Initial OCR extraction prompt
The following prompt was used to extract the 2019 election results from the uploaded PDFs into the supplied ODS template. The original wording is preserved.
“The pdf file i have uploaded contains the results of the election. Each page contains results for a electoral District. The results are in tables, where four columns exist with the column names Name of the candidate, Party Abbreviation, Votes Received, Percentage. There are 35 candidates. After the candidate results the summary is Total Valid Votes, Rejected Votes, Total Polled, Registered No of Electors as consecutive rows. These summaries also contain counts and percentages. i have uploaded an ods file with candidate names and summaries, but not the counts and percentages. Can you extract the counts and percentages and transfer it to the ods file for each electoral district, it should be one sheet for one electoral district. Can you do that ? Use OCR ?”
The same request was subsequently applied to the remaining district PDFs. Counts and percentages were extracted for all 35 candidates and the four summary rows in each result table. Subsequent validation identified the missing tables, misplaced pages, and source discrepancies documented below.
Preparation scope
This article documents the changes implemented for the 2019 presidential election. It preserves the exact capitalisation of the labels used in that script. Replacement values and source findings have not been independently verified for this documentation.
Supplemented missing results
The narrative documents adding Polling Division K — Kilinochchi, which was omitted from the 12-page Jaffna PDF. The added table contained all 35 candidates and four summary rows and became the thirteenth workbook sheet. The narrative attributes the replacement to the official Election Commission result; the supplied R code imports the completed workbook rather than downloading or performing that replacement.
Replaced misplaced result tables
The following replacements are documented. Page numbers refer to the original PDFs, counting the first page as page 1.
| Uploaded PDF | Page | Supplied result | Required result |
|---|---|---|---|
| Colombo | 2 | Digamadulla postal votes | Colombo postal votes |
| Colombo | 5 | Anuradhapura East | Borella |
| Kegalle | 2 | Kurunegala postal votes | Kegalle postal votes |
| Ratnapura | 10 | Kolonnawa in Colombo | Kolonna in Ratnapura |
The narrative states that the extracted tables were corrected while the original PDFs were retained unchanged.
Preserved registered-elector discrepancies
Printed district totals and component totals were retained and documented rather than forced to agree. Component totals include postal-vote and polling-division elector counts.
| District | Printed district total | Component total | Component minus district |
|---|---|---|---|
| Hambantota | 485,786 | 485,986 | +200 |
| Jaffna | 564,714 | 564,984 | +270 |
| Kalutara | 955,079 | 955,080 | +1 |
| Ratnapura | 864,978 | 864,979 | +1 |
These are registered-elector disagreements, not candidate-vote disagreements. The comparison alone does not identify the incorrect individual count.
Imported and checked the ODS tables
Each sheet is read from A1:D41, excluding notes below
the result table. The first sheet’s name supplies the district name.
Labels are trimmed and internal whitespace condensed without
lowercasing. Counts and percentages are converted to numeric values.
The code stops if a table has other than 35 candidate rows, other than four summary rows, or any missing counts. It also stops on duplicate combinations of year, district, electorate, and row label.
Aligned summary and reporting labels
| Original summary label | Replacement label |
|---|---|
| Total Valid Votes | Total No of Valid Votes |
| Rejected Votes | No of Rejected Votes |
| Total Polled | Total No of Votes Polled |
| Registered No of Electors | No of Registered Electors |
District-total tables are labelled
Final District Results, and postal tables are labelled
Postal District Votes. Polling-division prefixes such as
A - are removed, and spaces in the remaining electorate
names are converted to hyphens.
ODS percentage proportions are multiplied by 100 and rounded to two decimal places, so 0.5949 becomes 59.49. This changes the storage scale, rather than recalculating a percentage from vote counts.
Aligned district names
| Original district label | Replacement label |
|---|---|
| Kandy | Mahanuwara |
| Moneragala | Monaragala |
| Nuwara Eliya | Nuwara-Eliya |
This recoding is applied to the combined Election_Data
without a year restriction. The subsequent electorate mapping is
restricted to 2019 records. Capitalisation is preserved in the explicit
replacement labels.
Aligned electorate names with 2015
The code applies the following 26 district-specific mappings to 2019:
| District | Original electorate label | Replacement label |
|---|---|---|
| Anuradhapura | Horowpothana | Horowapatana |
| Anuradhapura | Madawachchiya | Medawachchiya |
| Anuradhapura | Mihinthale | Mihintale |
| Badulla | Haliela | Hali-Ela |
| Badulla | Haputhale | Haputale |
| Badulla | Mahiyangana | Mahiyangane |
| Badulla | Uvaparanagama | Uva-Paranagama |
| Batticaloa | Kalkudah | Kalkuda |
| Galle | Benthara-Elpitiya | Bentara-Elpitiya |
| Galle | Rathgama | Ratgama |
| Gampaha | Aththanagalla | Attanagalla |
| Gampaha | Meerigama | Mirigama |
| Jaffna | Uduppiddy | Udupiddy |
| Kalutara | Agalawaththa | Agalawatta |
| Kalutara | Bulathsinghala | Bulathsinhala |
| Kalutara | Mathugama | Matugama |
| Kegalle | Aranayaka | Aranayake |
| Mahanuwara | Galgedara | Galagedara |
| Mahanuwara | Harispatthuwa | Harispattuwa |
| Mahanuwara | Pathadumbara | Patha-Dumbara |
| Mahanuwara | Udadumbara | Udu-Dumbara |
| Matale | Ratthota | Rattota |
| Monaragala | Bibila | Bibile |
| Nuwara-Eliya | Kothmale | Kotmale |
| Trincomalee | Muthur | Mutur |
| Trincomalee | Seruvila | Seruwila |
Validation findings and export scope
The narrative reports correcting OCR errors against scans and checking candidate totals against valid votes, valid plus rejected votes against polled votes, and component counts against district totals. Those vote-reconciliation checks are described in prose; the executable 2019 import implements structural and duplicate checks, not those arithmetic assertions.
Presence tables are constructed before and after the name mappings to support inspection across election years. They do not automatically assert that every expected label matches.