Changed some dataset card phrasings for consistency.

README.md

@@ -5,6 +5,11 @@ dataset_info:
     dtype: string
   - name: label
     dtype: string
+  splits:
+  - name: train
+    num_examples: 8175
+  - name: test
+    num_examples: 2725
 license: apache-2.0
 task_categories:
 - text-classification
@@ -20,22 +25,23 @@ source_datasets:
 language_creators:
 - found
 annotations_creators:
-- unknown
+- found
 size_categories:
 - 10K<n<100K
 ---
 
 # Dataset Card for Email Classification Dataset
 
-The Email Classification Dataset contains **10,900 English emails** published as a binary spam/not_spam dataset. 
+The Email Classification Dataset contains **10,900 English texts** labelled as `spam` or `not_spam`. 
+It is used as training and evaluation data for the spam detection component of the MLOps course project at FIB-UPC.
 
 ## Dataset Details
 
 ### Dataset Description
 
-This dataset contains 10,900 English emails for text classification tasks. 
+This dataset contains 10,900 English messages for text classification tasks. 
 The dataset is split into train (8,175 rows, 75%) and test (2,725 rows, 25%) splits. It 
-contains two columns: `text` (full email body) and `label` (binary label [spam, not_spam]).
+contains two columns: `text` (full message content) and `label` (binary label [spam, not_spam]).
 
 - **Curated by:** FIB-UPC MDS students, see [Dataset Card Contact](#dataset-card-contact). Source data by [Deysi](https://huggingface.co/datasets/Deysi/spam-detection-dataset).
 - **Language(s) (NLP):** English
@@ -51,8 +57,8 @@ Due to the original dataset lacking a dataset card, the sources that originated
 
 ### Direct Use
 
-- **Binary Spam/Ham Email Classification:** Use the `label` column for traditional spam detection tasks.
-- **Raw Text for LLM Training:** The `text` column provides raw English email content suitable for language model pre-training or fine-tuning.
+- **Binary spam classification:** The `label` column supports training and evaluating models that distinguish spam from legitimate (ham) messages.
+- **Raw Text for LLM Training:** The `text` column provides raw English messages content suitable for language model pre-training or fine-tuning.
 
 ### Out-of-Scope Use
 
@@ -67,12 +73,12 @@ Each instance contains the following fields:
 
 | Field | Type | Description |
 |-------|------|-------------|
-| `text` | `string` | Full email body as plain text |
+| `text` | `string` | Full message text |
 | `label` | `string` | Binary label `spam` or `not_spam` |
 
 Example:
 
-```python
+```json
 {
   "text": "Get rich quick! Make millions in just days with our new and revolutionary system! Don't miss out on this amazing opportunity!",
   "label": "spam"
@@ -87,13 +93,24 @@ Example:
 | `test` | 2,725 | 25% |
 | **Total** | **10,900** | **100%** |
 
+The split was inherited unchanged from the source dataset.
+
+### Label Distribution
+
+| Label | Train | Test | Total |
+|-------|-------|------|-------|
+| `spam` | TODO | TODO | TODO |
+| `not_spam` | TODO | TODO | TODO |
+
+Text length ranges from TODO to approximately TODO characters.
+
 ## Dataset Creation
 
 ### Curation Rationale
 
-The dataset was originally published as a binary spam detection dataset without a dataset card. 
-As part of the MLOPS course in the Master's in Data Science at FIB-UPC, a dataset card was created to 
-document it properly.
+The dataset was selected to train the spam detection component of the MLOps course project in the Master in Data Science at FIB-UPC. 
+It provides binary spam labels in English with a predefined train and test split. The source dataset was published without documentation. 
+It was therefore republished under the team's account with a dataset card that documents its content, limitations and intended use within the project.
 
 ### Source Data
 
@@ -104,7 +121,7 @@ containing 10,900 English emails with binary spam/not_spam labels.
 
