1935 — Page 398

Blue Books 香港計冊 All AI Reviewed

The user wants me to proofread OCR output of a historical table about daily average prisoners in Hong Kong from 1926-1935. The OCR is messy with broken lines, misaligned columns, and garbled text. I need to reconstruct this as a proper Markdown table.

First, let me understand the structure. The title says "STATEMENT of the DAILY AVERAGE of PRISONERS for each Month of the Year, from 1926 to 1935." But the years shown are 1926-1935, though the header shows 1926, 1927, 1928, 1929, 1930, 1931, 1932, 1933, 1934, 1925? Wait, the last column says 1925 but that might be a typo for 1935.

Looking at the data: There are columns for E (Europeans), I (Indians), C (Chinese) for each year. The months are January through December. Then totals and averages.

The OCR has many artifacts: "Jutinary" = January, "Mareli" = March, "KA" maybe a page marker, "Y 16" maybe a reference.

I need to reconstruct the table with proper columns. Let me parse the data row by row.

The header row seems to be: Months, then for each year 1926-1935, three sub-columns: E, I, C.

But the OCR shows: "1926. 1927. 1928. 1929. 1930. 1931. 1932. 1933. 1934. 1925" - likely 1935.

Then "396 (Y 16) E. : C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C." This is messy.

Actually, looking at the data rows, each month has numbers for E, I, C for each year. For example, January row: "X 50 6 958 6 8 | 1,080 11 10 1,103 5 14 970 12 61,198 9 1,149 16 6 1,044 22 G 1,245 18 41,462 20 5 1,667"

This is garbled. Let me try to interpret.

Better approach: The table likely has 10 years (1926-1935), each with 3 columns (E, I, C). So 30 data columns plus month column. That's huge. But the OCR might have merged columns.

Wait, the title says "from 1926 to 1935" inclusive, that's 10 years. But the header shows 1926, 1927, 1928, 1929, 1930, 1931, 1932, 1933, 1934, 1925 (probably 1935). So 10 years.

Each year has E, I, C. So 30 numeric columns.

But the OCR text shows numbers like "958", "1,080", "1,103", "970", "1,198", "1,149", "1,044", "1,245", "1,462", "1,667" for January across years? Let's count: 10 numbers? Actually there are more.

Let me look at the January line: "X 50 6 958 6 8 | 1,080 11 10 1,103 5 14 970 12 61,198 9 1,149 16 6 1,044 22 G 1,245 18 41,462 20 5 1,667"

Maybe "X 50" is something else. "6 958" could be E=6, I=958? But 958 seems high for Indians. Actually, the numbers might be daily averages. For 1926, Europeans maybe 6, Indians 958? That seems off.

Wait, the legend at bottom: "E.=Europeans. I.Indians, C. Chinese." So three categories.

Perhaps the table is structured with months as rows, and for each year, three columns. But the OCR has lost the column boundaries.

I need to reconstruct based on typical Hong Kong prison statistics. The total prisoners per day average might be around 1000-2000. The numbers like 958, 1080, 1103, 970, 1198, 1149, 1044, 1245, 1462, 1667 could be the Chinese (C) column for each year? Or totals?

Look at the TOTAL row: "TOTAL 74 133 12,436 | 87 8114,095 103 139 12,600|117 112 12,671 102 85 19.918 141 90 12,901 221 40 13,110 260 52 17,380 283 44 18,096 225 6521,258"

This is messy. "74 133 12,436" might be totals for 1926: E=74, I=133, C=12,436? But 12,436 for Chinese seems high for daily average? Actually daily average per month? Wait, the table is "DAILY AVERAGE of PRISONERS for each Month". So each cell is the daily average for that month. The TOTAL row might be sum of monthly averages? Or total for the year? The AVERAGE row shows "AVERAGE row: "AVERAGE,... 6 11 1,087 7 7 1,175 9 11 1,051 10 9 1,056 9 7 1,160 12 81,082 18 8 1,093 22 10 5 1,416 24 41,589 19 5 1,772"

This suggests for each year, three numbers: E, I, C averages. For 1926: 6, 11, 1,087. For 1927: 7, 7, 1,175. For 1928: 9, 11, 1,051. For 1929: 10, 9, 1,056. For 1930: 9, 7, 1,160. For 1931: 12, 8, 1,082? Wait "81,082" maybe 8, 1,082. For 1932: 18, 8, 1,093. For 1933: 22, 10, 1,416? "22 10 5 1,416" maybe 22, 10, 1,416. For 1934: 24, 4, 1,589? "24 41,589" maybe 24, 4, 1,589. For 1935: 19, 5, 1,772.

That seems plausible: Europeans small (6-24), Indians small (7-11), Chinese large (1000-1772). So the table has 10 years, each with E, I, C.

Now the monthly rows should have similar triplets per year.

Let's parse January row with this hypothesis. The row starts with "Jutinary." (January). Then numbers: "X 50 6 958 6 8 | 1,080 11 10 1,103 5 14 970 12 61,198 9 1,149 16 6 1,044 22 G 1,245 18 41,462 20 5 1,667"

We need to extract 30 numbers (10 years * 3). But there are many numbers. Let's list all numbers in that line: 50, 6, 958, 6, 8, 1080, 11, 10, 1103, 5, 14, 970, 12, 61, 1198? "61,198" might be 61 and 1198? Or 61,198 as one number? But 61,198 is too large. Probably "61,198" is two numbers: 61 and 1198? But 61 for Europeans? Could be. Then 9, 1149, 16, 6, 1044, 22, 1245, 18, 41, 1462? "41,462" -> 41 and 1462. Then 20, 5, 1667.

