 206: "1,064", line 207: "那些", line 208: "1,124". So "Mar 2", "9", "1,064", "1,124". "9" European males? "1,064" Chinese males? "1,124" Chinese total? Missing Chinese females.

For Mar 9: line 229: "9", line 230: "1,126", line 231: "48", line 232: "1,182". So "9", "1,126", "48", "1,182". No European number shown.

For Mar 16: line 215: "31"? Wait line 215 is "31", line 216: "6", line 217: "??", line 218: "1,119", line 219: "15", line 220: "55", line 221: "1,190". This is messy. "31" might be Mar 31? But Mar 31 is week 13. Let's list weeks chronologically.

I think the best approach is to recognize that the OCR has completely mangled the table. Given the instruction to "reconstruct tabular data" using Markdown table syntax, and "restore column reading order", I should attempt to reconstruct the table as it logically should be: a chronological list of weeks with six numeric columns.

But the OCR data is too corrupted to reliably extract all 52 weeks * 6 columns. Many numbers are missing, garbled, or misaligned.

However, the user expects me to proofread the OCR output. The output should be the proofread text in Markdown. Since the original is a table, I should output a Markdown table.

Given the severe corruption, I might need to indicate missing data with ... as per rule 7.

But perhaps I can salvage by separating the two columns. The OCR interleaves left and right column lines. Let me try to split the text into two streams: odd lines (left column) and even lines (right column)? Not exactly, because the line breaks are arbitrary.

Another approach: The text appears to have two distinct date sequences interleaved. I can extract all date-like tokens and numbers and try to assign them to weeks.

But this is extremely time-consuming and error-prone. Given the constraints, perhaps the expected output is a cleaned-up version of the OCR text with paragraphs fixed, but for tables, reconstruct as best as possible.

The instruction says: "Tables — if data is clearly in table format, reconstruct the table using Markdown table syntax."

So I must produce a Markdown table.

Let me try to parse the data by assuming the original table had two columns side by side, each with 26 rows (weeks 1-26 and 27-52). The OCR read it row by row across: row1_col1, row1_col2, row2_col1, row2_col2, etc.

But the OCR output is not line-aligned; it's a stream of tokens. However, the tokens appear in groups.

Let me write a script mentally to parse. But as an AI, I can process the text sequentially and try to identify patterns.

I'll go through the raw lines and group by week.

First, note that the header lines (1-42) are just header. The data starts at line 43 "1927." then "January 5".

Let me list all tokens in order, ignoring empty lines.

Tokens:

  1. (Y 11 )
  2. PRISON.
  3. 1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927.
  4. EUROPEANS.
  5. CHINESE OR
  6. EUROPEANS.
  7. COLOURED.
  8. CHINESE OK COLOURED.
  9. WEEK.
  10. TOTAL.
  11. WEEK.
  12. IOTAL.
  13. Males.
  14. Fe- males.
  15. I
  16. Males.
  17. Fe- males.
  18. Males.
  19. Fe- males.
  20. Males.
  21. Fe- males.
  22. 1927.
  23. January 5
  24. 1927.
  25. :
  26. 1,017
  27. 42 1,063
  28. July
  29. 6
  30. 13
  31. :
  32. 1,149
  33. 98 1.251
  34. 12
  35. 1,039
  36. }
  37. 41 1,088
  38. 13
  39. 5
  40. 1.135
  41. 88
  42. "}
  43. 1.228
  44. 19
  45. "
  46. 1,054
  47. 44
  48. | 1,105
  49. 20
  50. 1.144
  51. 94
  52. "
  53. 1.243
  54. 26
  55. 1,069
  56. 45 1,120
  57. 27
  58. 像像
  59. 1,144
  60. 85
  61. 1,224
  62. February 2
  63. 1,052
  64. 14
  65. 1,101
  66. August
  67. *
  68. 5
  69. 1,154
  70. 69
  71. 1,228
  72. 9
  73. 1,014
  74. 46
  75. 1,066
  76. 10
  77. "
  78. 1,123
  79. 61
  80. 1,190
  81. 16
  82. 那些
  83. 5
  84. 1,036
  85. 49
  86. ""
  87. 1,090
  88. 17
  89. 6
  90. 17
  91. 1,115
  92. 57
  93. 1,181
  94. 23
  95. 3
  96. :
  97. 1,040
  98. 42 1,085
  99. 24
  100. 6 ¦
  101. 11.08
  102. 57
  103. 那些
  104. 1,171
  105. March
  106. 2
  107. 9
  108. 1,064
  109. 那些
  110. 1,124
  111. 31
  112. 6
  113. ??
  114. 1,119
  115. 15
  116. 55
  117. 1,190
  118. 9
  119. 1,126
  120. 48
  121. 1,182
  122. September 7
  123. 9
  124. 1.138
  125. 51
  126. 16
  127. 9
  128. 27
  129. 1,092
  130. 51
  131. 1,155
  132. 14
  133. 11
  134. 23
  135. 19
  136. 1,082
  137. 59
  138. 1,148
  139. 30
  140. 11
  141. 1,119
  142. 62
  143. 1.193
  144. 12 12
  145. 24
  146. "
  147. April
  148. :
  149. :
  150. 1,198
  151. 1,141
  152. 55
  153. 1,205
  154. 1,164 56
  155. 1,227
  156. 1,212
  157. 59
  158. 1,278
  159. co
  160. 6
  161. 7
  162. 1,125
  163. 61
  164. 1,193 October
  165. 7
  166. 1,212 | 54
  167. 1,273
  168. 13
  169. 7
  170. **
  171. 1,106
  172. 72
  173. 1,185
  174. 12
  175. 10
  176. 1,203
  177. 59
  178. 1,272
  179. "
  180. 20
  181. 27
  182. 20
  183. +
  184. 1,111
  185. 75
  186. 1,195
  187. 19
  188. LI
  189. "
  190. 1,192
  191. 61
  192. 1,254
  193. **
  194. :
  195. 1,141
  196. 77
  197. 1.225
  198. 26
  199. 9
  200. 1,153
  201. T
  202. 56
  203. 1.220
  204. May
  205. Go
  206. 3
  207. 5
  208. 1,123
  209. 2
  210. 1,206
  211. November 2
  212. :
  213. 1,161
  214. 48
  215. 1,210
  216. 10
  217. 6
  218. "
  219. 1,106
  220. 77
  221. 1,189
  222. 9
  223. 1 |
  224. 1,173
  225. 47
  226. 1,221
  227. 17
  228. 7
  229. **
  230. :
  231. 1,105
  232. 77
  233. 1,189
  234. 15
  235. 1,149
  236. 19
  237. 1,199
  238. | 24
  239. 6
  240. 13
  241. 1,125
  242. 81
  243. 1,212
  244. 23
  245. 1.166
  246. 52
  247. "
  248. 1,229
  249. June
  250. 1
  251. 7
  252. ...
  253. 1,196
  254. 3
  255. NO
  256. 1,283
  257. 30
  258. "
  259. 1,146 | 43
  260. 1,202
  261. x
  262. 7
  263. 1,170
  264. 82
  265. 1,259
  266. December 7
  267. 6
  268. 1,146
  269. 7
  270. 51
  271. 1,205
  272. 15
  273. B
  274. C
  275. 1.162
  276. 那些
  277. 1,251
  278. 14
  279. 10
  280. H
  281. 1,142
  282. 57
  283. 1,209
  284. 22
  285. 7
  286. #
  287. 1,144
  288. 75
  289. 1,229
  290. 21
  291. 10
  292. 1,148 61
  293. 1,219
  294. 29
  295. *
  296. 6 |
  297. 1,147
  298. 102
  299. 1.255
  300. 28
  301. 1,112
  302. 53
  303. 1.174
  304. 395

Now, I need to pair these into weeks. The pattern seems to be: for each week, there is a date, then some numbers for Europeans, then numbers for Chinese. But the Europeans numbers are often missing or single digit.

Observing the first few:

  • Token 23: "January 5" (date)
  • Token 24: "1927." (year)
  • Token 25: ":" (separator)
  • Token 26: "1,017" (number)
  • Token 27: "42 1,063" (two numbers: 42 and 1,063)
  • Token 28: "July" (next date month)
  • Token 29: "6" (day)
  • Token 30: "13" (number)
  • Token 31: ":" (separator)
  • Token 32: "1,149" (number)
  • Token 33: "98 1.251" (two numbers)

So for Jan 5, we have three numbers: 1017, 42, 1063. For Jul 6, we have 13, 1149, 98, 1251? Actually four numbers: 13, 1149, 98, 1251.

But the header suggests 6 numbers per week. Perhaps the table has two sections: left column for Europeans, right for Chinese? No.

Maybe the table is structured as:

Week | European Males | European Females | European Total | Chinese Males | Chinese Females | Chinese Total

But the data shows only 3 or 4 numbers per week in the OCR.

Let's look at the next week: Token 34: "12" (Jan 12), Token 35: "1,039", Token 36: "}", Token 37: "41 1,088". So numbers: 1039, 41, 1088. Three numbers.

Token 38: "13" (Jul 13?), Token 39: "5", Token 40: "1.135", Token 41: "88", Token 42: "\"}", Token 43: "1.228". So for Jul 13: 5, 1135, 88, 1228? Four numbers.

Token 44: "19" (Jan 19), Token 45: "\"", Token 46: "1,054", Token 47: "44", Token 48: "| 1,105". So 1054, 44, 1105.

Token 49: "20" (Jul 20?), Token 50: "1.144", Token 51: "94", Token 52: "\"", Token 53: "1.243". So 1144, 94, 1243? Three numbers.

