File size: 2,371 Bytes
0dcdcd6
8d48408
 
 
 
 
 
 
 
0dcdcd6
990635a
 
 
8d48408
990635a
8d48408
990635a
 
 
 
 
8d48408
 
 
 
 
990635a
8d48408
 
 
 
 
990635a
 
 
 
 
8d48408
990635a
8d48408
990635a
8d48408
 
 
 
 
990635a
8d48408
 
990635a
 
 
8d48408
990635a
8d48408
 
 
 
 
 
 
990635a
8d48408
990635a
 
 
8d48408
990635a
8d48408
990635a
8d48408
990635a
 
8d48408
990635a
 
 
 
 
 
 
 
 
8d48408
990635a
8d48408
 
 
 
 
 
 
 
990635a
 
 
 
 
8d48408
 
 
 
990635a
 
 
 
 
8d48408
990635a
8d48408
990635a
 
 
 
 
 
8d48408
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
---
language:
- en
task_categories:
- text-classification
task_ids:
- multi-label-classification
pretty_name: Scikit-learn GitHub Issues (Multilabel)
license: apache-2.0
---

# 🧩 Scikit-learn GitHub Issues – Multilabel Dataset

This dataset contains **GitHub issues from the `scikit-learn` repository**, prepared for **multilabel NLP tasks** such as issue tagging, automated triage, and semantic search.

Each row corresponds to **one issue-comment context**, making the dataset suitable for real-world developer tooling.

---

## 📌 Motivation

GitHub issues are a critical signal in open-source projects:
- Bug tracking
- Feature requests
- Documentation improvements
- Module-specific discussions

This dataset enables:
- Multilabel text classification
- Label recommendation systems
- Semantic search over issues
- Downstream LLM & RAG pipelines

---

## 📦 Dataset Construction

**Source**
- Repository: `scikit-learn/scikit-learn`
- Collected using the GitHub REST API

**Included**
- Open & closed **issues**
- Issue **title + body**
- All **comments**
- Original **GitHub labels**

**Excluded**
- Pull requests

---

## 🔄 Preprocessing Pipeline

1. Retrieved issues up to API safety limits  
2. Removed pull requests  
3. Downloaded all issue comments  
4. Exploded issues by comments  
5. Constructed a unified `text` field:
   - title  
   - body  
   - comments  
6. Extracted label names into a multilabel format

---

## 📊 Dataset Overview

- **Samples**: ~12,500
- **Labels per sample**: 1–6
- **Unique labels**: ~20+
- **Language**: English

Example labels:
- `Bug`
- `Documentation`
- `New Feature`
- `module:linear_model`
- `Build / CI`
- `Needs Triage`

---

## 🧱 Dataset Columns

| Column | Description |
|------|------------|
| `html_url` | GitHub issue URL |
| `labels` | List of labels (multilabel target) |
| `text` | Issue title + body + comments |
| `issue_number` | Original GitHub issue number |

> Column types are inferred automatically from the dataset files.

---

## 🚀 Intended Use

- Multilabel classification
- Issue triage automation
- Semantic search
- Developer-facing ML tools

---

## ⚠️ Limitations

- Natural class imbalance
- Domain-specific to scikit-learn
- Label taxonomy evolves over time

---

## 👤 Author

**Talip7**  
Focused on applied NLP, real-world datasets, and production ML pipelines.