its-autumn/burmese-wiktionary

Burmese Wiktionary Cleaned Lexicon

A systematically parsed, cleaned, and structured lexical dataset extracted from Myanmar Wiktionary (my.wiktionary.org). This resource is tailored specifically for computational linguistics, corpus analysis, and Natural Language Processing (NLP) tasks targeting Burmese—a low-resource language with limited structured lexical availability.

---

Dataset Overview & Provenance

• Source: Myanmar Wiktionary (my.wiktionary.org)
• Data Snapshot / Dump Date: 2026-08-01 (mywiktionary-20260801-pages-articles.xml.bz2)
• Data Integrity & As-Is Fidelity: All textual definitions, lexical entries, grammatical labels, and transliterations directly mirror the source content in my.wiktionary.org as of the snapshot date. No synthetic definitions or lexical alterations were introduced; processing was strictly constrained to structural extraction, Wikitext markup stripping, HTML artifact normalization, and relational schema normalization.
• License: Distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Users are free to share, adapt, transform, and build upon this material for any purpose, including commercial applications, provided appropriate attribution is given.

---

Technical Architecture & Schema

Each entry in the dataset represents a distinct Word–POS tuple. Words containing multiple parts of speech are split into individual, discrete rows to ensure clean downstream classification labels.

Data Fields

Field	Type	Description
word	string	Primary Burmese headword (lexical entry) in Unicode.
mlcts	string	Standard MLC (Myanmar Language Commission) Transcription System Romanization. Complete (100%) coverage across all records.
ipa	string or null	International Phonetic Alphabet representation, when available in the raw entry.
pos	string	Myanmar grammatical category / Part of Speech (e.g., နာမ်, ကြိယာ, နာမဝိသေသန).
definitions	list[string]	Array of clean dictionary definitions stripped of Wiki markups, links, and HTML entities.
classifier	string or null	Numerical classifier associated with the entity. Grammatically constrained strictly to entities where pos == "နာမ်" (Nouns). Set to null for all other syntactic classes.

---

Dataset Summary & Statistics

Corpus Metrics

Metric	Count
Total Cleaned Records	22,191
Unique Headwords (Words)	20,759
Total Definitions	28,402
Avg. Definitions per Record	1.28
Avg. Definition Length	42.0 characters
Max Definitions on a Single Word	29 (ထိုး)

Part of Speech (POS) Distribution

Part of Speech (POS)	Count	Percentage
နာမ် (Noun)	12,878	58.03%
ကြိယာ (Verb)	7,845	35.35%
နာမဝိသေသန (Adjective)	907	4.09%
ဝါစင်္ဂ (Particle / Functional)	375	1.69%
သမ္ဗန္ဓ (Conjunction)	107	0.48%
ဝိဘတ် (Postposition)	61	0.27%
အာမေဍိတ် (Interjection)	15	0.07%
ပစ္စည်း (Affix / Enclitic)	3	0.01%

Metadata Field Coverage

Field	Available Count	Coverage (%)	Description / Constraints
word	22,191	100.00%	Primary lexical headword
mlcts	22,191	100.00%	Standard MLC Romanization
definitions	22,190	100.00%	Clean, normalized definitions list
classifier	7,821	35.24%	Syntactically constrained to Nouns only
ipa	113	0.51%	IPA pronunciation phonetic strings

---

Intended Use Cases

• Part-of-Speech (POS) Tagging & Parsing: Provides high-density lexical ground truth for training sequence-labeling models (CRFs, BiLSTM-CRF, Transformer-based taggers).
• Noun-Classifier Semantics: Benchmark dataset for evaluating lexical dependency and noun-to-classifier selection tasks in Burmese.
• Grapheme-to-Phoneme (G2P) & Transcription: High-coverage parallel corpus for training grapheme-to-MLCTS transliteration engines.
• Lexical Sense Disambiguation & Tokenization: Grounding dictionary entries for word boundary disambiguation and dictionary-based segmenters.

---

Acknowledgements & Technical Credits

Special thanks to Khant Sint Heinn (Kalix Louis) for technical assistance and engineering contributions toward developing the extraction scripts, streaming parsers, and data cleaning workflows used in this project.

---

Citation & Attribution

If you utilize this dataset in research, tool development, or technical benchmarking, please attribute it as follows:

@dataset{burmese_wiktionary_lexicon_2026,
  author       = {Autumn and Khant Sint Heinn},
  title        = {Burmese Wiktionary Cleaned Lexicon},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {https://huggingface.co/datasets/its-autumn/burmese-wiktionary},
  note         = {Extracted and normalized from Myanmar Wiktionary dump dated 2026-08-01 under CC BY 4.0}
}