Token 54: "26" (Jan 26), Token 55: "1,069", Token 56: "45 1,120". So 1069, 45, 1120.

Token 57: "27" (Jul 27), Token 58: "那些", Token 59: "1,144", Token 60: "85", Token 61: "1,224". So 1144, 85, 1224.

Token 62: "February 2", Token 63: "1,052", Token 64: "14", Token 65: "1,101". So 1052, 14, 1101.

Token 66: "August", Token 67: "", Token 68: "5", Token 69: "1,154", Token 70: "69", Token 71: "1,228". So Aug 5: 1154, 69, 1228? But token 67 is "", token 68 is "5" (day). So date Aug 5. Numbers: 1154, 69, 1228.

Token 72: "9" (Feb 9), Token 73: "1,014", Token 74: "46", Token 75: "1,066". So 1014, 46, 1066.

Token 76: "10" (Aug 10), Token 77: "\"", Token 78: "1,123", Token 79: "61", Token 80: "1,190". So 1123, 61, 1190.

Token 81: "16" (Feb 16), Token 82: "那些", Token 83: "5", Token 84: "1,036", Token 85: "49", Token 86: "\"\"", Token 87: "1,090". So numbers: 5, 1036, 49, 1090. Four numbers.

Token 88: "17" (Aug 17), Token 89: "6", Token 90: "17", Token 91: "1,115", Token 92: "57", Token 93: "1,181". So 6, 17, 1115, 57, 1181? Five numbers? "6" and "17" might be European males and females? Then 1115 Chinese males, 57 Chinese females, 1181 Chinese total.

Token 94: "23" (Feb 23), Token 95: "3", Token 96: ":", Token 97: "1,040", Token 98: "42 1,085". So 3, 1040, 42, 1085. Four numbers.

Token 99: "24" (Aug 24), Token 100: "6 ¦", Token 101: "11.08", Token 102: "57", Token 103: "那些", Token 104: "1,171". So 6, 1108? 57, 1171. Garbled.

Token 105: "March", Token 106: "2", Token 107: "9", Token 108: "1,064", Token 109: "那些", Token 110: "1,124". So Mar 2: 9, 1064, 1124? Three numbers.

Token 111: "31" (Sep 31? No, Sep 30), Token 112: "6", Token 113: "??", Token 114: "1,119", Token 115: "15", Token 116: "55", Token 117: "1,190". This might be Sep 7? But token 122 is "September 7". So token 111 "31" might be Mar 31? March has 31 days, week ending Mar 31. Token 111: "31", token 112: "6" (European males?), token 113: "??", token 114: "1,119" (Chinese males), token 115: "15" (Chinese females?), token 116: "55" (??), token 117: "1,190" (Chinese total). Messy.

Token 118: "9" (Mar 9), Token 119: "1,126", Token 120: "48", Token 121: "1,182". So 1126, 48, 1182.

Token 122: "September 7", Token 123: "9", Token 124: "1.138", Token 125: "51", Token 126: "16", Token 127: "9", Token 128: "27", Token 129: "1,092", Token 130: "51", Token 131: "1,155". This is interleaved: Sep 7 and Sep 16? Token 122: "September 7", token 123: "9" (European males?), token 124: "1.138" (Chinese males?), token 125: "51" (Chinese females?), token 126: "16" (next week date?), token 127: "9" (European males?), token 128: "27" (Chinese males?), token 129: "1,092" (Chinese males?), token 130: "51" (Chinese females?), token 131: "1,155" (Chinese total). Very messy.

Given the extreme corruption, it's impossible to accurately reconstruct the full table. However, the task is to proofread the OCR output. The output should be the corrected text in Markdown. Since the original is a table, I should output a Markdown table with the data as best as I can, but I must not invent data. Rule 7: if missing, insert ....

But the user provided the OCR text and expects a proofread version. Perhaps the proofread version should be a cleaned-up version of the text, preserving the table structure as much as possible, with OCR errors corrected (like removing garbled characters, fixing numbers).

Given the instructions: "Reconstruct Sentences and Paragraphs" but also "Tables — if data is clearly in table format, reconstruct the table using Markdown table syntax."

The text is primarily a table. So I should output a Markdown table.

I think the best I can do is to create a table with the weeks in chronological order, and for each week, the six columns, filling in what can be deciphered and using ... for missing.

But that would be a huge table (52 rows). And I don't have reliable data for all.

Alternatively, I could output the text as a series of paragraphs describing the table? No, the instruction says to use Markdown table syntax.

Given the complexity, perhaps the expected answer is to format the header and then the data rows as they appear in the OCR but cleaned up, acknowledging the two-column layout.

But the instruction: "restore column reading order — if text was originally in multiple columns but OCR read it in the wrong order, reorder it to make logical sense (left-to-right, top-to-bottom)."

The original was two columns side by side. Logical reading order for a human is to read left column top to bottom, then right column top to bottom. But the content is chronological, so left column is first half year, right column second half. So the logical order is chronological: left column rows 1-26, then right column rows 1-26.

The OCR read it row by row across: left1, right1, left2, right2, etc. So to restore, I need to separate the two columns and then concatenate left then right.

But the OCR output is a linear stream. I need to de-interleave.

From the tokens, it seems the first week (Jan 5) and the 27th week (Jul 6) are interleaved. Then Jan 12 and Jul 13, etc.

So the stream is: Week1_left, Week1_right, Week2_left, Week2_right, ...

Thus, I can split the data tokens into two sequences: even-indexed weeks (left) and odd-indexed weeks (right) after the header.

But the tokens are not grouped per week cleanly.

Given the time, I might produce a simplified output: a Markdown table with the header and then note that the data is corrupted. But the instruction says to proofread, not to comment.

I'll attempt to reconstruct a table for the first few weeks to demonstrate, but the user likely expects the full table.

Wait, the OCR text ends with "395". That might be a page number or total.

Given the difficulty, I'll produce a Markdown table with the columns as per header, and fill rows for each week by extracting from the token stream in chronological order, using the left-column dates (Jan-Jun) then right-column dates (Jul-Dec). I'll parse the left-column data from the tokens that correspond to Jan-Jun dates, and right-column from Jul-Dec dates.

Let me identify all date tokens and their following numbers.

Left column dates (Jan-Jun):

  • January 5
  • January 12
  • January 19
  • January 26
  • February 2
  • February 9
  • February 16
  • February 23
  • March 2
  • March 9
  • March 16? (not seen, but March 23? Token 111 is 31, maybe March 30)
  • March 23? Token 111 "31" could be March 31.
  • April 6? Token 148 "April", token 161 "6", token 162 "7" - maybe April 6 and 13?
  • April 13? Token 161 "6", 162 "7" - confusing.
  • April 20? Token 181 "20", 182 "27", 183 "20" - messy.
  • May 4? Token 206 "May", 207 "Go", 208 "3", 209 "5" - maybe May 4? "Go" might be "4" misread.
  • May 11? Token 218 "10", 219 "6" - maybe May 11?
  • May 18? Token 229 "17", 230 "上", 231 "7" - maybe May 18?
  • May 25? Token 241 "| 24", 242 "6", 243 "13" - maybe May 25?
  • June 1? Token 252 "June", 253 "1", 254 "7" - June 1 and 8?
  • June 8? Token 254 "7"?
  • June 15? Token 264 "x", 265 "7" - maybe June 15?
  • June 22? Token 260 "30", 261 "\"" - June 30?
  • June 29? Not clear.

Right column dates (Jul-Dec):

  • July 6
  • July 13
  • July 20
  • July 27
  • August 5
  • August 10
  • August 17
  • August 24
  • August 31? Token 99 "24", then maybe 31 not seen.
  • September 7
  • September 14? Token 126 "16" but that's 16.
  • September 21? Token 136 "19"?
  • September 28? Token 140 "30"?
  • October 7? Token 165 "October", 166 "7"
  • October 12? Token 169 "13", 170 "7" - maybe Oct 12?
  • October 19? Token 181 "20", 182 "27" - Oct 20?
  • October 26? Token 199 "26"?
  • November 2? Token 213 "November 2"
  • November 9? Token 218 "10", 219 "6" - Nov 9?
  • November 16? Token 229 "17" - Nov 16?
  • November 23? Token 241 "| 24", 242 "6" - Nov 23?
  • November 30? Token 260 "30" - Nov 30?
  • December 7? Token 269 "December 7"
  • December 14? Token 281 "14"
  • December 21? Token 287 "22", 288 "7" - Dec 21?
  • December 28? Token 297 "29", 298 "*", 299 "6" - Dec 28?
  • December 31? Token 303 "28", 304 "1,112" - maybe Dec 28.

This is too messy.

Given the constraints, I think the best approach is to output the proofread text as a cleaned-up version of the OCR text, preserving the table structure by using Markdown table for the header and then representing the data as a list of weeks with numbers, but given the corruption, I'll produce a table with the header and then a note that data rows are corrupted? But the instruction says not to add commentary.

Perhaps I can reconstruct the table by assuming the OCR tokens for each week are in groups of 7: date, euro_m, euro_f, euro_t, chi_m, chi_f, chi_t. But the data doesn't match.

Another idea: The original image might be a standard Hong Kong Blue Book statistical table. I could look up the typical format. But I don't have access.

Given the time, I'll produce a Markdown table with the header and then for each week, I'll put the date and the numbers as they appear in the OCR for that week in the left column (first half) and right column (second half) but in chronological order. I'll do my best to parse the first few and last few.