That gives: 50, 6, 958, 6, 8, 1080, 11, 10, 1103, 5, 14, 970, 12, 61, 1198, 9, 1149, 16, 6, 1044, 22, 1245, 18, 41, 1462, 20, 5, 1667. That's 27 numbers. Need 30. Missing three.

Maybe "X" is not a number. "50" might be something else. Could be the month number? No.

Let's look at February row: "February, 7 952 5 61,078) 1,037 6 12 # 97 10 7 1,168 9 9 1,167 23 41,056 24 61,303 18 x"

Numbers: 7, 952, 5, 61, 1078? "61,078)" -> 61, 1078? Then 1037, 6, 12, 97, 10, 7, 1168, 9, 9, 1167, 23, 41, 1056? "41,056" -> 41, 1056. Then 24, 61, 1303? "61,303" -> 61, 1303. Then 18.

That's messy.

Perhaps the OCR has merged columns and there are vertical bars indicating separation. In January line there is a "|" after "8". In February there is ")" and "#". In March: "Mareli, 9 993 >>> 9 81,139 10 91,022 11 973 71,140 5 6 1,053 16 21,067 22 61,340 22". Numbers: 9, 993, 9, 81, 1139? "81,139" -> 81, 1139. 10, 91, 1022? "91,022" -> 91, 1022. 11, 973, 71, 1140? "71,140" -> 71, 1140. 5, 6, 1053, 16, 21, 1067? "21,067" -> 21, 1067. 22, 61, 1340? "61,340" -> 61, 1340. 22.

April: "April, 5 9 1,018 8 71,174 9 7 1,049 91 101,022 4 30 7 1,142 7 6 1,037 13 31,053 26". Numbers: 5, 9, 1018, 8, 71, 1174? "71,174" -> 71, 1174. 9, 7, 1049, 91, 101, 1022? "101,022" -> 101, 1022. 4, 30, 7, 1142, 7, 6, 1037, 13, 31, 1053? "31,053" -> 31, 1053. 26.

May: "May, 6 1,066 6 71,197 t 8 1,086 7 10 1,013 7 7 1,148 * 7 998 11 21,070 24". Numbers: 6, 1066, 6, 71, 1197? "71,197" -> 71, 1197. 8, 1086, 7, 10, 1013, 7, 7, 1148, 7, 998, 11, 21, 1070? "21,070" -> 21, 1070. 24.

June: "June, 6 9 1,078 8 1,225 8 7 1,106 00 8 9 996 7 7 1,117 12 9 1,077 12 2 1,050 25 4 1,396 23 3 1,537 23 21,549 25". Numbers: 6, 9, 1078, 8, 1225, 8, 7, 1106, 00, 8, 9, 996, 7, 7, 1117, 12, 9, 1077, 12, 2, 1050, 25, 4, 1396, 23, 3, 1537, 23, 21, 1549? "21,549" -> 21, 1549. 25.

July: "July, 6 9 1,078 5 81,224 8 15 1,079 7 6 1,028 7 51,113 I 7 1,065 20". Numbers: 6, 9, 1078, 5, 81, 1224? "81,224" -> 81, 1224. 8, 15, 1079, 7, 6, 1028, 7, 51, 1113? "51,113" -> 51, 1113. 7, 1065, 20.

August: "August, September,... October, 6 15 1,100 6 51,168 11 141,035 10 8 1,070 10 7 1,192 00 8 7 1,065 22 22 3 1,075 20 2 1,582 23 I 1,061 19". This line seems to combine three months? "August, September,... October," then numbers. Might be three rows merged.

November: "November, December, 5 4 15 1,050 10 11 1,038 9 5 1,196 10 1,181 9 16 1,022 13 6 19 985 16 6 1,174 D 81,193 18 6 1,184 8 1,127 19 8 1,143 7 1,139 22 8 11,094 26 23 223 4 1,127 18 4 1,185 6 | 1,221 18 21 51,491 30 61,350 23 6 1,471 21 5 1,607 20 51,644 18 61,681 14 31,50 23 4 1,709 21,449 24 1|1,493 .17 21,590 17 3 1,658 16 3 1.621 19 41,683 18 + 1,787 51,718 5 1,749 41,813 61,802 6 1,841 1,857 6 1,684 6 1,806 61,805"

This is a mess. It seems the OCR has run multiple rows together.

Given the complexity, perhaps the original table is too large to reconstruct perfectly from this OCR. But the user expects me to produce a Markdown table. I need to do my best.

Maybe the table is actually two separate tables? The title says "2.--STATEMENT of the DAILY AVERAGE of PRISONERS for each Month of the Year, from 1926 to 1935." That's one table.

But the OCR includes "396 (Y 16)" at top, maybe a page number.

Given the difficulty, I should reconstruct a simplified table with the data I can parse, perhaps only the totals and averages, or note that the monthly data is too garbled.

But the instruction: "If data is clearly in table format, reconstruct the table using Markdown table syntax." I must try.

Let me think: The table likely has 12 months + Total + Average rows. Columns: Month, then for each year 1926-1935, three subcolumns: E, I, C. That's 31 columns. Too wide for Markdown but possible.

Alternatively, the table might be split by year: each year has a sub-table? But the OCR shows all years horizontally.

Given the OCR is so corrupted, perhaps the best is to present the data as a series of tables per year? But the original is one table.

I could reconstruct the table with the years as columns and the three categories as sub-rows? But the original has months as rows.

Let me try to parse the monthly data for each year from the OCR by aligning with the average row.

From the average row, we have for each year (1926-1935) the average daily prisoners: E, I, C.

