wpseven/sms-spam-burmese

SMS Spam Collection – Burmese Translation

A Burmese (Myanmar, Unicode) translation of the UCI SMS Spam Collection v.1, created for spam-detection research in Burmese, where no public SMS spam corpus exists.

Dataset structure

column	description
label	ham (legitimate) or spam
text_my	Burmese translation
text_en	original English message

• 5,574 messages: 4,827 ham, 747 spam (same labels and order as the original)
• Also included: SMSSpamCollection_my — tab-separated label<TAB>text file in the same format as the original corpus

Translation notes

• Machine-translated with an LLM (Claude); not native Burmese SMS.
• Ham was translated in colloquial spoken Burmese; spam in formal written register.
• Phone numbers, short codes, URLs, prices (£), keywords (e.g. CLAIM, STOP) and HTML entities (<#>) were kept unchanged.
• Duplicate messages in the original have identical translations.
• A small number of messages (e.g. "Ok", emoticons, romanized Hindi/Tamil/Malayalam) were left untranslated.

Limitations

• Spam content reflects UK-specific scams from the 2000s and may not represent real Myanmar spam (e.g. mobile-money or betting scams).
• The colloquial-ham / formal-spam register difference may let models learn style rather than content.
• Burmese has no word spaces; use a syllable/word segmenter before bag-of-words or TF-IDF features.
• Contains explicit and offensive language present in the original corpus.

Source & citation

Original corpus by Tiago A. Almeida and José María Gómez Hidalgo (CC BY 4.0):

@inproceedings{almeida2011sms,
  title={Contributions to the Study of SMS Spam Filtering: New Collection and Results},
  author={Almeida, Tiago A. and G{\'o}mez Hidalgo, Jos{\'e} Mar{\'i}a and Yamakami, Akebo},
  booktitle={Proceedings of the 11th ACM Symposium on Document Engineering (DocEng'11)},
  year={2011}
}