A cleaned Nepali-language text corpus extracted from official Ministry of Finance (MoF), Nepal ministry-wise progress reports published on mof.gov.np. The corpus spans fiscal years 2072/73 – 2080/81 (approximately 2015–2024), covering budget implementation, ministry-level expenditure, program progress, and financial reporting across all government ministries of Nepal.

Each row is a self-contained chunk of Nepali text (~300–1200 characters), filtered from mixed Nepali/English content, cleaned of OCR artifacts, and annotated with metadata including fiscal year, source filename, and content type.

Dataset Structure

Splits

Split	Rows
train	~8,000
validation	~1,000
test	~1,000
total	~10,000

Exact counts depend on MIN_CHARS/MAX_CHARS settings at generation time.

Fields

Field	Type	Description
`text`	string	Cleaned Nepali text chunk (300–1200 chars)
`id`	string	Unique chunk ID — format: `mof-{source_id}-{chunk_id}`
`chunk_global_id`	int	Global chunk index across all sources
`chunk_local_id`	int	Chunk index within its source file
`source_id`	int	Source block number
`source_filename`	string	Inner PDF/source filename
`outer_file`	string	Outer merged file this source came from
`fiscal_year`	string	Fiscal year extracted from filename (e.g. `2073-74`)
`char_count`	int	Character count of this chunk
`nepali_char_ratio`	float	Fraction of Devanagari characters (0.0–1.0)
`content_type`	string	Semantic content category (see below)
`language`	string	ISO 639-1: `ne` (Nepali)
`script`	string	ISO 15924: `Deva` (Devanagari)
`domain`	string	`government_finance`
`country`	string	`NP` (Nepal)
`organization`	string	`Ministry of Finance, Nepal (MoF)`
`document_type`	string	`ministry_progress_report`
`license`	string	`public_domain_gov_np`
`dataset_version`	string	`1.0`
`created_date`	string	Date dataset was generated (YYYY-MM-DD)

Content Types

Value	Description
`report_narrative`	Main body text of progress reports
`policy_text`	Budget policy, legal, or regulatory passages
`table_data`	Numeric, expenditure, or tabular content
`table_of_contents`	TOC sections with page references
`abbreviations`	Abbreviation or glossary lists
`appendix`	Annexure / appendix / schedule sections
`other`	Miscellaneous short passages

Source Documents

Fiscal Year	Document
2072/73	मन्त्रालयगत प्रगति विवरण (Ministry-wise Progress Report)
2073/74	Ministrywise Progress 2073-74
2074/75	प्रगति विवरण सम्पादन
2075/76	मन्त्रालयगत प्रगति विवरण २०७५
2076/77	मन्त्रालयगत प्रगति 2076
2077/78	मन्त्रालयगत प्रगति विवरण_2077
2078/79	मन्त्रालयगत प्रगति विवरण २०७८-७९
2079/80	मन्त्रालयगत प्रगति विवरण २०७८
2080/81	MoF Annual Report 2080.81
Various	एकीकृत वित्तीय विवरण (NPSAS-based Integrated Financial Report)

Data Collection & Processing

Source

All documents were sourced from mof.gov.np — the official website of the Ministry of Finance, Government of Nepal. Documents are publicly available government publications. The raw text was pre-extracted from PDFs and merged into a single .txt file with FILE: / फाइल: section headers.

File Structure

The source file has a nested structure handled automatically by the parser:

FILE: merged_nepalimof1.txt          ← outer wrapper file
  फाइल: actualpdf1.txt              ← inner source (parsed as individual block)
  फाइल: actualpdf2.txt              ← inner source
FILE: MoF Annual Report 2080.81.txt  ← direct content block

Each inner block or direct content block becomes a separate source in the dataset, traceable via source_id, source_filename, and outer_file.