1926: 6, 11, 1087

1927: 7, 7, 1175

1928: 9, 11, 1051

1929: 10, 9, 1056

1930: 9, 7, 1160

1931: 12, 8, 1082

1932: 18, 8, 1093

1933: 22, 10, 1416

1934: 24, 4, 1589

1935: 19, 5, 1772

Now, the monthly rows should have similar magnitudes. For January, the Chinese numbers around 1000-1600. The numbers I extracted: 958, 1080, 1103, 970, 1198, 1149, 1044, 1245, 1462, 1667. That's 10 numbers, matching 10 years. And they are in the range of Chinese averages. So likely those are the Chinese (C) column for each year for January.

Similarly, the Europeans and Indians numbers are smaller. In January line, before each Chinese number there are two numbers? Let's see: "6 958" -> maybe E=6, I=958? But 958 is too high for Indians. Wait, the average for Indians in 1926 is 11. So 958 cannot be Indians. So "6 958" might be E=6, C=958? But then where is Indians? The pattern might be E, C, I? But the legend says E, I, C.

Look at the average row: "6 11 1,087" for 1926. So order is E, I, C.

In January, for 1926, we might have E, I, C. The numbers "6 958" - if 6 is E, 958 is I? No. Could be that the OCR missed a column. Perhaps the table has only two categories? But legend says three.

Maybe the table is: Year, European, Indian, Chinese. But the monthly data might be presented as: Month, then for each year: European, Indian, Chinese. But the OCR has lost the Indian column for some rows.

Let's check February: "7 952" - 7 could be European, 952 Chinese? But 952 is lower than January's 958? Possible. Then "5 61,078)" - 5 European, 61 Indian? 61 Indian possible? Average Indian for 1927 is 7. 61 is high. "1,037" Chinese? Then "6 12" - 6 European, 12 Indian? "97 10" - 97 European? Too high. "7 1,168" - 7 European, 1168 Chinese. "9 9 1,167" - 9 European, 9 Indian, 1167 Chinese. "23 41,056" - 23 European, 41 Indian, 1056 Chinese? "24 61,303" - 24 European, 61 Indian, 1303 Chinese. "18 x" - 18 European.

This is inconsistent.

Given the time, perhaps the best is to output the table as it appears in the OCR but cleaned up, with a note that the data is garbled. But the instruction says to correct unambiguous OCR errors and reconstruct tables.

I think I need to make a judgment: The table is too corrupted to fully reconstruct. However, I can reconstruct the structure and fill in what I can, using ... for missing.

But the user might expect a full table. Let me look at the OCR again: It might be from a known Hong Kong Blue Book? This looks like a statistical table from Hong Kong Blue Book 1935. The table "Daily Average of Prisoners" might be standard.

I could search my memory: In Hong Kong Blue Books, there is a table "Daily Average Number of Prisoners" by month and year, with breakdown by Europeans, Indians, Chinese. The years 1926-1935.

Given the difficulty, I will reconstruct a Markdown table with the years as columns and months as rows, but only include the total and average rows which are clearer, and for monthly rows, I will put the Chinese numbers I extracted for January, and for others I'll try to parse similarly.

But the instruction: "Preserve original content — 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 must not invent data.

Thus, I should present the text as a cleaned-up version of the OCR, preserving the table structure as best as possible, with ... for illegible parts.

I'll output the text in Markdown, with a table for the monthly data, but given the corruption, I might represent it as a code block? No, the instruction says use Markdown table syntax.

Let me try to create a table with the columns: Month, 1926 E, 1926 I, 1926 C, 1927 E, 1927 I, 1927 C, ... up to 1935.

But I don't have all data. I could leave cells empty or with ... .

Given the complexity, I'll produce a Markdown table that captures the structure and the data that is legible, using ... for missing.

First, the header: "MONTHS. 1926. 1927. 1928. 1929. 1930. 1931. 1932. 1933. 1934. 1935." (correct 1925 to 1935).

Then the subheader: "E. I. C." repeated for each year.

Then rows.

I'll parse each month row from the OCR text, splitting by the apparent delimiters.

Let me write a script in my mind to parse. But as an AI, I can do it manually.

I'll go through the OCR line by line.

The OCR text provided:

MONTHS. 

1926. 


2.--STATEMENT of the DAILY AVERAGE of PRISONERS for each Month of the Year, from 1926 to 1935. 

1927. 

1928. 

1929. 

1930. 

1931. 

1932. 

1933. 

1934. 

1925, 


396 

(Y 16 ) 

E. 

:་ 

C. E. 

I. C. E. 

I. C. 

E. 

I. C. E. 

I. 

C. E. I. 

C. 

E. 

I. C. 

E. I. C. E. 

I. 

C. E. 

I. C. 

Jutinary. 

X 

50 

6 958 

6 

8 | 1,080 

11 

10 1,103 

5 

14 

970 12 

61,198 

ן[ 

9 1,149 

16 

6 1,044 

22 G 1,245 18 

41,462 20 5 1,667 

| 

February, 

7 

952 

5 

61,078) 

1,037 

G 

12 

# 

97 10 

7 1,168 9 

9 1,167 

23 

41,056 24 

61,303 18 

x 

Mareli, 

9 993 

>>> 

9 

81,139 

10 

91,022 

11 

973 

71,140 

5 6 1,053 

16 

21,067 22 

61,340 22 

April, 

5 9 1,018 8 

71,174 

9 

7 1,049 

91 

101,022 ४ 

30 

7 1,142 

7 6 1,037 

13 

31,053 26 

May, 

6 1,066 

6 

71,197 

t 

8 1,086 

7 

10 1,013 7 

7 1,148 

* 

7 998 

11 

21,070 24 

. June, 

6 

9 1,078 

8 1,225 

8 

7 1,106 

00 

8 

9 996 

7 

7 1,117 

12 

9 1,077 12 

2 1,050 25 

4 1,396 23 

3 1,537 23 

21,549 25 

July, 

6 

9 1,078 

5 

81,224 

8 

15 1,079 

7 

6 1,028 

7 

51,113 

I 

7 1,065 20 

August, 

September,... 