But the user might expect the full table. However, as an AI, I can process the entire token stream and attempt to assign.

Let me write a programmatic approach in my reasoning.

I'll simulate parsing: The tokens after header (token 22 onward) come in pairs of weeks: left week, right week.

Define a function to parse a week from tokens starting at index. But variable length.

Given the difficulty, I'll output a simplified Markdown table with the columns and then rows for each week with the data I can extract, using ... for missing.

But the instruction: "Do not add or remove substantive words. Do not rephrase, rewrite, or change the author's style, tone, or word choice. Only correct what is clearly wrong."

So I should not create a new table with guessed data. I should correct the OCR text as a text, preserving paragraphs. The text is a table, so I should correct the table representation.

Perhaps the best is to output the text as a Markdown table with the exact same rows as the OCR but with OCR errors fixed (garbled characters removed, numbers corrected). But the OCR text is not in row format; it's a stream.

I think I need to interpret the OCR output as a two-column table and render it as two separate tables or one combined.

Given the instruction "restore column reading order", I will separate the two columns and present them as two tables: one for Jan-Jun, one for Jul-Dec, each with the proper columns.

But the OCR doesn't clearly separate.

Let me try to split the token stream after the header into two streams: Stream A (left column) and Stream B (right column). The pattern: the first week of left (Jan 5), then first week of right (Jul 6), then second week of left (Jan 12), then second week of right (Jul 13), etc.

So if I take every other week-group, I can separate.

But the week-groups are not delimited.

However, the dates are recognizable. I can scan for month names.

Month tokens: "January", "July", "February", "August", "March", "September", "April", "October", "May", "November", "June", "December".

In the token stream, they appear in order: January (token 23), July (28), February (62), August (66), March (105), September (122), April (148), October (165), May (206), November (213), June (252), December (269).

This is exactly the interleaved order: Jan, Jul, Feb, Aug, Mar, Sep, Apr, Oct, May, Nov, Jun, Dec.

So the left column months: Jan, Feb, Mar, Apr, May, Jun.

Right column months: Jul, Aug, Sep, Oct, Nov, Dec.

Each month has about 4-5 weeks.

So the token stream after header is grouped by month pairs.

For each month pair, the weeks are interleaved.

For example, January weeks: Jan 5, 12, 19, 26. July weeks: Jul 6, 13, 20, 27.

In tokens: after "January 5" (23), we have numbers, then "July" (28), "6" (29), numbers, then "12" (34) (Jan 12), numbers, then "13" (38) (Jul 13), numbers, then "19" (44) (Jan 19), numbers, then "20" (49) (Jul 20), numbers, then "26" (54) (Jan 26), numbers, then "27" (57) (Jul 27), numbers.

Then "February 2" (62), numbers, then "August" (66), "5" (68), numbers, then "9" (72) (Feb 9), numbers, then "10" (76) (Aug 10), numbers, then "16" (81) (Feb 16), numbers, then "17" (88) (Aug 17), numbers, then "23" (94) (Feb 23), numbers, then "24" (99) (Aug 24), numbers.

Then "March" (105), "2" (106), numbers, then "31" (111) (Mar 31?), but September appears at 122. Actually after March, we have "September 7" (122). But March has weeks: 2, 9, 16, 23, 30. September: 7, 14, 21, 28.

In tokens: after "March" (105), "2" (106), then numbers, then "9" (118) (Mar 9), numbers, then "31" (111) appears before "9"? Token 111 is "31", token 112 "6", etc. Then token 118 "9". Then token 122 "September 7". So March 31 (or 30) appears before March 9? That's out of order. Token 111 "31" might be March 31, but it appears after March 2 and before March 9? Actually token 105 March, 106 2, 107 9, 108 1064, 109 those, 110 1124, 111 31, 112 6, 113 ??, 114 1119, 115 15, 116 55, 117 1190, 118 9, 119 1126, 120 48, 121 1182, 122 September 7.

So the March weeks are interleaved with September? But September starts at 122. The tokens 111-117 might be for September 7? But "31" is not September. September 7 is token 122.

This is confusing.

Given the time, I will produce a proofread version that presents the header clearly and then the data in a simplified table format, acknowledging the corruption by using ... for unreadable parts. But the instruction says not to add commentary.

I think the safest is to output the text as a Markdown table with the header row and then each subsequent line as a row, but since the OCR is not row-aligned, I'll try to reconstruct rows by grouping tokens between dates.

Given the extreme difficulty, I'll provide a cleaned-up version of the OCR text in a code block? But the instruction says not to wrap in code fences.

The instruction: "Return the proofread text in standard Markdown: ... Do not wrap your response in code fences."

So I must output plain Markdown.

I'll output a Markdown table with the columns: Week, European Males, European Females, European Total, Chinese Males, Chinese Females, Chinese Total.

Then I'll list the weeks in chronological order with data extracted as best as I can from the left-column tokens (Jan-Jun) and right-column tokens (Jul-Dec). I'll do this for all 52 weeks, but given the token limit, I'll do a representative sample? No, the user gave the full OCR, so they expect full proofread.

But the output token limit might be large. I'll try to produce the full table.

Let me attempt to parse the left column (Jan-Jun) from the tokens that belong to Jan, Feb, Mar, Apr, May, Jun.

I'll go through the token stream and pick out the left-column weeks by their dates.

Left column dates and following numbers (until next left date or right date).

I'll write a parser in my mind.

Start at token 23: "January 5" -> left week 1.

Tokens until next left date "12" (token 34) or right date "July" (28). But right date appears at 28. So left week 1 tokens: 23-27? Tokens: 23 Jan5, 24 1927., 25 :, 26 1017, 27 42 1063. Then token 28 July (right). So left week 1 data: 1017, 42, 1063. But we need 6 numbers. Only 3. Perhaps the European numbers are not present or are zero. Maybe the table only shows Chinese? But header says both.

Token 27 is "42 1,063" which is two numbers. So three numbers total.

Similarly, right week 1: token 28 July, 29 6, 30 13, 31 :, 32 1149, 33 98 1.251. So numbers: 13, 1149, 98, 1251. Four numbers.

Left week 2: token 34 "12", 35 1039, 36 }, 37 41 1088. So 1039, 41, 1088.

Right week 2: token 38 "13", 39 5, 40 1.135, 41 88, 42 "}", 43 1.228. So 5, 1135, 88, 1228.

Left week 3: token 44 "19", 45 ", 46 1054, 47 44, 48 | 1105. So 1054, 44, 1105.

Right week 3: token 49 "20", 50 1.144, 51 94, 52 ", 53 1.243. So 1144, 94, 1243.

Left week 4: token 54 "26", 55 1069, 56 45 1120. So 1069, 45, 1120.

Right week 4: token 57 "27", 58 those, 59 1144, 60 85, 61 1224. So 1144, 85, 1224.

Left week 5: token 62 "February 2", 63 1052, 64 14, 65 1101. So 1052, 14, 1101.

Right week 5: token 66 "August", 67 *, 68 5, 69 1154, 70 69, 71 1228. So 1154, 69, 1228. (Date Aug 5)

Left week 6: token 72 "9", 73 1014, 74 46, 75 1066. (Feb 9) So 1014, 46, 1066.

Right week 6: token 76 "10", 77 ", 78 1123, 79 61, 80 1190. (Aug 10) So 1123, 61, 1190.

Left week 7: token 81 "16", 82 those, 83 5, 84 1036, 85 49, 86 "", 87 1090. (Feb 16) Numbers: 5, 1036, 49, 1090. Four numbers.

Right week 7: token 88 "17", 89 6, 90 17, 91 1115, 92 57, 93 1181. (Aug 17) Numbers: 6, 17, 1115, 57, 1181. Five numbers.

Left week 8: token 94 "23", 95 3, 96 :, 97 1040, 98 42 1085. (Feb 23) Numbers: 3, 1040, 42, 1085. Four numbers.

Right week 8: token 99 "24", 100 6 ¦, 101 11.08, 102 57, 103 those, 104 1171. (Aug 24) Numbers: 6, 1108, 57, 1171. (11.08 -> 1108)

Left week 9: token 105 "March", 106 "2", 107 9, 108 1064, 109 those, 110 1124. (Mar 2) Numbers: 9, 1064, 1124. Three numbers.

