Update README.md

README.md

@@ -18,7 +18,7 @@ size_categories:
 A large **Sámi** text corpus: **172,455 documents / 68.5M words**, covering five written varieties --
 Northern (`sme`), Inari (`smn`), Southern (`sma`), Lule (`smj`), Skolt (`sms`) and Ume (`sju`)
 Sámi. Sámi varieties total 55M words (80%); the remainder is mostly co-occurring
-Norwegian/Finnish/Swedish/English in bilingual texts. The corpus is assembled from a custom web crawl, Common Crawl & Internet Archive recovery, the SIKOR / GiellaLT corpus, and Sámi Wikipedia(s). The texts were extracted from HTML using [DeepSeek V4 Flash](huggingface.co/deepseek-ai/DeepSeek-V4-Flash) (with reasoning enabled) and from PDF files using [Qwen3 VL 235B](https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Instruct).
+Norwegian/Finnish/Swedish/English in bilingual texts. The corpus is assembled from a custom web crawl, Common Crawl & Internet Archive recovery, the SIKOR / GiellaLT corpus, and Sámi Wikipedia(s). The texts were all extracted to Markdown, specifically from HTML using [DeepSeek V4 Flash](huggingface.co/deepseek-ai/DeepSeek-V4-Flash) (with reasoning enabled), and from PDF files using [Qwen3 VL 235B](https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Instruct).
 
 
 ## Language and source distribution