Update README.md

README.md

@@ -1,23 +1,143 @@
 ---
 dataset_info:
   features:
-  - name: html_url
-    dtype: string
-  - name: labels
-    list: string
-  - name: text
-    dtype: string
-  - name: issue_number
-    dtype: int64
+    - name: html_url
+      dtype: string
+    - name: labels
+      dtype: list
+    - name: text
+      dtype: string
+    - name: issue_number
+      dtype: int64
   splits:
-  - name: train
-    num_bytes: 50874000
-    num_examples: 12484
-  download_size: 10226655
-  dataset_size: 50874000
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
+    - name: train
+      num_examples: 12484
+  task_categories:
+    - text-classification
+  task_ids:
+    - multi-label-classification
+  languages:
+    - en
 ---
+
+# 🧩 Scikit-learn GitHub Issues – Multilabel Dataset
+
+This dataset contains **GitHub issues from the `scikit-learn` repository**, prepared specifically for **multilabel NLP tasks** such as issue tagging, automated triage, and semantic search.
+
+Each sample represents **one issue + one comment context**, making it suitable for fine-grained text understanding and real-world ML workflows.
+
+---
+
+## 📌 Motivation
+
+Open-source projects rely heavily on labeled GitHub issues:
+- bug tracking
+- feature requests
+- documentation updates
+- module-specific discussions
+
+This dataset aims to support:
+- **multilabel classification**
+- **label recommendation systems**
+- **semantic search over issues**
+- downstream LLM and RAG pipelines
+
+---
+
+## 📦 Dataset Construction
+
+### Source
+- Repository: `scikit-learn/scikit-learn`
+- Data collected using the **GitHub REST API**
+
+### Included
+✅ Open and closed **issues**  
+✅ Full issue **title + body**  
+✅ All **issue comments**  
+✅ Original **GitHub labels**
+
+### Excluded
+❌ Pull requests  
+❌ System / bot content (where possible)
+
+---
+
+## 🔄 Preprocessing Steps
+
+1. **Fetched issues** up to API safety limits
+2. **Filtered out pull requests**
+3. Retrieved **all comments** for each issue
+4. **Exploded comments**, creating multiple rows per issue
+5. Combined:
+   - issue title
+   - issue body
+   - comments  
+   into a single `text` field
+6. Extracted label names into a clean multilabel format
+
+---
+
+## 📊 Dataset Statistics
+
+- **Total samples**: ~12,500
+- **Unique labels**: ~20+
+- **Labels per sample**: 1–6
+- **Language**: English
+
+### Example labels
+- `Bug`
+- `Documentation`
+- `New Feature`
+- `module:linear_model`
+- `Build / CI`
+- `Needs Triage`
+
+---
+
+## 🧱 Dataset Schema
+
+| Column        | Type   | Description |
+|--------------|--------|-------------|
+| `html_url`   | string | URL of the GitHub issue |
+| `labels`     | list   | List of issue labels (multilabel target) |
+| `text`       | string | Issue title + body + comments |
+| `issue_number` | int | Original GitHub issue number |
+
+---
+
+## 🚀 Intended Use
+
+This dataset is suitable for:
+- Multilabel text classification
+- Label recommendation models
+- Semantic search and retrieval
+- Training LLM-powered developer tools
+
+Example use cases:
+- Auto-tagging incoming GitHub issues
+- Finding similar historical issues
+- Prioritization and triage dashboards
+
+---
+
+## ⚠️ Limitations
+
+- Strong **class imbalance** (natural for GitHub data)
+- Domain-specific to scikit-learn
+- Labels evolve over time as project conventions change
+
+---
+
+## 🔗 Related Resources
+
+- Trained model on this dataset:  
+  👉 `Talip7/scikit-learn-multilabel-classifier`
+- Hugging Face LLM Course – Chapter 5 (Datasets & Semantic Search)
+
+---
+
+## 👤 Author
+
+**Talip7**  
+AI / NLP practitioner focusing on **real-world datasets**,  
+end-to-end ML pipelines, and applied transformer models.