Right week 9: token 122 "September 7", 123 9, 124 1.138, 125 51, 126 16, 127 9, 128 27, 129 1092, 130 51, 131 1155. This is messy because it includes next week? Token 126 "16" might be Sep 16 (next week). So right week 9 (Sep 7): tokens 122-125? 9, 1138,

Edit History

2026-07-14 03:31:10 · NVIDIA / nvidia/nemotron-3-ultra-550b-a55b
Live
View comparison
AI Proofread
The user wants me to proofread OCR output of a historical prison record from Hong Kong (1927). The text appears to be a table showing weekly prisoner counts broken down by Europeans and Chinese/Coloured, with males and females. Let me analyze the OCR text carefully. It's a table with columns: WEEK (date) EUROPEANS: Males, Females, Total CHINESE OR COLOURED: Males, Females, Total The OCR has many errors: garbled characters (喃喃, นค, པ, ་, 上, etc.), misaligned columns, broken numbers, and formatting issues. I need to reconstruct this as a proper Markdown table. Let me parse the data week by week. First, let me understand the structure. The header shows: EUROPEANS: Males, Females, Total CHINESE OR COLOURED: Males, Females, Total But the OCR shows "EUROPEANS." then "CHINESE OR" then "EUROPEANS." then "COLOURED." then "CHINESE OK COLOURED." - this is messy. Looking at the data rows, each week seems to have two lines: one for Europeans, one for Chinese/Coloured? Or maybe it's side-by-side columns. Actually, looking at the first data row: "January 5 1927. : 1,017 42 1,063 July 6 13 : 1,149 98 1.251" This seems like two separate weeks side by side? January 5 and July 6? That doesn't make sense chronologically. Wait, the OCR might have read two columns of the table side by side. The original table probably had two columns: left side for first half of year, right side for second half? Or maybe it's a wide table with European and Chinese sections. Let me look at the pattern. The header says: "1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927." Then columns: EUROPEANS (Males, Females, Total), CHINESE OR COLOURED (Males, Females, Total) But the data shows dates like "January 5", "12", "19", "26", "February 2", "9", "16", "23", "March 2", "9", "16", "23", "30", "April 6", "13", "20", "27", "May 4", "11", "18", "25", "June 1", "8", "15", "22", "29", "July 6", "13", "20", "27", "August 3", "10", "17", "24", "31", "September 7", "14", "21", "28", "October 5", "12", "19", "26", "November 2", "9", "16", "23", "30", "December 7", "14", "21", "28" That's 52 weeks (one per week). The OCR seems to have interleaved them or read in two columns. Looking at the raw text, it appears the OCR read the table in two vertical columns (left and right) but merged them line by line. For example, first line has "January 5" and "July 6" - those are week 1 and week 27 approximately. Let me try to separate the two columns. The original table was likely printed in two columns on the page to save space: left column Jan-Jun, right column Jul-Dec. So I need to split the data into two halves and then interleave them chronologically? Or present as two separate tables? The instruction says "restore column reading order — if text was originally in multiple columns but OCR read it in the wrong order, reorder it to make logical sense (left-to-right, top-to-bottom)." Since the original was two columns side by side (left column first half, right column second half), the logical reading order is to read the left column top to bottom, then right column top to bottom. But for a weekly time series, chronological order makes more sense. However, the instruction says "reorder it to make logical sense (left-to-right, top-to-bottom)" - meaning the original physical layout was left column then right column. But the content is chronological weeks. The most logical presentation for the data is chronological order. I think I should reconstruct the table in chronological order (Week 1 to Week 52) with proper columns. Let me parse all the data. First, let me extract all the week entries from the OCR text. The text is a mess with many OCR artifacts. I'll go line by line. The OCR text lines (as provided): Line 1: (Y 11 ) Line 2: Line 3: PRISON. Line 4: Line 5: 1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927. Line 6: Line 7: EUROPEANS. Line 8: Line 9: CHINESE OR Line 10: Line 11: EUROPEANS. Line 12: Line 13: COLOURED. Line 14: Line 15: CHINESE OK COLOURED. Line 16: Line 17: WEEK. Line 18: Line 19: TOTAL. Line 20: Line 21: WEEK. Line 22: Line 23: IOTAL. Line 24: Line 25: Males. Line 26: Line 27: Fe- males. Line 28: Line 29: I Line 30: Line 31: Males. Line 32: Line 33: Fe- males. Line 34: Line 35: Males. Line 36: Line 37: Fe- males. Line 38: Line 39: Males. Line 40: Line 41: Fe- males. Line 42: Line 43: 1927. Line 44: Line 45: January 5 Line 46: Line 47: 1927. Line 48: Line 49: : Line 50: Line 51: 1,017 Line 52: Line 53: 42 1,063 Line 54: Line 55: July Line 56: Line 57: 6 Line 58: Line 59: 13 Line 60: Line 61: : Line 62: Line 63: 1,149 Line 64: Line 65: 98 1.251 Line 66: Line 67: 12 Line 68: Line 69: 1,039 Line 70: Line 71: } Line 72: Line 73: 41 1,088 Line 73: Line 74: 13 Line 75: Line 76: 5 Line 77: Line 78: 1.135 Line 79: Line 80: 88 Line 81: Line 82: "} Line 83: Line 84: 1.228 Line 85: Line 86: 19 Line 87: Line 88: " Line 89: 1,054 Line 90: Line 91: 44 Line 92: Line 93: | 1,105 Line 94: Line 95: 20 Line 96: Line 97: 1.144 Line 98: Line 99: 94 Line 100: Line 101: " Line 102: 1.243 Line 103: Line 104: 26 Line 105: Line 106: 1,069 Line 107: Line 108: 45 1,120 Line 109: Line 110: 27 Line 111: Line 112: 像像 Line 113: Line 114: 1,144 Line 115: Line 116: 85 Line 117: Line 118: 1,224 Line 119: Line 120: February 2 Line 121: Line 122: 1,052 Line 123: Line 124: 14 Line 125: Line 126: 1,101 Line 127: Line 128: August Line 129: Line 130: * Line 131: 5 Line 132: Line 133: 1,154 Line 134: Line 135: 69 Line 136: Line 137: 1,228 Line 138: Line 139: 9 Line 140: Line 141: 1,014 Line 142: Line 143: 46 Line 144: Line 145: 1,066 Line 146: Line 147: 10 Line 148: Line 149: " Line 150: 1,123 Line 151: Line 152: 61 Line 153: Line 154: 1,190 Line 155: Line 156: 16 Line 157: Line 158: นค Line 159: Line 160: 5 Line 161: Line 162: 1,036 Line 163: Line 164: 49 Line 165: Line 166: "" Line 167: Line 168: 1,090 Line 169: Line 170: 17 Line 171: Line 172: 6 Line 172: Line 173: 17 Line 174: Line 175: 1,115 Line 176: Line 177: 57 Line 178: Line 179: 1,181 Line 180: Line 181: 23 Line 182: Line 183: 3 Line 184: Line 185: : Line 186: Line 187: 1,040 Line 188: Line 189: 42 1,085 Line 190: Line 191: 24 Line 192: Line 193: 6 ¦ Line 194: Line 195: 11.08 Line 196: Line 197: 57 Line 198: Line 199: པ Line 200: Line 201: 1,171 Line 202: Line 203: March Line 204: Line 205: 2 Line 206: Line 207: 9 Line 208: Line 209: 1,064 Line 210: Line 211: 那些 Line 212: Line 213: 1,124 Line 214: Line 215: 31 Line 216: Line 217: 6 Line 218: Line 219: ?? Line 220: Line 221: 1,119 Line 222: Line 223: 15 Line 224: Line 225: 55 Line 226: Line 227: 1,190 Line 228: Line 229: 9 Line 229: Line 230: 1,126 Line 231: Line 232: 48 Line 233: Line 234: 1,182 Line 235: Line 236: September 7 Line 237: Line 238: 9 Line 239: 1.138 Line 240: Line 241: 51 Line 242: Line 243: 16 Line 244: Line 245: 9 Line 246: Line 247: 27 Line 248: Line 249: 1,092 Line 250: Line 251: 51 Line 252: Line 253: 1,155 Line 254: Line 255: 14 Line 256: Line 257: 11 Line 258: Line 259: 23 Line 260: Line 261: ་ Line 262: Line 263: 19 Line 264: Line 265: 1,082 Line 266: Line 267: 59 Line 268: Line 269: 1,148 Line 270: Line 271: 30 Line 272: Line 273: 11 Line 274: Line 275: 1,119 Line 276: Line 277: 62 Line 278: Line 279: 1.193 Line 280: Line 281: 12 12 Line 282: Line 283: 24 Line 283: Line 284: " Line 285: Line 286: April Line 287: Line 288: : Line 289: : Line 290: 1,198 Line 291: Line 292: 1,141 Line 293: Line 294: 55 Line 295: Line 296: 1,205 Line 297: Line 298: 1,164 56 Line 299: Line 300: 1,227 Line 301: Line 302: 1,212 Line 303: Line 304: 59 Line 305: Line 306: 1,278 Line 307: Line 308: co Line 309: Line 310: 6 Line 311: Line 312: 7 Line 313: Line 314: 1,125 Line 315: Line 316: 61 Line 317: Line 318: 1,193 October Line 319: Line 320: 7 Line 321: 1,212 | 54 Line 322: Line 323: 1,273 Line 324: Line 325: 13 Line 326: Line 327: 7 Line 328: Line 329: ** Line 330: Line 331: 1,106 Line 332: Line 333: 72 Line 334: Line 335: 1,185 Line 336: Line 337: 12 Line 338: Line 339: 10 Line 340: Line 341: 1,203 Line 342: Line 343: 59 Line 344: Line 345: 1,272 Line 346: Line 347: " Line 348: Line 349: 20 Line 350: Line 351: 27 Line 352: Line 353: 20 Line 354: Line 355: + Line 356: Line 357: 1,111 Line 358: Line 359: 75 Line 360: Line 361: 1,195 Line 362: Line 363: 19 Line 364: Line 365: LI Line 366: Line 367: " Line 368: 1,192 Line 369: Line 370: 61 Line 371: Line 372: 1,254 Line 373: Line 374: ** Line 375: Line 376: : Line 377: Line 378: 1,141 Line 379: Line 380: 77 Line 381: Line 382: 1.225 Line 383: Line 384: 26 Line 385: Line 386: 9 Line 387: Line 388: 1,153 Line 389: Line 390: T Line 391: Line 392: 56 Line 393: Line 394: 1.220 Line 395: Line 396: • Line 397: Line 398: May Line 399: Line 400: Go Line 401: 3 Line 402: 5 Line 403: 1,123 Line 404: Line 405: 2 Line 406: Line 407: 1,206 Line 408: Line 409: November 2 Line 410: Line 411: : Line 412: 1,161 Line 413: Line 414: 48 Line 415: Line 416: 1,210 Line 417: Line 418: 10 Line 419: Line 420: 6 Line 421: Line 422: " Line 423: 1,106 Line 424: Line 425: 77 Line 426: Line 427: 1,189 Line 428: Line 429: 9 Line 430: Line 431: 1 | Line 432: Line 433: 1,173 Line 434: Line 435: 47 Line 436: Line 437: 1,221 Line 438: Line 439: 17 Line 439: Line 440: 上 Line 441: Line 442: 7 Line 443: Line 444: ** Line 445: Line 446: : Line 447: 1,105 Line 448: Line 449: 77 Line 450: Line 451: 1,189 Line 452: Line 453: 15 Line 454: Line 455: 1,149 Line 456: Line 457: 19 Line 458: Line 459: 1,199 Line 460: Line 461: | 24 Line 462: Line 463: 6 Line 464: Line 465: 13 Line 466: Line 467: 1,125 Line 468: Line 469: 81 Line 470: Line 471: 1,212 Line 472: Line 473: 23 Line 474: Line 475: 1.166 Line 476: Line 477: 52 Line 478: Line 479: " Line 480: Line 481: 1,229 Line 482: Line 483: June Line 484: Line 485: 1 Line 486: Line 487: 7 Line 488: Line 489: ... Line 489: Line 490: 1,196 Line 491: Line 492: 3 Line 493: Line 494: NO Line 495: Line 496: 1,283 Line 497: Line 498: 30 Line 499: Line 500: " Line 501: 1,146 | 43 Line 502: Line 503: 1,202 Line 504: Line 505: x Line 506: Line 507: 7 Line 508: Line 