Purposed changes:

• use init.sql script during spinning container up via docker compose up -d
• use copy SQL command to load CSV data to declared table. This will serve a use case of Import CSV data with a single command

This module allow you to start a local database, import CSV data to it and run query to look for more insignts.

Feature added:

• [x] Starting local DB
• [x] Connect to local db with a single command
• [x] Import CSV data with a single command

I will convert this to a real PR when the import script is done.

This is an update on coordinates approach and its results. I edited original OCR json since there are many misplace coordinates and missing text.

Still in the middle of fixing the result in 3.15 - 3.17 files but I wanted to share the results as soon as possible so others can review it.

As the format of ITEM_ID is xxxx.x.x.xx, can you make it clear in the metadata which section refers to REF_DOC and which one refer to RUNNING_NO? Or maybe clarify each section means if possible. Thx.

I translated most of the character columns to English using Google Translation API and uploaded the translated data to Google Sheet. Feel free to merge if you think this might be useful to others as well. :)

MANY thanks to the team for all the hard work! |||

Do you think we can programmatically translate the data to English too? :)

@nitikornbunya Maybe with LONGDO Dict? :)

PDF is not really a programming friendly format. HTML would make it easier to parse data and would let more people with different expertise in different programming language to take part.

I'm using this library to do the parsing: https://github.com/pdf2htmlEX/pdf2htmlEX

It is quite simple to use as they provide the docker image. See example here

There is a minor unicode issue that shifted า to ำ in some places but it is deterministic and should be easily fixed.

Purposed changes:

• Use init.sql script during spinning container up via docker compose up -d. That script will do ETL from CSV file to containerized Postgres database.
• Make output_example_v2.csv more machine-friendly by removing non-realistic lines

Note

Even though this PR is similar to #20 , there were no real change applied. So this PR aims for change in main branch.

Add references to visualizations powered by this dataset in README.md including:

• Thanawit Prasongpongchai's https://taepras.github.io/thaibudget65/#/
• korlan rayong's https://public.tableau.com/app/profile/korlan.rayong2953/viz/OverviewBudget65/Dashboard1

Removed single Thai char that are mistakes from OCR.

โครงการพัฒนาครูผู้สอนวิทยาศาสตร์ ข ข คณิตศาสตร์ และ เทคโนโลยี to โครงการพัฒนาครูผู้สอนวิทยาศาสตร์ คณิตศาสตร์ และ เทคโนโลยี.

โครงการการพัฒนาอุตสาหกรรมการท่องเที่ยวอย่างยั่งยืน ญ 4 to โครงการการพัฒนาอุตสาหกรรมการท่องเที่ยวอย่างยั่งยืน 4

Fixed bug in first document. Added log message as #14 reported.

The leaf node will be item description of each entry. example

1. งบเงินอุดหนุน 47,034,100 บาท
    1.1 เงินอุดหนุนทั่วไป 38,887,100 บาท
        1) ค่าใช้จ่ายบุคลากร 918,500 บาท
        2) ค่าใช้จ่ายดำเนินงาน 23,870,700 บาท
        3) เงินอุดหนุนดำเนินการตามอำนาจหน้าที่และภารกิจถ่ายโอน 14,097,900 บาท
    1.2 เงินอุดหนุนเฉพาะกิจ 8,147,000 บาท

result will be |CATEGORY_LV1|CATEGORY_LV2|CATEGORY_LV3|...|ITEM_DESCRIPTION|FISCAL_YEAR|AMOUNT| |---|---|---|---|---|---|---| |งบเงินอุดหนุน|เงินอุดหนุนทั่วไป|| |ค่าใช้จ่ายบุคลากร|2022|918,500 |งบเงินอุดหนุน|เงินอุดหนุนทั่วไป|| |ค่าใช้จ่ายดำเนินงาน|2022|23,870,700 |งบเงินอุดหนุน|เงินอุดหนุนทั่วไป|| |เงินอุดหนุนดำเนินการตามอำนาจหน้าที่และภารกิจถ่ายโอน|2022|14,097,900 |งบเงินอุดหนุน||||เงินอุดหนุนเฉพาะกิจ |2022|8,147,000

Hello there, According to README.md, it said "หมวดงบรายจ่าย level-1 จะประกอบไปด้วย งบบุคลากร, งบดำเนินงาน, งบลงทุน, งบเงินอุดหนุน, งบรายจ่ายอื่น เท่านั้น" for CATEGORY_LV1 But this is what I got when using python pandas

df.groupby(["CATEGORY_LV1"])[["ITEM_ID"]].agg(["count"]) 

So, I would like to state the issues here and hope that it will be fixed soon

While the team and the kaogeek community are working together to make the data more accurate, I would like to continue to add more value to the data.

Would be great if we can extract all addresses from the data, and MAYBE even geocode them. For geocoding, we can use Google Geocoding API (again) to do this. Happy to contribute, but I can't promise when. :)

Hi,

Thank you very much for your effort!

I've played with field (on Google Spreadsheet) a bit and found that there are some issues on parsing AMOUNT. Many values there don't look correct (in numerical sense). You can try to reproduce the screenshot by sorting AMOUNT.

Because the information in the file is quite important, we really need to make sure that these issue get fixed asap before anyone would try to use it and compute statistics from the data.