 #### Who are the source data producers?
 
-The original binary data was sourced from Deysi and published on Hugging Face. Information from the original dataset's sources is not clear.
+The source dataset was published on Hugging Face by the user Deysi. Its dataset card provides no information about the origin of the data.
 
 ### Annotations
 
@@ -114,16 +131,16 @@ The annotation process for the binary labels is unknown.
 
 #### Who are the annotators?
 
-Unknown. Original method not documented.
+The annotators of the binary labels are not documented in the source dataset.
 
-#### Personal and Sensitive Information
+### Personal and Sensitive Information
 
-Contains full email bodies. Anonymization not documented. Assume Personally Identifiable Information present, such as names and addresses. 
-No PII audit was performed. Use with care. Do not try to identify individuals.
+Anonymization not documented in the original dataset. However, the `not_spam` instances seem to be public forum posts. Some of them contain URLs, organisation names and details that authors disclosed about themselves, such as their employer or field of study. 
+The `spam` instances appear to be synthetic and are not expected to contain real personal data. No anonymisation or PII audit was performed on the texts.
 
 ## Bias, Risks, and Limitations
 
-- **Synthetic spam style.** Original spam texts contain exaggerated, promotional language that may not reflect real-world phishing tactics. Models trained on this dataset may not generalize to real-life spam.
+- **Synthetic spam style.** Spam texts contain exaggerated, promotional language that may not reflect real-world phishing tactics. Models trained on this dataset may not generalize to real-life spam.
 - **Temporal drift.** Dataset likely reflects spam patterns from its original (and unknown) collection period. Current spam tactics may differ.
 - **English-only.** No multilingual coverage. Performance on non-English email is not tested and likely to be worse.
 
@@ -131,30 +148,39 @@ No PII audit was performed. Use with care. Do not try to identify individuals.
 
 Users should be made aware of the risks, biases, and limitations of the dataset. Specifically, users should:
 
-- Be cautious when applying models trained on this dataset to real-world spam detection, as the spam style may not generalize.
+- Validate models trained on this dataset on real email data before using them for email filtering, as neither the domain nor the spam style may generalize.
 - Be aware that temporal and linguistic biases may limit applicability to other contexts.
 
 ## Citation
 
 **BibTeX:**
 
-```bibtex
-@dataset{email_classification_2026,
-  author = {Carles Aguilera, Alex Bueno, Joel Delgado, Berta Torrents, Diego Velilla},
-  title = {Email Classification Dataset},
-  year = {2026},
+@misc{email_classification_2026,
+  author    = {Aguilera, Carles and Bueno, Alex and Delgado, Joel and Torrents, Berta and Velilla, Diego},
+  title     = {Email Classification Dataset},
+  year      = {2026},
   publisher = {Hugging Face},
-  url = {https://huggingface.co/datasets/diegovelilla/email-classification-dataset}
+  url       = {https://huggingface.co/datasets/diegovelilla/email-classification-dataset}
 }
-```
 
 **APA:**
 
 > Aguilera, C., Bueno, A., Delgado, J., Torrents, B., & Velilla, D. (2026). Email Classification Dataset. Hugging Face. https://huggingface.co/datasets/diegovelilla/email-classification-dataset
 
+Users of this dataset should also acknowledge the source dataset, [Deysi/spam-detection-dataset](https://huggingface.co/datasets/Deysi/spam-detection-dataset).
+
+## Glossary
+
+- **Spam:** Unsolicited message, usually with promotional or fraudulent intent.
+- **Ham:** Legitimate message that is not spam. It corresponds to the `not_spam` label.
+
+## More Information
+
+The code and documentation of the project are available in the team's repository, https://github.com/mlops-2627q1-mds-upc/MLOps-Datatouille.git
+
 ## Dataset Card Contact
 
-This project has been carried out by:
+Questions about the dataset can be addressed to TODO. The project was carried out by the following team members.
 
 - [Carles Aguilera](https://github.com/carles-aguilera-pilo)
 - [Alex Bueno](https://github.com/AlexBuenoL)