509: 1,170 Line 509: Line 510: 82 Line 511: Line 512: 1,259 Line 513: Line 514: December 7 Line 515: Line 516: 6 Line 517: 1,146 Line 518: Line 519: 7 Line 520: Line 521: 51 Line 522: Line 523: 1,205 Line 524: Line 525: 15 Line 526: Line 527: B Line 528: Line 529: C Line 530: Line 531: 1.162 Line 532: Line 533: 那些 Line 534: Line 535: 1,251 Line 536: Line 537: 14 Line 538: Line 539: 10 Line 539: Line 540: H Line 541: Line 542: 1,142 Line 543: Line 544: 57 Line 545: Line 546: 1,209 Line 547: Line 548: 22 Line 549: Line 550: 7 Line 551: Line 552: # Line 553: Line 554: 1,144 Line 555: Line 556: 75 Line 557: Line 558: 1,229 Line 559: Line 560: 21 Line 561: Line 562: 10 Line 563: Line 564: 1,148 61 Line 565: Line 566: 1,219 Line 567: Line 568: 29 Line 569: Line 570: * Line 571: Line 572: 6 | Line 573: Line 574: 1,147 Line 575: Line 576: 102 Line 577: Line 578: 1.255 Line 579: Line 580: 28 Line 581: Line 582: 1,112 Line 583: Line 584: 53 Line 585: Line 586: 1.174 Line 587: Line 588: 395 This is extremely messy. The OCR has read a two-column table as a single stream, interleaving left and right columns line by line. The original table likely had two columns on the page: left column for weeks 1-26 (Jan-Jun), right column for weeks 27-52 (Jul-Dec). Each column has the same structure: Week date, European Males, European Females, European Total, Chinese Males, Chinese Females, Chinese Total. But the OCR read it row by row across both columns, so line 1 of left column, line 1 of right column, line 2 of left, line 2 of right, etc. Let me try to separate the two columns. First, identify the week dates in order. Left column (Jan-Jun) weeks: January 5, 12, 19, 26 February 2, 9, 16, 23 March 2, 9, 16, 23, 30 April 6, 13, 20, 27 May 4, 11, 18, 25 June 1, 8, 15, 22, 29 Right column (Jul-Dec) weeks: July 6, 13, 20, 27 August 3, 10, 17, 24, 31 September 7, 14, 21, 28 October 5, 12, 19, 26 November 2, 9, 16, 23, 30 December 7, 14, 21, 28 Now, the OCR text seems to have mixed these. Let me try to parse the data by matching patterns. Looking at the first data lines: "January 5" then "1927." then ":" then "1,017" then "42 1,063" then "July 6" then "13" then ":" then "1,149" then "98 1.251" This suggests the first row of left column is Jan 5, and first row of right column is Jul 6. But the numbers: For Jan 5: European Males 1,017? That seems high for Europeans. Wait, the header says EUROPEANS and CHINESE OR COLOURED. But the numbers are in thousands. That seems more like Chinese prisoners. Perhaps the columns are swapped? Or "EUROPEANS" might be a mislabel? Actually, in Hong Kong prisons, the majority were Chinese. But 1,000 Europeans seems too high. Let me check typical numbers. In 1927, Hong Kong population was ~800,000, Europeans maybe 10,000. Prisoners maybe hundreds. But 1,000 European prisoners? Unlikely. Perhaps the first set of numbers is for Chinese/Coloured, and the second for Europeans? But the header says EUROPEANS first. Wait, the header: "EUROPEANS. CHINESE OR EUROPEANS. COLOURED. CHINESE OK COLOURED." This is garbled. Probably the table has two main sections: EUROPEANS and CHINESE OR COLOURED. Each with Males, Females, Total. But the OCR shows "EUROPEANS." then "CHINESE OR" then "EUROPEANS." then "COLOURED." then "CHINESE OK COLOURED." This might be a two-page table or two columns with repeated headers. Given the numbers are around 1,000-1,200 for males and 40-100 for females, totals 1,060-1,280. This is plausible for total prisoners (mostly Chinese). The European prisoners would be much fewer. So perhaps the first numeric column after the date is Chinese Males, not European. But the header says "EUROPEANS. Males. Fe- males. I Males. Fe- males." Then "CHINESE OR COLOURED. Males. Fe- males. Males. Fe- males." This is confusing. Let me look at the data for January 5: "1,017 42 1,063". Then July 6: "13 : 1,149 98 1.251". The "13" might be European males for July? 13 European males, then Chinese males 1,149, females 98, total 1,251. That makes sense! For January 5, European males might be missing or the "1,017" is actually Chinese males? But the pattern: Jan 5 line has "1,017 42 1,063" - three numbers. Jul 6 line has "13 : 1,149 98 1.251" - four numbers? "13" then "1,149" "98" "1,251". Actually "13" might be European males, then Chinese males 1,149, Chinese females 98, Chinese total 1,251. But where is European females and total? Maybe European females is 0 or not shown. Let's check the next line: "12" then "1,039" then "} 41 1,088". "12" is Jan 12. "1,039" might be European males? No, 1,039 Europeans unlikely. Perhaps the first number after date is Chinese males, then Chinese females, then Chinese total. And European numbers are in a separate column not captured well. But the header suggests six numeric columns: Euro M, Euro F, Euro Total, Chinese M, Chinese F, Chinese Total. The OCR might have dropped the European columns because they were small numbers (single/double digits) and misread. Let's look at the line for Jan 5: "January 5 1927. : 1,017 42 1,063". The ":" might be a separator. Then three numbers. Then "July 6 13 : 1,149 98 1.251". Here "13" before colon. So perhaps the format is: Date, European Males, European Females, European Total, Chinese Males, Chinese Females, Chinese Total. But for Jan 5, European numbers are missing? Or "1,017" is European Males? No. Wait, the OCR shows "January 5" then "1927." then ":" then "1,017 42 1,063". Then "July 6" then "13" then ":" then "1,149 98 1.251". The "13" is on a separate line. In the raw lines, line 55: "July", line 56: "6", line 57: "13", line 58: ":", line 59: "1,149", line 60: "98 1.251". So "13" is a separate token. Similarly, for Jan 12: line 67: "12", line 68: "1,039", line 69: "}", line 70: "41 1,088". So "12" date, "1,039" first number, "41 1,088" next. For Jan 19: line 86: "19", line 87: "\"", line 88: "1,054", line 89: "44", line 90: "| 1,105". So "19", "1,054", "44", "1,105". For Jan 26: line 104: "26", line 105: "1,069", line 106: "45 1,120". So "26", "1,069", "45", "1,120". For Feb 2: line 120: "February 2", line 121: "1,052", line 122: "14", line 123: "1,101". So "Feb 2", "1,052", "14", "1,101". For Feb 9: line 139: "9", line 140: "1,014", line 141: "46", line 142: "1,066". So "9", "1,014", "46", "1,066". For Feb 16: line 156: "16", line 157: "那些" (garbled), line 158: "5", line 159: "1,036", line 160: "49", line 161: "\"\"", line 162: "1,090". So "16", "5"? Wait "那些" might be "1" or something. "5" might be European males? Then "1,036" Chinese males, "49" Chinese females, "1,090" Chinese total. For Feb 23: line 181: "23", line 182: "3", line 183: ":", line 184: "1,040", line 185: "42 1,085". So "23", "3", "1,040", "42", "1,085". Here "3" might be European males. For Mar 2: line 203: "March", line 204: "2", line 205: "9", line 206: "1,064", line 207: "那些", line 208: "1,124". So "Mar 2", "9", "1,064", "1,124". "9" European males? "1,064" Chinese males? "1,124" Chinese total? Missing Chinese females. For Mar 9: line 229: "9", line 230: "1,126", line 231: "48", line 232: "1,182". So "9", "1,126", "48", "1,182". No European number shown. For Mar 16: line 215: "31"? Wait line 215 is "31", line 216: "6", line 217: "??", line 218: "1,119", line 219: "15", line 220: "55", line 221: "1,190". This is messy. "31" might be Mar 31? But Mar 31 is week 13. Let's list weeks chronologically. I think the best approach is to recognize that the OCR has completely mangled the table. Given the instruction to "reconstruct tabular data" using Markdown table syntax, and "restore column reading order", I should attempt to reconstruct the table as it logically should be: a chronological list of weeks with six numeric columns. But the OCR data is too corrupted to reliably extract all 52 weeks * 6 columns. Many numbers are missing, garbled, or misaligned. However, the user expects me to proofread the OCR output. The output should be the proofread text in Markdown. Since the original is a table, I should output a Markdown table. Given the severe corruption, I might need to indicate missing data with ... as per rule 7. But perhaps I can salvage by separating the two columns. The OCR interleaves left and right column lines. Let me try to split the text into two streams: odd lines (left column) and even lines (right column)? Not exactly, because the line breaks are arbitrary. Another approach: The text appears to have two distinct date sequences interleaved. I can extract all date-like tokens and numbers and try to assign them to weeks. But this is extremely time-consuming and error-prone. Given the constraints, perhaps the expected output is a cleaned-up version of the OCR text with paragraphs fixed, but for tables, reconstruct as best as possible. The instruction says: "Tables — if data is clearly in table format, reconstruct the table using Markdown table syntax." So I must produce a Markdown table. Let me try to parse the data by assuming the original table had two columns side by side, each with 26 rows (weeks 1-26 and 27-52). The OCR read it row by row across: row1_col1, row1_col2, row2_col1, row2_col2, etc. But the OCR output is not line-aligned; it's a stream of tokens. However, the tokens appear in groups. Let me write a script mentally to parse. But as an AI, I can process the text sequentially and try to identify patterns. I'll go through the raw lines and group by week. First, note that the header lines (1-42) are just header. The data starts at line 43 "1927." then "January 5". Let me list all tokens in order, ignoring empty lines. Tokens: (Y 11 ) PRISON. 1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927. EUROPEANS. CHINESE OR EUROPEANS. COLOURED. CHINESE OK COLOURED. WEEK. TOTAL. WEEK. IOTAL. Males. Fe- males. I Males. Fe- males. Males. Fe- males. Males. Fe- males. 1927. January 5 1927. : 1,017 42 1,063 July 6 13 : 1,149 98 1.251 12 1,039 } 41 1,088 13 5 1.135 88 "} 1.228 19 " 1,054 44 | 1,105 20 1.144 94 " 1.243 26 1,069 45 1,120 27 像像 1,144 85 1,224 February 2 1,052 14 1,101 August * 5 1,154 69 1,228 9 1,014 46 1,066 10 " 1,123 61 1,190 16 那些 5 1,036 49 "" 1,090 17 6 17 1,115 57 1,181 23 3 : 1,040 42 1,085 24 6 ¦ 11.08 57 那些 1,171 March 2 9 1,064 那些 1,124 31 6 ?? 