October, 

6 

15 1,100 

6 

51,168 

11 

141,035 10 

8 1,070 10 

7 1,192 

00 

8 

7 1,065 22 

22 

3 1,075 20 

2 1,582 23 

I 1,061 19 

7 

17 1,079 7 

4 1,200 

સ 

13 1,041 17 

11 1,108 

t- 

7 

.8 473 18 

7 1,004 17 

21,071 2】 

3 1,561 28 

31.525 29 

6 1,599 19 

KA 

5 

19 1,005 

9 

5 1,233 

8 

131,041 15) 

10 1,159 

9 

8 1,207 18 

November, 

December, 

5 

4 15 1,050 10 

11 1,038 9 

5 1,196 

10 1,181 

9 16 1,022 13 

6 19 985 16 

6 1,174 

D 

81,193 18 

6 1,184 

8 1,127 19 

8 1,143 

7 1,139 22 

8 11,094 26 

23 

223 

4 1,127 18 

4 1,185 

6 | 1,221 

18 

21 

51,491 30 

61,350 23 

6 1,471 21 

5 1,607 20 

51,644 18 

61,681 14 

31,50 23 4 1,709 

21,449 24 

1|1,493 .17 

21,590 17 

3 1,658 16 

3 1.621 19 

41,683 18 

+ 1,787 

51,718 

5 1,749 

41,813 

61,802 

6 1,841 

1,857 

6 1,684 

6 1,806 

61,805 


TOTAL 74 133 12,436 | 87 

8114,095 103 139 12,600|117 112 12,671 102 

85 19.918 141 90 12,901 221 

40 13,110 260 

52 17,380 283 

44 18,096 225 

6521,258 

AVERAGE,... 

6 11 1,087 7 

7 1,175 

9 11 1,051 10 

9 1,056 9 

7 1,160 

12 

81,082 18 

8 1,093 

22 

10 

5 

1,416 

24 

41,589 19 

5 1,772 

E.=Europeans. 

I.Indians, 

C. Chinese.

I need to clean this up.

First, the title: "2.--STATEMENT of the DAILY AVERAGE of PRISONERS for each Month of the Year, from 1926 to 1935."

The header: "MONTHS." then years 1926 to 1935. The last year is mis-OCRed as 1925, should be 1935.

Then there is a line "396 (Y 16)" probably a page number: "396 (Y 16)".

Then column headers: "E. I. C." repeated for each year. The OCR shows a mess: "E. : C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C. E. I. C." So 10 times.

Now the monthly rows. Each month row starts with month name. Then for each year, three numbers (E, I, C). But the OCR has lost alignment.

Given the average row is clear, I can use that to verify the number of columns.

The average row: "AVERAGE,... 6 11 1,087 7 7 1,175 9 11 1,051 10 9 1,056 9 7 1,160 12 8 1,082 18 8 1,093 22 10 1,416 24 4 1,589 19 5 1,772"

I've parsed that as 10 triplets.

Similarly, the TOTAL row: "TOTAL 74 133 12,436 87 114,095 103 139 12,600 117 112 12,671 102 85 19,918 141 90 12,901 221 40 13,110 260 52 17,380 283 44 18,096 225 65 21,258" (I cleaned up the OCR: "8114,095" -> "114,095"? Actually "8114,095" might be "114,095"? But 81,14,095? Let's see: "87 8114,095" -> maybe 87, 114,095? But 114,095 is huge. The total for a year might be sum of monthly averages? For 1926, total Chinese 12,436, European 74, Indian 133. That sums to 12,643. For 1927, 87, 114,095? That seems off. Wait, "8114,095" could be "114,095"? But 114,095 for Chinese total? That would be sum of monthly averages? If average Chinese is 1175, times 12 = 14,100. So 114,095 is too high. Maybe it's 11,409.5? The OCR "8114,095" might be "11,409.5"? But the pattern: "74 133 12,436" for 1926. For 1927: "87 114,095"? That doesn't match. Let's look at the OCR: "TOTAL 74 133 12,436 | 87 8114,095 103 139 12,600|117 112 12,671 102 85 19.918 141 90 12,901 221 40 13,110 260 52 17,380 283 44 18,096 225 6521,258"

Maybe the totals are for each category across months? For 1926: European total 74, Indian 133, Chinese 12,436. For 1927: European 87, Indian 114? "8114,095" could be "114,095"? But 114,095 is not plausible. Could be "11,409.5"? But the Chinese total for 1927 should be around 1175*12=14,100. 12,600 appears for 1928? "103 139 12,600" for 1928? That matches: European 103, Indian 139, Chinese 12,600. For 1929: "117 112 12,671". For 1930: "102 85 19,918"? 19,918 seems high. For 1931: "141 90 12,901". For 1932: "221 40 13,110". For 1933: "260 52 17,380". For 1934: "283 44 18,096". For 1935: "225 65 21,258". That seems plausible: European totals increase over years, Indian totals vary, Chinese totals increase.

So the TOTAL row has 10 triplets.

Thus the table has 10 years, each with E, I, C.

Now for monthly rows, each should have 30 numbers. But the OCR is missing many.

Given the difficulty, I will reconstruct the table in Markdown with the data I can extract for January (the first month) as an example, and for the rest, I will indicate the data is illegible. But the instruction says to reconstruct the table. Perhaps I can create a table with the months and the years, and fill in the Chinese column only? But that would be altering.

Given the constraints, I think the best is to output the text as a cleaned-up version of the OCR, preserving the table structure with Markdown table syntax, but with many cells containing "..." due to OCR corruption. I'll note that the monthly data is largely illegible.

But the user might want the table as it appears in the original. Since I cannot know the original, I must do my best.

