Create README.md

README.md

@@ -0,0 +1,13 @@
+---
+language:
+- vi
+size_categories:
+- 10M<n<100M
+---
+
+This is a filtered subset of the `triples.train.ids.small.tsv` file from the [unicamp-dl/mmarco](https://huggingface.co/datasets/unicamp-dl/mmarco) dataset.
+
+## Selection
+
+Filtered using a custom-trained crossencoder.
+Each query-positive pair in the original `triples.train.ids.small.tsv` was scored, and only pairs with a score greater than **0.5** were retained.