1,119 15 55 1,190 9 1,126 48 1,182 September 7 9 1.138 51 16 9 27 1,092 51 1,155 14 11 23 ་ 19 1,082 59 1,148 30 11 1,119 62 1.193 12 12 24 " April : : 1,198 1,141 55 1,205 1,164 56 1,227 1,212 59 1,278 co 6 7 1,125 61 1,193 October 7 1,212 | 54 1,273 13 7 ** 1,106 72 1,185 12 10 1,203 59 1,272 " 20 27 20 + 1,111 75 1,195 19 LI " 1,192 61 1,254 ** : 1,141 77 1.225 26 9 1,153 T 56 1.220 • May Go 3 5 1,123 2 1,206 November 2 : 1,161 48 1,210 10 6 " 1,106 77 1,189 9 1 | 1,173 47 1,221 17 上 7 ** : 1,105 77 1,189 15 1,149 19 1,199 | 24 6 13 1,125 81 1,212 23 1.166 52 " 1,229 June 1 7 ... 1,196 3 NO 1,283 30 " 1,146 | 43 1,202 x 7 1,170 82 1,259 December 7 6 1,146 7 51 1,205 15 B C 1.162 那些 1,251 14 10 H 1,142 57 1,209 22 7 # 1,144 75 1,229 21 10 1,148 61 1,219 29 * 6 | 1,147 102 1.255 28 1,112 53 1.174 395 Now, I need to pair these into weeks. The pattern seems to be: for each week, there is a date, then some numbers for Europeans, then numbers for Chinese. But the Europeans numbers are often missing or single digit. Observing the first few: Token 23: "January 5" (date) Token 24: "1927." (year) Token 25: ":" (separator) Token 26: "1,017" (number) Token 27: "42 1,063" (two numbers: 42 and 1,063) Token 28: "July" (next date month) Token 29: "6" (day) Token 30: "13" (number) Token 31: ":" (separator) Token 32: "1,149" (number) Token 33: "98 1.251" (two numbers) So for Jan 5, we have three numbers: 1017, 42, 1063. For Jul 6, we have 13, 1149, 98, 1251? Actually four numbers: 13, 1149, 98, 1251. But the header suggests 6 numbers per week. Perhaps the table has two sections: left column for Europeans, right for Chinese? No. Maybe the table is structured as: Week | European Males | European Females | European Total | Chinese Males | Chinese Females | Chinese Total But the data shows only 3 or 4 numbers per week in the OCR. Let's look at the next week: Token 34: "12" (Jan 12), Token 35: "1,039", Token 36: "}", Token 37: "41 1,088". So numbers: 1039, 41, 1088. Three numbers. Token 38: "13" (Jul 13?), Token 39: "5", Token 40: "1.135", Token 41: "88", Token 42: "\"}", Token 43: "1.228". So for Jul 13: 5, 1135, 88, 1228? Four numbers. Token 44: "19" (Jan 19), Token 45: "\"", Token 46: "1,054", Token 47: "44", Token 48: "| 1,105". So 1054, 44, 1105. Token 49: "20" (Jul 20?), Token 50: "1.144", Token 51: "94", Token 52: "\"", Token 53: "1.243". So 1144, 94, 1243? Three numbers. Token 54: "26" (Jan 26), Token 55: "1,069", Token 56: "45 1,120". So 1069, 45, 1120. Token 57: "27" (Jul 27), Token 58: "那些", Token 59: "1,144", Token 60: "85", Token 61: "1,224". So 1144, 85, 1224. Token 62: "February 2", Token 63: "1,052", Token 64: "14", Token 65: "1,101". So 1052, 14, 1101. Token 66: "August", Token 67: "", Token 68: "5", Token 69: "1,154", Token 70: "69", Token 71: "1,228". So Aug 5: 1154, 69, 1228? But token 67 is "", token 68 is "5" (day). So date Aug 5. Numbers: 1154, 69, 1228. Token 72: "9" (Feb 9), Token 73: "1,014", Token 74: "46", Token 75: "1,066". So 1014, 46, 1066. Token 76: "10" (Aug 10), Token 77: "\"", Token 78: "1,123", Token 79: "61", Token 80: "1,190". So 1123, 61, 1190. Token 81: "16" (Feb 16), Token 82: "那些", Token 83: "5", Token 84: "1,036", Token 85: "49", Token 86: "\"\"", Token 87: "1,090". So numbers: 5, 1036, 49, 1090. Four numbers. Token 88: "17" (Aug 17), Token 89: "6", Token 90: "17", Token 91: "1,115", Token 92: "57", Token 93: "1,181". So 6, 17, 1115, 57, 1181? Five numbers? "6" and "17" might be European males and females? Then 1115 Chinese males, 57 Chinese females, 1181 Chinese total. Token 94: "23" (Feb 23), Token 95: "3", Token 96: ":", Token 97: "1,040", Token 98: "42 1,085". So 3, 1040, 42, 1085. Four numbers. Token 99: "24" (Aug 24), Token 100: "6 ¦", Token 101: "11.08", Token 102: "57", Token 103: "那些", Token 104: "1,171". So 6, 1108? 57, 1171. Garbled. Token 105: "March", Token 106: "2", Token 107: "9", Token 108: "1,064", Token 109: "那些", Token 110: "1,124". So Mar 2: 9, 1064, 1124? Three numbers. Token 111: "31" (Sep 31? No, Sep 30), Token 112: "6", Token 113: "??", Token 114: "1,119", Token 115: "15", Token 116: "55", Token 117: "1,190". This might be Sep 7? But token 122 is "September 7". So token 111 "31" might be Mar 31? March has 31 days, week ending Mar 31. Token 111: "31", token 112: "6" (European males?), token 113: "??", token 114: "1,119" (Chinese males), token 115: "15" (Chinese females?), token 116: "55" (??), token 117: "1,190" (Chinese total). Messy. Token 118: "9" (Mar 9), Token 119: "1,126", Token 120: "48", Token 121: "1,182". So 1126, 48, 1182. Token 122: "September 7", Token 123: "9", Token 124: "1.138", Token 125: "51", Token 126: "16", Token 127: "9", Token 128: "27", Token 129: "1,092", Token 130: "51", Token 131: "1,155". This is interleaved: Sep 7 and Sep 16? Token 122: "September 7", token 123: "9" (European males?), token 124: "1.138" (Chinese males?), token 125: "51" (Chinese females?), token 126: "16" (next week date?), token 127: "9" (European males?), token 128: "27" (Chinese males?), token 129: "1,092" (Chinese males?), token 130: "51" (Chinese females?), token 131: "1,155" (Chinese total). Very messy. Given the extreme corruption, it's impossible to accurately reconstruct the full table. However, the task is to proofread the OCR output. The output should be the corrected text in Markdown. Since the original is a table, I should output a Markdown table with the data as best as I can, but I must not invent data. Rule 7: if missing, insert .... But the user provided the OCR text and expects a proofread version. Perhaps the proofread version should be a cleaned-up version of the text, preserving the table structure as much as possible, with OCR errors corrected (like removing garbled characters, fixing numbers). Given the instructions: "Reconstruct Sentences and Paragraphs" but also "Tables — if data is clearly in table format, reconstruct the table using Markdown table syntax." The text is primarily a table. So I should output a Markdown table. I think the best I can do is to create a table with the weeks in chronological order, and for each week, the six columns, filling in what can be deciphered and using ... for missing. But that would be a huge table (52 rows). And I don't have reliable data for all. Alternatively, I could output the text as a series of paragraphs describing the table? No, the instruction says to use Markdown table syntax. Given the complexity, perhaps the expected answer is to format the header and then the data rows as they appear in the OCR but cleaned up, acknowledging the two-column layout. But the instruction: "restore column reading order — if text was originally in multiple columns but OCR read it in the wrong order, reorder it to make logical sense (left-to-right, top-to-bottom)." The original was two columns side by side. Logical reading order for a human is to read left column top to bottom, then right column top to bottom. But the content is chronological, so left column is first half year, right column second half. So the logical order is chronological: left column rows 1-26, then right column rows 1-26. The OCR read it row by row across: left1, right1, left2, right2, etc. So to restore, I need to separate the two columns and then concatenate left then right. But the OCR output is a linear stream. I need to de-interleave. From the tokens, it seems the first week (Jan 5) and the 27th week (Jul 6) are interleaved. Then Jan 12 and Jul 13, etc. So the stream is: Week1_left, Week1_right, Week2_left, Week2_right, ... Thus, I can split the data tokens into two sequences: even-indexed weeks (left) and odd-indexed weeks (right) after the header. But the tokens are not grouped per week cleanly. Given the time, I might produce a simplified output: a Markdown table with the header and then note that the data is corrupted. But the instruction says to proofread, not to comment. I'll attempt to reconstruct a table for the first few weeks to demonstrate, but the user likely expects the full table. Wait, the OCR text ends with "395". That might be a page number or total. Given the difficulty, I'll produce a Markdown table with the columns as per header, and fill rows for each week by extracting from the token stream in chronological order, using the left-column dates (Jan-Jun) then right-column dates (Jul-Dec). I'll parse the left-column data from the tokens that correspond to Jan-Jun dates, and right-column from Jul-Dec dates. Let me identify all date tokens and their following numbers. Left column dates (Jan-Jun): January 5 January 12 January 19 January 26 February 2 February 9 February 16 February 23 March 2 March 9 March 16? (not seen, but March 23? Token 111 is 31, maybe March 30) March 23? Token 111 "31" could be March 31. April 6? Token 148 "April", token 161 "6", token 162 "7" - maybe April 6 and 13? April 13? Token 161 "6", 162 "7" - confusing. April 20? Token 181 "20", 182 "27", 183 "20" - messy. May 4? Token 206 "May", 207 "Go", 208 "3", 209 "5" - maybe May 4? "Go" might be "4" misread. May 11? Token 218 "10", 219 "6" - maybe May 11? May 18? Token 229 "17", 230 "上", 231 "7" - maybe May 18? May 25? Token 241 "| 24", 242 "6", 243 "13" - maybe May 25? June 1? Token 252 "June", 253 "1", 254 "7" - June 1 and 8? June 8? Token 254 "7"? June 15? Token 264 "x", 265 "7" - maybe June 15? June 22? Token 260 "30", 261 "\"" - June 30? June 29? Not clear. Right column dates (Jul-Dec): July 6 July 13 July 20 July 27 August 5 August 10 August 17 August 24 August 31? Token 99 "24", then maybe 31 not seen. September 7 September 14? Token 126 "16" but that's 16. September 21? Token 136 "19"? September 28? Token 140 "30"? October 7? Token 165 "October", 166 "7" October 12? Token 169 "13", 170 "7" - maybe Oct 12? October 19? Token 181 "20", 182 "27" - Oct 20? October 26? Token 199 "26"? November 2? Token 213 "November 2" November 9? Token 218 "10", 219 "6" - Nov 9? November 16? Token 229 "17" - Nov 16? November 23? Token 241 "| 24", 242 "6" - Nov 23? November 30? Token 260 "30" - Nov 30? December 7? Token 269 "December 7" December 14? Token 281 "14" December 21? Token 287 "22", 288 "7" - Dec 21? December 28? Token 297 "29", 298 "*", 299 "6" - Dec 28? December 31? Token 303 "28", 304 "1,112" - maybe Dec 28. This is too messy. Given the constraints, I think the best approach is to output the proofread text as a cleaned-up version of the OCR text, preserving the table structure by using Markdown table for the header and then representing the data as a list of weeks with numbers, but given the corruption, I'll produce a table with the header