Update description

README.md

@@ -37,4 +37,9 @@ language:
 pretty_name: (Subset of) Aya dataset translated to Bengali
 size_categories:
 - 1K<n<10K
----
+---
+
+
+`aya_dataset_ben_translated` is a subset of the [aya_dataset](https://huggingface.co/datasets/CohereForAI/aya_dataset), with some modifications. In particular, the original data points in Bengali (indicated by the `language` or `language_code` columns) are retained. In addition, the English and Hindi data points are translated into Bengali using Google Cloud Translation API. All columns from the original dataset are retained.
+
+A handful of inaccuracies arising out of translation have been fixed so far. Therefore, the dataset can be a bit noisy. This is particularly true for coding related questions and answers. Moreover, some non-Bengali characters can be found in the text. In addition, potential duplicates from the original dataset are retained as well.