Human-Written Myanmar Text Corpus & API
A standardized, curated, and machine-readable Burmese text repository extracted from genuine, human-verified reviews and technical literature. Free for open-source AI training, LLM pre-training, evaluation benchmarks, and linguistic research.
Domain & Category Distribution
Linguistic balance across genres, weighted by rule-based segmented syllables.
1
100% Human-Authored151
Unicode (Mymr)46
Rule-based Tokenized1
Clean SegmentationsCC-BY-4.0
Open Research & CommercialDirect Downloads (One-Click)
Select your preferred format to immediately download the latest corpus dump snapshot.
Standard JSON Array
Full dataset with rich document-level metadata, paragraph splits, and metrics.
JSON Lines
Optimal format for OpenAI, Hugging Face Datasets, and distributed streaming.
Plain Text Corpus
Clean Burmese text without metadata headers. Ready for Byte-Pair (BPE) tokenizers.
Comma-Separated
Ready for spreadsheet inspection, Pandas tabular pipelines, and SQL ingestion.
Myanmar Sentence Corpus
Clean, segmented Burmese sentences split line-by-line (။). Perfect for NLP sentence transformers & translation models.
Tab-Separated
Clean delimiter structure that avoids comma conflicts within complex prose.
API Reference & Direct Endpoints
Access static endpoints programmatically without API keys or rate limits.
Programmatic Ingestion Examples
Copy and execute directly in your Jupyter Notebooks, Google Colab, or pipelines.
import pandas as pd
# Load CSV corpus directly into a DataFrame
df = pd.read_csv("https://articles.datalamhnyin.site/api/corpus.csv")
print(f"Loaded {len(df)} documents")
print(df[["title", "char_count", "published_date"]].head())import requests
import json
response = requests.get("https://articles.datalamhnyin.site/api/corpus.jsonl")
# Stream document entries
for line in response.iter_lines():
if line:
record = json.loads(line.decode("utf-8"))
print("ID:", record["id"])
print("Burmese Text:", record["text"][:100], "...")curl -s "https://articles.datalamhnyin.site/api/corpus.txt" -o myanmar_raw_corpus.txtWhy Human-Crafted Burmese Data Matters
Low-resource language AI systems often suffer from synthetic hallucination loops caused by training on machine-translated or uncurated web-scraped noise. The Data Lam Hnyin corpus initiative provides human-authored, peer-reviewed, and technically aligned Myanmar text.
Every scrapable entry on this platform is verified for standard Myanmar Unicode encoding, linguistic naturalness, and domain accuracy. We encourage developers, researchers, and NLP enthusiasts to utilize this data to benchmark tokenizers, train language models, and evaluate machine translation performance.