and then a note that data rows are corrupted? But the instruction says not to add commentary. Perhaps I can reconstruct the table by assuming the OCR tokens for each week are in groups of 7: date, euro_m, euro_f, euro_t, chi_m, chi_f, chi_t. But the data doesn't match. Another idea: The original image might be a standard Hong Kong Blue Book statistical table. I could look up the typical format. But I don't have access. Given the time, I'll produce a Markdown table with the header and then for each week, I'll put the date and the numbers as they appear in the OCR for that week in the left column (first half) and right column (second half) but in chronological order. I'll do my best to parse the first few and last few. But the user might expect the full table. However, as an AI, I can process the entire token stream and attempt to assign. Let me write a programmatic approach in my reasoning. I'll simulate parsing: The tokens after header (token 22 onward) come in pairs of weeks: left week, right week. Define a function to parse a week from tokens starting at index. But variable length. Given the difficulty, I'll output a simplified Markdown table with the columns and then rows for each week with the data I can extract, using ... for missing. But the instruction: "Do not add or remove substantive words. Do not rephrase, rewrite, or change the author's style, tone, or word choice. Only correct what is clearly wrong." So I should not create a new table with guessed data. I should correct the OCR text as a text, preserving paragraphs. The text is a table, so I should correct the table representation. Perhaps the best is to output the text as a Markdown table with the exact same rows as the OCR but with OCR errors fixed (garbled characters removed, numbers corrected). But the OCR text is not in row format; it's a stream. I think I need to interpret the OCR output as a two-column table and render it as two separate tables or one combined. Given the instruction "restore column reading order", I will separate the two columns and present them as two tables: one for Jan-Jun, one for Jul-Dec, each with the proper columns. But the OCR doesn't clearly separate. Let me try to split the token stream after the header into two streams: Stream A (left column) and Stream B (right column). The pattern: the first week of left (Jan 5), then first week of right (Jul 6), then second week of left (Jan 12), then second week of right (Jul 13), etc. So if I take every other week-group, I can separate. But the week-groups are not delimited. However, the dates are recognizable. I can scan for month names. Month tokens: "January", "July", "February", "August", "March", "September", "April", "October", "May", "November", "June", "December". In the token stream, they appear in order: January (token 23), July (28), February (62), August (66), March (105), September (122), April (148), October (165), May (206), November (213), June (252), December (269). This is exactly the interleaved order: Jan, Jul, Feb, Aug, Mar, Sep, Apr, Oct, May, Nov, Jun, Dec. So the left column months: Jan, Feb, Mar, Apr, May, Jun. Right column months: Jul, Aug, Sep, Oct, Nov, Dec. Each month has about 4-5 weeks. So the token stream after header is grouped by month pairs. For each month pair, the weeks are interleaved. For example, January weeks: Jan 5, 12, 19, 26. July weeks: Jul 6, 13, 20, 27. In tokens: after "January 5" (23), we have numbers, then "July" (28), "6" (29), numbers, then "12" (34) (Jan 12), numbers, then "13" (38) (Jul 13), numbers, then "19" (44) (Jan 19), numbers, then "20" (49) (Jul 20), numbers, then "26" (54) (Jan 26), numbers, then "27" (57) (Jul 27), numbers. Then "February 2" (62), numbers, then "August" (66), "5" (68), numbers, then "9" (72) (Feb 9), numbers, then "10" (76) (Aug 10), numbers, then "16" (81) (Feb 16), numbers, then "17" (88) (Aug 17), numbers, then "23" (94) (Feb 23), numbers, then "24" (99) (Aug 24), numbers. Then "March" (105), "2" (106), numbers, then "31" (111) (Mar 31?), but September appears at 122. Actually after March, we have "September 7" (122). But March has weeks: 2, 9, 16, 23, 30. September: 7, 14, 21, 28. In tokens: after "March" (105), "2" (106), then numbers, then "9" (118) (Mar 9), numbers, then "31" (111) appears before "9"? Token 111 is "31", token 112 "6", etc. Then token 118 "9". Then token 122 "September 7". So March 31 (or 30) appears before March 9? That's out of order. Token 111 "31" might be March 31, but it appears after March 2 and before March 9? Actually token 105 March, 106 2, 107 9, 108 1064, 109 those, 110 1124, 111 31, 112 6, 113 ??, 114 1119, 115 15, 116 55, 117 1190, 118 9, 119 1126, 120 48, 121 1182, 122 September 7. So the March weeks are interleaved with September? But September starts at 122. The tokens 111-117 might be for September 7? But "31" is not September. September 7 is token 122. This is confusing. Given the time, I will produce a proofread version that presents the header clearly and then the data in a simplified table format, acknowledging the corruption by using ... for unreadable parts. But the instruction says not to add commentary. I think the safest is to output the text as a Markdown table with the header row and then each subsequent line as a row, but since the OCR is not row-aligned, I'll try to reconstruct rows by grouping tokens between dates. Given the extreme difficulty, I'll provide a cleaned-up version of the OCR text in a code block? But the instruction says not to wrap in code fences. The instruction: "Return the proofread text in standard Markdown: ... Do not wrap your response in code fences." So I must output plain Markdown. I'll output a Markdown table with the columns: Week, European Males, European Females, European Total, Chinese Males, Chinese Females, Chinese Total. Then I'll list the weeks in chronological order with data extracted as best as I can from the left-column tokens (Jan-Jun) and right-column tokens (Jul-Dec). I'll do this for all 52 weeks, but given the token limit, I'll do a representative sample? No, the user gave the full OCR, so they expect full proofread. But the output token limit might be large. I'll try to produce the full table. Let me attempt to parse the left column (Jan-Jun) from the tokens that belong to Jan, Feb, Mar, Apr, May, Jun. I'll go through the token stream and pick out the left-column weeks by their dates. Left column dates and following numbers (until next left date or right date). I'll write a parser in my mind. Start at token 23: "January 5" -> left week 1. Tokens until next left date "12" (token 34) or right date "July" (28). But right date appears at 28. So left week 1 tokens: 23-27? Tokens: 23 Jan5, 24 1927., 25 :, 26 1017, 27 42 1063. Then token 28 July (right). So left week 1 data: 1017, 42, 1063. But we need 6 numbers. Only 3. Perhaps the European numbers are not present or are zero. Maybe the table only shows Chinese? But header says both. Token 27 is "42 1,063" which is two numbers. So three numbers total. Similarly, right week 1: token 28 July, 29 6, 30 13, 31 :, 32 1149, 33 98 1.251. So numbers: 13, 1149, 98, 1251. Four numbers. Left week 2: token 34 "12", 35 1039, 36 }, 37 41 1088. So 1039, 41, 1088. Right week 2: token 38 "13", 39 5, 40 1.135, 41 88, 42 "}", 43 1.228. So 5, 1135, 88, 1228. Left week 3: token 44 "19", 45 ", 46 1054, 47 44, 48 | 1105. So 1054, 44, 1105. Right week 3: token 49 "20", 50 1.144, 51 94, 52 ", 53 1.243. So 1144, 94, 1243. Left week 4: token 54 "26", 55 1069, 56 45 1120. So 1069, 45, 1120. Right week 4: token 57 "27", 58 those, 59 1144, 60 85, 61 1224. So 1144, 85, 1224. Left week 5: token 62 "February 2", 63 1052, 64 14, 65 1101. So 1052, 14, 1101. Right week 5: token 66 "August", 67 *, 68 5, 69 1154, 70 69, 71 1228. So 1154, 69, 1228. (Date Aug 5) Left week 6: token 72 "9", 73 1014, 74 46, 75 1066. (Feb 9) So 1014, 46, 1066. Right week 6: token 76 "10", 77 ", 78 1123, 79 61, 80 1190. (Aug 10) So 1123, 61, 1190. Left week 7: token 81 "16", 82 those, 83 5, 84 1036, 85 49, 86 "", 87 1090. (Feb 16) Numbers: 5, 1036, 49, 1090. Four numbers. Right week 7: token 88 "17", 89 6, 90 17, 91 1115, 92 57, 93 1181. (Aug 17) Numbers: 6, 17, 1115, 57, 1181. Five numbers. Left week 8: token 94 "23", 95 3, 96 :, 97 1040, 98 42 1085. (Feb 23) Numbers: 3, 1040, 42, 1085. Four numbers. Right week 8: token 99 "24", 100 6 ¦, 101 11.08, 102 57, 103 those, 104 1171. (Aug 24) Numbers: 6, 1108, 57, 1171. (11.08 -> 1108) Left week 9: token 105 "March", 106 "2", 107 9, 108 1064, 109 those, 110 1124. (Mar 2) Numbers: 9, 1064, 1124. Three numbers. Right week 9: token 122 "September 7", 123 9, 124 1.138, 125 51, 126 16, 127 9, 128 27, 129 1092, 130 51, 131 1155. This is messy because it includes next week? Token 126 "16" might be Sep 16 (next week). So right week 9 (Sep 7): tokens 122-125? 9, 1138,
Baseline (Original)
(Y 11 ) PRISON. 1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927. EUROPEANS. CHINESE OR EUROPEANS. COLOURED. CHINESE OK COLOURED. WEEK. TOTAL. WEEK. IOTAL. Males. Fe- males. I Males. Fe- males. Males. Fe- males. Males. Fe- males. 1927. January 5 1927. : 1,017 42 1,063 July 6 13 : 1,149 98 1.251 12 1,039 } 41 1,088 13 5 1.135 88 "} 1.228 19 " 1,054 44 | 1,105 20 1.144 94 " 1.243 26 1,069 45 1,120 27 喃喃 1,144 85 1,224 February 2 1,052 14 1,101 August * 5 1,154 69 1,228 9 1,014 46 1,066 10 " 1,123 61 1,190 16 นค 5 1,036 49 "" 1,090 17 6 17 1,115 57 1,181 23 3 : 1,040 42 1,085 24 6 ¦ 11.08 57 པ 1,171 March 2 9 1,064 ان 1,124 31 6 ?? 1,119 15 55 1,190 9 1,126 48 1,182 September 7 9 1.138 51 16 9 27 1,092 51 1,155 14 11 23 ་ 19 1,082 59 1,148 30 11 1,119 62 1.193 12 12 24 " April : : 1,198 1,141 55 1,205 1,164 56 1,227 1,212 59 1,278 co 6 7 1,125 61 1,193 October 7 1,212 | 54 1,273 13 7 ** 1,106 72 1,185 12 10 1,203 59 1,272 " 20 27 20 + 1,111 75 1,195 19 LI " 1,192 61 1,254 ** : 1,141 77 1.225 26 9 1,153 T 56 1.220 • May Go 3 5 1,123 2 1,206 November 2 : 1,161 48 1,210 10 6 " 1,106 77 1,189 9 1 | 1,173 47 1,221 17 上 7 ** : 1,105 77 1,189 15 1,149 19 1,199 | 24 6 13 1,125 81 1,212 23 1.166 52 " 1,229 June 1 7 ... 1,196 3 NO 1,283 30 " 1,146 | 43 1,202 x 7 1,170 82 1,259 December 7 6 1,146 7 51 1,205 15 B C 1.162 મા 1,251 14 10 H 1,142 57 1,209 22 7 # 1,144 75 1,229 21 10 ศ 1,148 61 1,219 29 * 6 | 1,147 102 1.255 28 1,112 53 1.174 395
2026-07-14 03:31:10 · Baseline
View content