I'll create a Markdown table with the header row and the total and average rows, and for monthly rows, I'll put the month name and then "..." for each year's triplet.

However, the instruction: "If data is clearly in table format, reconstruct the table using Markdown table syntax." The data is clearly in table format, but the OCR is too garbled. I should still attempt to reconstruct the table with the available data.

Let me try to parse each month row into 30 numbers by using the pattern: each year has three numbers, often separated by spaces, and the Chinese number is usually the largest and often has a comma (e.g., 1,080). The European and Indian are small integers.

I'll write a parsing algorithm mentally for each month.

But given the time, I'll produce a simplified table: only the Total and Average rows, and mention that monthly data is illegible. But that might not satisfy.

Alternatively, I can output the entire OCR text cleaned up as paragraphs, but the table must be in Markdown table.

I'll do the following: Create a Markdown table with columns: Month, 1926 E, 1926 I, 1926 C, 1927 E, 1927 I, 1927 C, 1928 E, 1928 I, 1928 C, 1929 E, 1929 I, 1929 C, 1930 E, 1930 I, 1930 C, 1931 E, 1931 I, 1931 C, 1932 E, 1932 I, 1932 C, 1933 E, 1933 I, 1933 C, 1934 E, 1934 I, 1934 C, 1935 E, 1935 I, 1935 C.

Then fill in the Total and Average rows fully. For monthly rows, I'll fill in what I can from the OCR, using ... for missing.

Let's extract January:

From OCR: "Jutinary. X 50 6 958 6 8 | 1,080 11 10 1,103 5 14 970 12 61,198 9 1,149 16 6 1,044 22 G 1,245 18 41,462 20 5 1,667"

I need to map to 10 triplets. Let's assume the numbers appear in order: for each year, E, I, C. But the OCR might have omitted some.

Let's list all numbers in order, ignoring non-numeric: 50, 6, 958, 6, 8, 1080, 11, 10, 1103, 5, 14, 970, 12, 61, 1198, 9, 1149, 16, 6, 1044, 22, 1245, 18, 41, 1462, 20, 5, 1667.

That's 28 numbers. We need 30. Maybe "X" is 10? No.

Perhaps the first "50" is not part of data. "Jutinary. X 50" might be "January 1926"? No.

Maybe the table has a column for "Month" and then the years. The "X 50" could be a page artifact.

Let's look at February: "February, 7 952 5 61,078) 1,037 G 12 # 97 10 7 1,168 9 9 1,167 23 41,056 24 61,303 18 x"

Numbers: 7, 952, 5, 61, 1078, 1037, 12, 97, 10, 7, 1168, 9, 9, 1167, 23, 41, 1056, 24, 61, 1303, 18. That's 21 numbers.

March: "Mareli, 9 993 9 81,139 10 91,022 11 973 71,140 5 6 1,053 16 21,067 22 61,340 22"

Numbers: 9, 993, 9, 81, 1139, 10, 91, 1022, 11, 973, 71, 1140, 5, 6, 1053, 16, 21, 1067, 22, 61, 1340, 22. 22 numbers.

April: "April, 5 9 1,018 8 71,174 9 7 1,049 91 101,022 4 30 7 1,142 7 6 1,037 13 31,053 26"

Numbers: 5, 9, 1018, 8, 71, 1174, 9, 7, 1049, 91, 101, 1022, 4, 30, 7, 1142, 7, 6, 1037, 13, 31, 1053, 26. 23 numbers.

May: "May, 6 1,066 6 71,197 8 1,086 7 10 1,013 7 7 1,148 7 998 11 21,070 24"

Numbers: 6, 1066, 6, 71, 1197, 8, 1086, 7, 10, 1013, 7, 7, 1148, 7, 998, 11, 21, 1070, 24. 19 numbers.

June: "June, 6 9 1,078 8 1,225 8 7 1,106 00 8 9 996 7 7 1,117 12 9 1,077 12 2 1,050 25 4 1,396 23 3 1,537 23 21,549 25"

Numbers: 6, 9, 1078, 8, 1225, 8, 7, 1106, 0, 8, 9, 996, 7, 7, 1117, 12, 9, 1077, 12, 2, 1050, 25, 4, 1396, 23, 3, 1537, 23, 21, 1549, 25. 30 numbers! That's promising. June has 30 numbers. Let's see if they group into 10 triplets.

Group every three: (6,9,1078), (8,1225,8), (7,1106,0), (8,9,996), (7,7,1117), (12,9,1077), (12,2,1050), (25,4,1396), (23,3,1537), (23,21,1549), (25) leftover. Actually 30 numbers give 10 triplets exactly. Let's list:

6, 9, 1078
8, 1225, 8
7, 1106, 0
8, 9, 996
7, 7, 1117
12, 9, 1077
12, 2, 1050
25, 4, 1396
23, 3, 1537
23, 21, 1549

But the 10th triplet has 23, 21, 1549. Then there's an extra 25? Wait, I counted 30 numbers: let's count again:

1.6

2.9

3.1078

4.8

5.1225

6.8

7.7

8.1106

9.00 (0)

10.8

11.9

12.996

13.7

14.7

15.1117

16.12

17.9

18.1077

19.12

20.2

21.1050

22.25

23.4

24.1396

25.23

26.3

27.1537

28.23

29.21

30.1549

31.25? The OCR ends with "25". Actually the line: "21,549 25" -> 21, 1549, 25. So there are 31 numbers. Let's include the last 25: 31 numbers. So not exact.

But maybe the first number "6" is not part of the first year? The month row might start with the month name, then the first year's E. For June, the first year 1926: E=6, I=9, C=1078. That matches the average for 1926: E=6, I=11, C=1087. Close: I=9 vs 11, C=1078 vs 1087. Good.

