Update README.md

README.md

@@ -23,3 +23,37 @@ configs:
   - split: train
     path: data/train-*
 ---
+# mlqa filtered version
+
+For a better dataset description, please visit the official site of the source dataset: [LINK](https://huggingface.co/datasets/mlqa) <br>
+<br>
+**This dataset was prepared by converting mlqa dataset**. I've concatenated versions of the dataset for languages of interest and retrieved a text answers from "answers" column. 
+
+**I additionaly share the code which I used to convert the original dataset to make everything more clear**
+```
+def download_mlqa(subset_name):
+    dataset_valid = load_dataset("mlqa", subset_name, split="validation").to_pandas()
+    dataset_test  = load_dataset("mlqa", subset_name, split="test").to_pandas()
+    full_dataset = pd.concat([dataset_valid, dataset_test])
+    full_dataset.reset_index(drop=True, inplace=True)
+    return full_dataset
+
+needed_langs = ["mlqa.en.en", "mlqa.de.de", "mlqa.ar.ar", "mlqa.es.es", "mlqa.vi.vi", "mlqa.zh.zh"]
+datasets = []
+for lang in tqdm(needed_langs):
+    dataset = download_mlqa(lang)
+    dataset["lang"] = lang.split(".")[2]
+    datasets.append(dataset)
+    
+full_mlqa = pd.concat(datasets)
+full_mlqa.reset_index(drop=True, inplace=True)
+full_mlqa["answer"] = [answer_dict["text"][0] for answer_dict in full_mlqa["answers"]]
+full_mlqa.drop("answers", axis=1, inplace=True)
+```
+
+**How to download**
+
+```
+from datasets import load_dataset
+data = load_dataset("dkoterwa/mlqa_filtered")
+```