Add link to paper

This PR adds a link to the paper "No One-Size-Fits-All: Building Systems For Translation to Bashkir, Kazakh, Kyrgyz, Tatar and Chuvash Using Synthetic And Original Data" (https://huggingface.co/papers/2602.04442) to the dataset card to provide more context for the dataset's creation and usage in machine translation for Turkic languages.

README.md

@@ -1,7 +1,4 @@
 ---
-task_categories:
-- translation
-- text-generation
 language:
 - ru
 - en
@@ -10,27 +7,30 @@ language:
 - kk
 - ky
 - cv
-pretty_name: yaturk-7
 size_categories:
 - 1M<n<10M
+task_categories:
+- translation
+- text-generation
+pretty_name: yaturk-7
 ---
 
-Dataset I have used for the online competition series "Machine Translation for Low-Resource Turkic Languages". This dataset was generated via Yandex.Translate.
-
-Column description:
-
-russian, english, tatar, bashkir, kazakh, kyrgyz, chuvash - phrases on the corresponding languages
-
-source -- huggingface name of dataset from which the phrase was taken (for datasets present on HuggingFace) or url (for datasets not present)
+# yaturk-7
 
-set_from_source - set on the source from which the phrase was taken
+This dataset was used in the research paper [No One-Size-Fits-All: Building Systems For Translation to Bashkir, Kazakh, Kyrgyz, Tatar and Chuvash Using Synthetic And Original Data](https://huggingface.co/papers/2602.04442).
 
-train_on : if 1 , finetune was done on this data sample, otherwise not
+Dataset used for the online competition series "Machine Translation for Low-Resource Turkic Languages". This dataset was generated via Yandex.Translate.
 
-pair -- language pair taken from the original dataset (or the only language if the original dataset was monolingual)
+### Column description:
 
-only_index1: if 1, this data sample was NOT used in the final English-Chuvash data index, otherwise it was used for it. 
+- **russian, english, tatar, bashkir, kazakh, kyrgyz, chuvash**: phrases on the corresponding languages
+- **source**: huggingface name of dataset from which the phrase was taken (for datasets present on HuggingFace) or url (for datasets not present)
+- **set_from_source**: set on the source from which the phrase was taken
+- **train_on**: if 1, finetune was done on this data sample, otherwise not
+- **pair**: language pair taken from the original dataset (or the only language if the original dataset was monolingual)
+- **only_index1**: if 1, this data sample was NOT used in the final English-Chuvash data index, otherwise it was used for it. 
 
-** For samples with only_index1=0, only English, Tatar and Chuvash translations are guaranteed to exist ( Russian also exist if it is from Alex Antonov's Russian-Chuvash corpus)**
+**Notes on data availability:**
 
-** For samples with only_index1=1 ,translations to all 7 languages exist**
+- For samples with `only_index1=0`, only English, Tatar and Chuvash translations are guaranteed to exist (Russian also exist if it is from Alex Antonov's Russian-Chuvash corpus).
+- For samples with `only_index1=1`, translations to all 7 languages exist.