Update README.md

README.md

@@ -7,4 +7,54 @@ language:
 pretty_name: Custom CNN/Daily Mail Summarization Dataset
 size_categories:
 - n<1K
----
+---
+# Dataset Card for Custom Text Dataset
+
+## Dataset Name
+Custom CNN/Daily Mail Summarization Dataset
+
+## Overview
+This dataset is a custom version of the CNN/Daily Mail dataset, designed for text summarization tasks. It contains news articles and their corresponding summaries.
+
+## Composition
+The dataset consists of two splits:
+- Train: 1 custom example
+- Test: 100 examples from the original CNN/Daily Mail dataset
+
+Each example contains:
+- 'sentence': The full text of a news article
+- 'labels': The summary of the article
+
+## Collection Process
+The training data is a custom example created manually, while the test data is sampled from the CNN/Daily Mail dataset (version 3.0.0) available on Hugging Face.
+
+## Preprocessing
+No specific preprocessing was applied beyond the original CNN/Daily Mail dataset preprocessing.
+
+## How to Use
+```python
+from datasets import load_from_disk
+
+# Load the dataset
+dataset = load_from_disk("./results/custom_dataset/")
+
+# Access the data
+train_data = dataset['train']
+test_data = dataset['test']
+
+# Example usage
+print(train_data['sentence'])
+print(train_data['labels'])
+```
+
+## Evaluation
+This dataset is intended for text summarization tasks. Common evaluation metrics include ROUGE scores, which measure the overlap between generated summaries and reference summaries.
+
+## Limitations
+- The training set is extremely small (1 example), which may limit its usefulness for model training.
+- The test set is a subset of the original CNN/Daily Mail dataset, which may not represent the full diversity of news articles.
+
+## Ethical Considerations
+- The dataset contains news articles, which may include sensitive or biased content.
+- Users should be aware of potential copyright issues when using news content for model training or deployment.
+- Care should be taken to avoid generating or propagating misleading or false information when using models trained on this dataset.