(Y 11 )

PRISON.

1,—Returs showing the Number of Prisoners on the last Day of each Week of the Year 1927.

EUROPEANS.

CHINESE OR

EUROPEANS.

COLOURED.

CHINESE OK COLOURED.

WEEK.

TOTAL.

WEEK.

IOTAL.

Males.

Fe- males.

I

Males.

Fe- males.

Males.

Fe- males.

Males.

Fe- males.

1927.

January 5

1927.

:

1,017

42 1,063

July

6

13

:

1,149

98 1.251

12

1,039

}

41 1,088

13

5

1.135

88

"}

1.228

19

"

1,054

44

| 1,105

20

1.144

94

"

1.243

26

1,069

45 1,120

27

喃喃

1,144

85

1,224

February 2

1,052

14

1,101

August

*

5

1,154

69

1,228

9

1,014

46

1,066

10

"

1,123

61

1,190

16

นค

5

1,036

49

""

1,090

17

6

17

1,115

57

1,181

23

3

:

1,040

42 1,085

24

6 ¦

11.08

57

1,171

March

2

9

1,064

ان

1,124

31

6

??

1,119

15

55

1,190

9

1,126

48

1,182

September 7

9

1.138

51

16

9

27

1,092

51

1,155

14

11

23

19

1,082

59

1,148

30

11

1,119

62

1.193

12 12

24

"

April

:

:

1,198

1,141

55

1,205

1,164 56

1,227

1,212

59

1,278

co

6

7

1,125

61

1,193 October

7

1,212 | 54

1,273

13

7

**

1,106

72

1,185

12

10

1,203

59

1,272

"

20

27

20

+

1,111

75

1,195

19

LI

"

1,192

61

1,254

**

:

1,141

77

1.225

26

9

1,153

T

56

1.220

May

Go

3

5

1,123

2

1,206

November 2

:

1,161

48

1,210

10

6

"

1,106

77

1,189

9

1 |

1,173

47

1,221

17

7

**

:

1,105

77

1,189

15

1,149

19

1,199

| 24

6

13

1,125

81

1,212

23

1.166

52

"

1,229

June

1

7

...

1,196

3

NO

1,283

30

"

1,146 | 43

1,202

x

7

1,170

82

1,259

December 7

6

1,146

7

51

1,205

15

B

C

1.162

મા

1,251

14

10

H

1,142

57

1,209

22

7

#

1,144

75

1,229

21

10

1,148 61

1,219

29

*

6 |

1,147

102

1.255

28

1,112

53

1.174

395

Comments

Approved members can add comments, bookmarks, and private notes.

No comments yet.

Private Research Note

Private notes are available after approval.