Second year 1927: E=8, I=1225? That's huge for Indians. Average for 1927 Indians is 7. So 1225 cannot be Indians. So the grouping is not E, I, C. Maybe it's E, C, I? But 1225 for Chinese? Average Chinese 1175, so 1225 possible for June. Then the third number 8 would be Indians? 8 Indians possible. So triplet: E=8, C=1225, I=8. But the order in header is E, I, C. The average row is E, I, C. So the monthly should be same order.

But 1225 for Indians is impossible. So perhaps the columns are not consistent. Maybe the table has only two categories: European and Chinese? But the legend says three.

Wait, the average row for 1927: 7, 7, 1175. So Indians=7. In June, if the second year's numbers are 8, 1225, 8, then maybe it's European=8, Chinese=1225, Indian=8? But Indian=8 is close to 7. Chinese=1225 close to 1175. European=8 close to 7. So the order might be E, C, I. But the header says E, I, C. Could be a mistake in the OCR header.

Let's check 1928 average: 9, 11, 1051. June third triplet: 7, 1106, 0. If order E, C, I: E=7, C=1106, I=0. But average E=9, C=1051, I=11. Not close.

If order E, I, C: E=7, I=1106, C=0. No.

Maybe the June row is not aligned with years because of missing months? The table might have months as rows, but the OCR has lost the column boundaries.

Given the difficulty, I think the only reliable rows are the Total and Average rows. The monthly rows are too corrupted.

Thus, I will reconstruct the table with only the Total and Average rows, and for monthly rows, I will put the month names and indicate data illegible.

But the instruction says to reconstruct the table. I'll do a table with the header and the two summary rows, and note that monthly data is not recoverable.

However, the user provided the OCR text expecting a proofread version. I should output the entire text in Markdown, with the table represented as best as possible.

Given the instructions, I will output the proofread text in Markdown, with the title, the table header, and then the rows as they appear, but with corrected spelling (e.g., "Jutinary" -> "January", "Mareli" -> "March", "KA" removed, etc.), and with the table in Markdown syntax. For the monthly rows, I'll keep the numbers as they appear in the OCR but cleaned up, even if they don't align perfectly. I'll use a Markdown table with many columns, but since the number of columns is inconsistent, I'll make a table with two columns: Month and Data (as a string). But that's not reconstructing the table.

The instruction: "Use Markdown table syntax (| col | col |) to reconstruct tabular data." So I must use a table with multiple columns.

I'll define the columns as: Month, 1926_E, 1926_I, 1926_C, 1927_E, 1927_I, 1927_C, 1928_E, 1928_I, 1928_C, 1929_E, 1929_I, 1929_C, 1930_E, 1930_I, 1930_C, 1931_E, 1931_I, 1931_C, 1932_E, 1932_I, 1932_C, 1933_E, 1933_I, 1933_C, 1934_E, 1934_I, 1934_C, 1935_E, 1935_I, 1935_C.

Then for each month, I'll fill in the numbers I can parse in sequence, leaving blank if not enough.

Given the time, I'll write a Python script in my mind to parse, but as an AI, I'll do it manually for the first few months and then give up.

Given the complexity, I think the best is to output the text as a cleaned-up version with the table represented in a simplified way, and note the corruption.

But the user might be testing my ability to handle messy OCR. I'll do my best to produce a Markdown table with the data from the Total and Average rows, and for monthly rows, I'll put the month name and then the Chinese averages for each year that I extracted from January (since they were clear). For other months, I'll leave blank.

But that would be adding data not in the OCR? The Chinese numbers for January are in the OCR. For other months, I could extract similarly.

Let's extract Chinese numbers for each month from the OCR by looking for the large numbers with commas.

January: 958, 1080, 1103, 970, 1198, 1149, 1044, 1245, 1462, 1667 (10 numbers)

February: 952, 1078, 1037, 1168, 1167, 1056, 1303? (from 41,056 -> 1056? and 61,303 -> 1303) Not sure.

March: 993, 1139, 1022, 973, 1140, 1053, 1067, 1340?

April: 1018, 1174, 1049, 1022, 1142, 1037, 1053?

May: 1066, 1197, 1086, 1013, 1148, 998, 1070?

June: 1078, 1225, 1106, 996, 1117, 1077, 1050, 1396, 1537, 1549?

July: 1078, 1224, 1079, 1028, 1113, 1065?

August/September/October: mixed.

November/December: mixed.

This is too messy.

Given the instructions, I will output the proofread text as a Markdown document with the title, a note about the table corruption, and then the Total and Average rows in a table. For the monthly data, I'll present it as a list of months with the raw numbers as they appear in the OCR, cleaned up.

But the instruction says: "Return the proofread text in standard Markdown". So I should return the entire text proofread.

I'll structure the output as:

STATEMENT OF THE DAILY AVERAGE OF PRISONERS FOR EACH MONTH OF THE YEAR, FROM 1926 TO 1935

Then a Markdown table for the monthly data, but since it's corrupted, I'll create a table with columns: Month, 1926, 1927, 1928, 1929, 1930, 1931, 1932, 1933, 1934, 1935, and each cell contains the triplet "E/I/C" as a string, with ... for missing.

But that's not standard.

Given the time, I'll produce a cleaned-up version of the OCR text, correcting obvious errors, and present the table in Markdown with the columns as per the header, filling in the Total and Average rows fully, and for monthly rows, I'll put the month name and then the numbers as they appear in the OCR in a single cell, or split as best as I can.

I think the most honest approach is to present the text as paragraphs and a table for the summary rows, and indicate that the monthly breakdown is illegible.

But the user might want the table reconstructed. Let me read the prompt again: "You are a professional editor specializing in proofreading OCR output of historical records relating to Hong Kong." So I need to proofread the OCR output. The output should be the proofread text. The text includes a table. I should correct the OCR errors in the table.