Cleaning Pipeline

Step	What was removed / fixed
English-only lines	Lines with zero Nepali characters and >5 chars (tables of contents in English, headings, footnotes)
`[Page N]` markers	Extractor-inserted page labels
`(cid:N)` sequences	Corrupt PDF glyph encoding artifacts
TOC dot leaders	Runs of 4+ dots collapsed to `…`
Unicode replacement char `U+FFFD`	Bad bytes from encoding errors
Private-use area chars `U+E000–U+F8FF`	PDF internal glyph references
Box-drawing chars `─`	Separator line artifacts
Cedilla `¸`	OCR noise character
OCR space injection — Rule 1	Space inserted before a vowel sign (matra); matras cannot start a Nepali word. Example: `सशु ासि` → `सशुासि`
OCR space injection — Rule 2	Matra + space + 1–2 char Devanagari fragment = mid-word OCR split. Example: `कायहि रू` → `कायहिरू`
Double spaces	Collapsed to single space
NFC normalization	Unicode canonical composition applied

Chunking Strategy

Lines filtered for Nepali content, then grouped into paragraphs
Paragraphs merged until chunk reaches 300–1200 characters
Any chunk where less than 30% of characters are Devanagari is discarded
Very short final segments are appended to the previous chunk

Note: The Nepali threshold is set slightly lower (30%) than the MoHA/Law Commission datasets because MoF reports mix Devanagari numerals, budget codes, and some English ministry acronyms within otherwise Nepali sentences.

Example Row

{
  "text": "नेपालको संविधानको धारा ११९ को उपधारा (२) मा प्रत्येक आर्थिक वर्षको वार्षिक अनुमान व्यवस्थापिका संसद समक्ष पेश गर्दा अघिल्लो आर्थिक वर्षमा प्रत्येक मन्त्रालयलाई छुट्याइएको रकम र त्यस्तो खर्च अनुसारको लक्ष्य हासिल भयो वा भएन त्यसको विवरण पनि साथै पेश गर्नुपर्ने व्यवस्था रहेको छ।",
  "id": "mof-002-0001",
  "chunk_global_id": 12,
  "chunk_local_id": 1,
  "source_id": 2,
  "source_filename": "Ministrywise Progress 2073-74_20170530090116.pdf",
  "outer_file": "merged_nepalimof1 (1).txt",
  "fiscal_year": "2073-74",
  "char_count": 298,
  "nepali_char_ratio": 0.8523,
  "content_type": "policy_text",
  "language": "ne",
  "script": "Deva",
  "domain": "government_finance",
  "country": "NP",
  "organization": "Ministry of Finance, Nepal (MoF)",
  "document_type": "ministry_progress_report",
  "license": "public_domain_gov_np",
  "dataset_version": "1.0",
  "created_date": "2025-04-26"
}

Comparison with Related Nepali Government Datasets

	MoHA Nepal	Law Commission	MoF Nepal
Organization	Ministry of Home Affairs	Nepal Law Commission	Ministry of Finance
Domain	`government`	`legal`	`government_finance`
Topic	Security, administration	Law reform, legislation	Budget, expenditure, fiscal policy
Fiscal years	2076/77–2082/83	2067/68–2081/82	2072/73–2080/81
English mixing	Minimal	Moderate	Moderate (budget codes, acronyms)
Nepali threshold	35%	35%	30%

Intended Uses

Financial NLP — Nepali budget and fiscal policy text understanding
Language model pre-training — Nepali government finance domain corpus
Named entity recognition — Ministry names, program names, fiscal year detection in Nepali
Summarization — Ministry-wise progress summarization in Nepali
Low-resource NLP — Nepali financial/government language is extremely under-resourced
Cross-lingual research — paired with English MoF documents for translation tasks

Limitations

OCR quality varies by year — documents from 2072/73 and 2073/74 have higher noise levels due to older scanning
Budget tables (table_data chunks) may have incomplete rows due to PDF multi-column extraction
Some English budget codes and acronyms (e.g. MTEF, PEFA, NPSAS) are retained within otherwise Nepali sentences
Fiscal year detection from filenames is heuristic — some chunks may show (unknown) fiscal year
The corpus does not include the actual budget law (Appropriation Act) text, only progress reports

License

All source documents are official publications of the Ministry of Finance, Government of Nepal, and are in the public domain under Nepal's government open publication policy. Released for research and educational use.

Citation

@dataset{mof_nepal_nepali_2025,
  title     = {Ministry of Finance Nepal -- Nepali Government Finance Corpus},
  author    = {Ministry of Finance, Nepal (source) and dataset contributors},
  year      = {2025},
  url       = {https://huggingface.co/datasets/YOUR_USERNAME/mof-nepal-nepali},
  note      = {Extracted and cleaned from mof.gov.np official ministry-wise progress reports}
}

Contact

For questions about this dataset, open a discussion on the dataset page. For questions about the source documents, visit mof.gov.np.

Downloads last month: 15

Total file size:

6 MB