Upload README.md with huggingface_hub

README.md

@@ -3,70 +3,125 @@ license: other
 language:
 - km
 - en
+multilinguality: multilingual
+size_categories:
+- 1M<n<10M
 task_categories:
 - text-generation
 - fill-mask
+task_ids:
+- language-modeling
+- masked-language-modeling
 tags:
 - khmer
 - cambodia
 - code-switching
 - sentencepiece
 - diffusion-lm
+- low-resource
 pretty_name: Khmer + English Mixed Text Corpus
+configs:
+- config_name: raw
+  default: true
+  data_files:
+  - split: train
+    path: all_text.txt
+- config_name: segmented
+  data_files:
+  - split: train
+    path: all_text_segmented.txt
 ---
 
 # Khmer + English Mixed Text Corpus
 
-A cleaned, deduplicated Khmer text corpus with naturally-occurring **Khmer/English code-switching**
-(English tech & finance terms, Latin script, and digits embedded in Khmer text). It is the training
-data for the [`Panhapich/khmer-sp-8k`](https://huggingface.co/Panhapich/khmer-sp-8k) SentencePiece tokenizer
-and the text-only warm-start of a shared Khmer diffusion decoder.
+A cleaned, deduplicated Khmer text corpus with naturally-occurring **Khmer/English
+code-switching** (English tech & finance terms, Latin script, and digits embedded in Khmer
+text). It is the training data for the
+[`Panhapich/khmer-sp-8k`](https://huggingface.co/Panhapich/khmer-sp-8k) SentencePiece tokenizer and the
+text-only warm-start of a shared Khmer diffusion decoder.
 
-## Contents
+## Dataset summary
 
-| File | Description |
-|---|---|
-| `all_text.txt` | The corpus. One normalized sentence per line, UTF-8. |
-| `metadata.json` | Corpus statistics and provenance. |
+- **4,893,739 sentences**, one per line, UTF-8, deduplicated and shuffled
+  (fixed seed 42, reproducible).
+- Two parallel versions of the same corpus are provided as separate configs — see
+  [Dataset structure](#dataset-structure).
+- Khmer/English mixing is **entirely organic**: it comes from the real sources below (receipt
+  line items, tech/finance terms in headlines and raw text). No synthetic code-switching is
+  injected anywhere in this corpus.
 
-- **Total sentences:** 4,893,739
+## Dataset structure
 
-## Sources
+### Configs / files
 
-Built from exactly four sources, mixed together and shuffled. A global cap
-(`5,000,000` sentences) bounds the total — small sources and the local file are taken
-in full, and the large raw-text source fills the remainder:
-
-| Source | Field | Contribution |
+| Config | File | Description |
 |---|---|---|
-| `Sokheng/khmer-synthetic-ocr-v1-100k` | `text` | Synthetic OCR receipt/label text (Khmer/English/digit mixed) |
-| `rinabuoy/aupp-assignment-data` | `title` | Khmer news headlines |
-| `nphearum/khmer-raw-text-3M-v2` | `text` | Large raw Khmer text (split on the Khmer full stop `។`) |
-| `khmer_corpus.txt` (local) | line | Pre-chunked ~1000-character Khmer text blocks |
+| `raw` (default) | `all_text.txt` | Natural Khmer text, no artificial word-boundary spaces. General-purpose use. |
+| `segmented` | `all_text_segmented.txt` | Same 4,893,739 sentences, run through `khmer-nltk` word segmentation + gazetteer/Latin masking (see [`Panhapich/khmer-sp-8k`](https://huggingface.co/Panhapich/khmer-sp-8k)). Ready for SentencePiece-style subword training without re-running segmentation yourself. |
+
+```python
+from datasets import load_dataset
+
+raw = load_dataset("Panhapich/khmer-text-corpus", "raw")             # or omit the config name, it's the default
+segmented = load_dataset("Panhapich/khmer-text-corpus", "segmented")
+```
+
+### Data instances
+
+Each row is a single field, `text`: one normalized sentence (or, in the `segmented` config,
+one word-segmented sentence with explicit spaces between Khmer words).
+
+### Data splits
+
+A single `train` split — this is raw pretraining/tokenizer-training text, not a supervised
+task with held-out evaluation labels.
+
+## Dataset creation
+
+### Source data
 