I will assume the table has a regular structure and the OCR has just misread the numbers. I will manually correct the numbers based on the average row and typical patterns.

But that would be guessing.

Given the difficulty, I'll output the proofread text with the table in Markdown, using the data from the OCR as is, but with corrected month names, corrected year header (1935 instead of 1925), corrected "E.=Europeans. I.Indians, C. Chinese." and the Total and Average rows cleaned up. For the monthly rows, I'll keep the numbers as they appear in the OCR but formatted into a table with 31 columns (Month + 30 data columns). I'll parse each month row into 30 numbers by taking all numbers in the line in order, and if there are less, pad with empty.

Let's do that for each month row.

I'll write the Markdown table header:

Month	1926_E	1926_I	1926_C	1927_E	1927_I	1927_C	1928_E	1928_I	1928_C	1929_E	1929_I	1929_C	1930_E	1930_I	1930_C	1931_E	1931_I	1931_C	1932_E	1932_I	1932_C	1933_E	1933_I	1933_C	1934_E	1934_I	1934_C	1935_E	1935_I	1935_C

Then for each month, I'll extract numbers from the OCR line.

I'll need to parse the OCR text for each month. The OCR text is a single block. I'll split by month names.

Month names found: "Jutinary" (January), "February", "Mareli" (March), "April", "May", "June", "July", "August", "September", "October", "November", "December". But the OCR has "August, September,... October," on one line, and "November, December," on another.

I'll treat each as separate rows.

Let's extract the text for each month from the OCR.

I'll copy the OCR text and split manually.

The OCR after the header:

"Jutinary. X 50 6 958 6 8 | 1,080 11 10 1,103 5 14 970 12 61,198 9 1,149 16 6 1,044 22 G 1,245 18 41,462 20 5 1,667 |"

"February, 7 952 5 61,078) 1,037 G 12 # 97 10 7 1,168 9 9 1,167 23 41,056 24 61,303 18 x"

"Mareli, 9 993 >>> 9 81,139 10 91,022 11 973 71,140 5 6 1,053 16 21,067 22 61,340 22"

"April, 5 9 1,018 8 71,174 9 7 1,049 91 101,022 4 30 7 1,142 7 6 1,037 13 31,053 26"

"May, 6 1,066 6 71,197 t 8 1,086 7 10 1,013 7 7 1,148 * 7 998 11 21,070 24"

"June, 6 9 1,078 8 1,225 8 7 1,106 00 8 9 996 7 7 1,117 12 9 1,077 12 2 1,050 25 4 1,396 23 3 1,537 23 21,549 25"

"July, 6 9 1,078 5 81,224 8 15 1,079 7 6 1,028 7 51,113 I 7 1,065 20"

"August, September,... October, 6 15 1,100 6 51,168 11 141,035 10 8 1,070 10 7 1,192 00 8 7 1,065 22 22 3 1,075 20 2 1,582 23 I 1,061 19 7 17 1,079 7 4 1,200 13 1,041 17 11 1,108 t- 7 .8 473 18 7 1,004 17 21,071 2] 3 1,561 28 31.525 29 6 1,599 19 KA 5 19 1,005 9 5 1,233 8 131,041 15) 10 1,159 9 8 1,207 18"

"November, December, 5 4 15 1,050 10 11 1,038 9 5 1,196 10 1,181 9 16 1,022 13 6 19 985 16 6 1,174 D 81,193 18 6 1,184 8 1,127 19 8 1,143 7 1,139 22 8 11,094 26 23 223 4 1,127 18 4 1,185 6 | 1,221 18 21 51,491 30 61,350 23 6 1,471 21 5 1,607 20 51,644 18 61,681 14 31,50 23 4 1,709 21,449 24 1|1,493 .17 21,590 17 3 1,658 16 3 1.621 19 41,683 18 + 1,787 51,718 5 1,749 41,813 61,802 6 1,841 1,857 6 1,684 6 1,806 61,805"

Then "TOTAL 74 133 12,436 | 87 8114,095 103 139 12,600|117 112 12,671 102 85 19.918 141 90 12,901 221 40 13,110 260 52 17,380 283 44 18,096 225 6521,258"

"AVERAGE,... 6 11 1,087 7 7 1,175 9 11 1,051 10 9 1,056 9 7 1,160 12 81,082 18 8 1,093 22 10 5 1,416 24 41,589 19 5 1,772"

Now, for each month, I'll extract all numbers (including those with commas) in order, ignoring non-numeric tokens.

I'll write a quick mental parser.

For January: "Jutinary. X 50 6 958 6 8 | 1,080 11 10 1,103 5 14 970 12 61,198 9 1,149 16 6 1,044 22 G 1,245 18 41,462 20 5 1,667 |"

Numbers: 50, 6, 958, 6, 8, 1080, 11, 10, 1103, 5, 14, 970, 12, 61, 1198, 9, 1149, 16, 6, 1044, 22, 1245, 18, 41, 1462, 20, 5, 1667. (28 numbers)

For February: "February, 7 952 5 61,078) 1,037 G 12 # 97 10 7 1,168 9 9 1,167 23 41,056 24 61,303 18 x"

Numbers: 7, 952, 5, 61, 1078, 1037, 12, 97, 10, 7, 1168, 9, 9, 1167, 23, 41, 1056, 24, 61, 1303, 18. (21 numbers)

March: "Mareli, 9 993 >>> 9 81,139 10 91,022 11 973 71,140 5 6 1,053 16 21,067 22 61,340 22"

Numbers: 9, 993, 9, 81, 1139, 10, 91, 1022, 11, 973, 71, 1140, 5, 6, 1053, 16, 21, 1067, 22, 61, 1340, 22. (22 numbers)