-**The Khmer/English mixing is entirely organic** — it comes from the real sources. No synthetic
-code-switching is injected.
+Built from exactly four sources, mixed together, deduplicated, and shuffled. A global cap
+(`5,000,000` sentences) bounds the total — small sources and the local file are
+taken in full, and the large raw-text source fills the remainder:
 
-## Preprocessing
+| Source | Field | Description | Collected (pre-dedup) | Share |
+|---|---|---|---|---|
+| `nphearum/khmer-raw-text-3M-v2` | text | Large raw Khmer text (split on the Khmer full stop `។`) | 4,836,355 | 96.7% |
+| `Sokheng/khmer-synthetic-ocr-v1-100k` | text | Synthetic OCR receipt/label text (Khmer/English/digit mixed) | 93,293 | 1.9% |
+| `khmer_corpus.txt` | line | Pre-chunked ~1000-character Khmer text blocks (local file) | 35,373 | 0.7% |
+| `rinabuoy/aupp-assignment-data` | title | Khmer news headlines | 34,991 | 0.7% |
 
-Every sentence is normalized: coerce to string, strip, replace newlines/tabs with a single space,
-collapse repeated whitespace, and drop empty/invalid labels (`???`, `NULL`, `N/A`, `UNKNOWN`).
-Khmer word-spacing, punctuation, Khmer + Arabic numerals, and English/Latin tokens are preserved.
+**The Khmer/English mixing is entirely organic** — it comes from the real sources above. No
+synthetic code-switching is injected.
 
-Hub-source sentences are kept only if 5-400 characters; the local blocks are exempt from the length
-cap. The corpus is deduplicated and shuffled with a fixed seed (42) for reproducibility.
+### Curation / preprocessing
 
-## Intended use
+Every sentence is normalized: coerce to string, strip, replace newlines/tabs with a single
+space, collapse repeated whitespace, and drop empty/invalid labels (`???`, `NULL`, `N/A`,
+`UNKNOWN`). Khmer word-spacing, punctuation, Khmer + Arabic numerals, and English/Latin tokens
+are preserved. Hub-source sentences are kept only if 5-400 characters; the local blocks are
+exempt from the length cap. The corpus is deduplicated and shuffled with a fixed seed (42) for
+reproducibility.
 
-Subword tokenizer training for Khmer with code-switched English, text-only diffusion/masked-LM
-pretraining of a Khmer decoder, and general mixed-script Khmer NLP.
+The `segmented` config additionally runs `khmer-nltk` word segmentation, with English/digit
+spans and known loanwords/acronyms (ATM, ABA, PDF, ...) masked out beforehand so the segmenter
+doesn't shatter them, and glued English/Latin runs (e.g. `tryAImodel`) further decomposed via
+frequency-based word segmentation, guarded by a domain exception list (ACLEDA, Bakong, COVID-19,
+5G, ...) so real terms aren't mangled. Full pipeline: `khmer_segmentation.py` in
+[`Panhapich/khmer-sp-8k`](https://huggingface.co/Panhapich/khmer-sp-8k).
 
-## Out of scope / limitations
+## Considerations for using the data
 
-Plain text only (no labels or translations). Includes synthetic OCR text and raw web-scraped text,
-so expect some noise and non-standard spelling. Not filtered for PII beyond the rules above.
+- **Known noise**: the local `khmer_corpus.txt` source and the raw-web-scraped Hub source
+  contain some non-standard spelling, stray symbols, and OCR artifacts. This is real-world
+  scraped text, not a curated, edited corpus.
+- **Not filtered for PII** beyond the basic invalid-label rules above.
+- **Plain text only** — no labels, translations, or task annotations.
+- **Segmentation is a statistical CRF model**, not a hand-built rule system — it generalizes
+  well to ordinary Khmer but was not evaluated line-by-line against this specific corpus; treat
+  the `segmented` config as "reduces cross-word token fusion," not "hand-verified perfect."
 
 ## Licensing
 
-This corpus is derived from third-party datasets and inherits their licenses. Confirm the terms of
-each source above and set the `license` field accordingly — `other` is a placeholder.
+This corpus is derived from third-party datasets and inherits their licenses. Confirm the terms
+of each source above and set the `license` field accordingly — `other` is a placeholder.