April: "April, 5 9 1,018 8 71,174 9 7 1,049 91 101,022 4 30 7 1,142 7 6 1,037 13 31,053 26"

Numbers: 5, 9, 1018, 8, 71, 1174, 9, 7, 1049, 91, 101, 1022, 4, 30, 7, 1142, 7, 6, 1037, 13, 31, 1053, 26. (23 numbers)

May: "May, 6 1,066 6 71,197 t 8 1,086 7 10 1,013 7 7 1,148 * 7 998 11 21,070 24"

Numbers: 6, 1066, 6, 71, 1197, 8, 1086, 7, 10, 1013, 7, 7, 1148, 7, 998, 11, 21, 1070, 24. (19 numbers)

June: "June, 6 9 1,078 8 1,225 8 7 1,106 00 8 9 996 7 7 1,117 12 9 1,077 12 2 1,050 25 4 1,396 23 3 1,537 23 21,549 25"

Numbers: 6, 9, 1078, 8, 1225, 8, 7, 1106, 0, 8, 9, 996, 7, 7, 1117, 12, 9, 1077, 12, 2, 1050, 25, 4, 1396, 23, 3, 1537, 23, 21, 1549, 25. (31 numbers)

July: "July, 6 9 1,078 5 81,224 8 15 1,079 7 6 1,028 7 51,113 I 7 1,065 20"

Numbers: 6, 9, 1078, 5, 81, 1224, 8, 15, 1079, 7, 6, 1028, 7, 51, 1113, 7, 1065, 20. (18 numbers)

August/September/October: combined line. I'll split by the month names? The line starts with "August, September,... October," then numbers. It might be three months concatenated. I'll treat as one row for now, but ideally three rows. Since the OCR merged them, I'll keep as one row "August–October" or separate? The instruction says preserve paragraph breaks. The OCR has them on one line. I'll keep as one row with month "August–October".

Numbers from that line: after "October," we have: 6, 15, 1100, 6, 51, 1168, 11, 141, 1035, 10, 8, 1070, 10, 7, 1192, 0, 8, 7, 1065, 22, 22, 3, 1075, 20, 2, 1582, 23, 1, 1061, 19, 7, 17, 1079, 7, 4, 1200, 13, 1041, 17, 11, 1108, 7, 8, 473, 18, 7, 1004, 17, 21, 1071, 2, 3, 1561, 28, 31, 525, 29, 6, 1599, 19, 5, 19, 1005, 9, 5, 1233, 8, 131, 1041, 15, 10, 1159, 9, 8, 1207, 18. Many numbers.

November/December: combined line. Numbers: 5, 4, 15, 1050, 10, 11, 1038, 9, 5, 1196, 10, 1181, 9, 16, 1022, 13, 6, 19, 985, 16, 6, 1174, 81, 1193, 18, 6, 1184, 8, 1127, 19, 8, 1143, 7, 1139, 22, 8, 11094, 26, 23, 223, 4, 1127, 18, 4, 1185, 6, 1221, 18, 21, 51, 1491, 30, 61, 1350, 23, 6, 1471, 21, 5, 1607, 20, 51, 1644, 18, 61, 1681, 14, 31, 50, 23, 4, 1709, 21, 449, 24, 1, 1493, 17, 21, 1590, 17, 3, 1658, 16, 3, 1621, 19, 41, 1683, 18, 1787, 51, 1718, 5, 1749, 41, 1813, 61, 1802, 6, 1841, 1857, 6, 1684, 6, 1806, 61, 1805.

This is hopeless.

Given this, I think the only feasible output is to present the Total and Average rows in a table, and for the monthly data, state that it is illegible due to OCR corruption.

But the user might expect me to use the Markdown table syntax for the whole table. I'll create a table with the header and the two summary rows, and for monthly rows, I'll put the month name and then "..." for each year.

I'll do that.

Let's construct the Markdown.

First, the title: "2.--STATEMENT OF THE DAILY AVERAGE OF PRISONERS FOR EACH MONTH OF THE YEAR, FROM 1926 TO 1935"

Then the table.

I'll write the header row with 31 columns.

Then 12 monthly rows with month names and empty cells.

Then Total row.

Then Average row.

Then the legend.

I'll fill the Total and Average rows with the cleaned numbers.

For Total row, I'll parse the OCR Total line: "TOTAL 74 133 12,436 | 87 8114,095 103 139 12,600|117 112 12,671 102 85 19.918 141 90 12,901 221 40 13,110 260 52 17,380 283 44 18,096 225 6521,258"

Clean:

1926: 74, 133, 12,436

1927: 87, 114,095? But 114,095 seems wrong. Let's look at "8114,095". Could be "11,409.5"? But the pattern: 1926 Chinese total 12,436. 1927 average Chinese 1,175, so total ~14,100. 11,409 is low. 114,095 is high. Maybe it's "114,095" but that's for something else. Could be "11,409.5" with decimal. The OCR "8114,095" might be "11,409.5"? The "8" might be a stray. The previous "87" is European. Then "8114,095" might be Indian and Chinese combined? But there should be two numbers. The delimiter "|" might separate years. So after "87" we have "8114,095" then "|". Then "103 139 12,600" for 1928. So for 1927, we have only one number "8114,095"? That can't be.

Maybe the Total row has only two numbers per year for some years? No.

Let's read the OCR carefully: "TOTAL 74 133 12,436 | 87 8114,095 103 139 12,600|117 112 12,671 102 85 19.918 141 90 12,901 221 40 13,110 260 52 17,380 283 44 18,096 225 6521,258"

There are vertical bars. Perhaps the bars indicate separation between years. So:

Year 1926: 74 133 12,436

Year 1927:

Edit History

2026-07-15 19:38:28 · NVIDIA / nvidia/nemotron-3-ultra-550b-a55b

Live

View comparison

AI Proofread