File size: 5,609 Bytes
ac9e8c2
 
 
 
 
 
 
 
 
 
 
 
3c20eaa
3a6f4f4
3c20eaa
3a6f4f4
3c20eaa
3a6f4f4
3c20eaa
3a6f4f4
3c20eaa
3a6f4f4
3c20eaa
 
 
3a6f4f4
3c20eaa
3a6f4f4
3c20eaa
3a6f4f4
ac9e8c2
3c20eaa
ac9e8c2
 
 
 
 
64efa04
ac9e8c2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6cb3484
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ac9e8c2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
---
license: cc-by-4.0
task_categories:
  - text-classification
language:
  - ar
size_categories:
  - 10K<n<600K
pretty_name: "Habibi: A Multi-Dialect, Multi-National Arabic Song Lyrics Corpus"

dataset_info:
  features:
    - name: songID
      dtype: int32
    - name: Singer
      dtype: string
    - name: SongTitle
      dtype: string
    - name: SongWriter
      dtype: string
    - name: Composer
      dtype: string
    - name: LyricsOrder
      dtype: int32
    - name: Lyrics
      dtype: string
    - name: SingerNationality
      dtype: string
    - name: SongDialect
      dtype: string


  splits:
    - name: train
    - name: validation
    - name: test
---

# Habibi: A Multi-Dialect, Multi-National Arabic Song Lyrics Corpus

**Habibi** is the first large-scale open corpus of Arabic song lyrics, comprising more than **30,000 songs** performed by artists from **18 Arab countries**. The collection represents **six major Arabic dialect groups**, segmented into over **520,000 verses** and containing more than **3.5 million words**. 

The corpus was designed to support computational research in **Arabic dialect identification**, **country-of-origin classification**, **authorship analysis**, and wider studies in Arabic language variation within contemporary music.

---

## 🌍 Corpus Overview

The Habibi corpus brings together lyrics from across the Arab world, reflecting the linguistic and cultural diversity of modern Arabic music. Each song includes:

- **Singer name**
- **Song title**
- **Country of origin**
- **Dialect category**  
  (Egyptian, Gulf, Levantine, Iraqi, Sudanese, Maghrebi)
- **Writer and composer** (where available)
- **Lyrics segmented into verses**

All lyrics were collected using a Web-as-Corpus approach and manually curated to maintain a clean, noise-free dataset free of emojis, hashtags, and social-media artefacts.

### Key Statistics

| Item | Count |
|------|-------|
| Songs | 30,072 |
| Verses (sentences) | 527,870 |
| Words | 3.57 million |
| Singers | 1,765 |
| Writers | 3,789 |
| Composers | 2,463 |
| Countries | 18 |
| Dialects | 6 |

---

## 🗂 Dialects and Geographic Coverage

Songs are categorised into six dialect groups widely recognised across modern Arabic music:

- **Egyptian**
- **Gulf**
- **Levantine**
- **Iraqi**
- **Sudanese**
- **Maghrebi**

Although singers often perform in dialects other than their own, the corpus assigns dialect based on the singer’s country of origin, following conventions used by major Arabic music platforms (Anghami, Spotify, Deezer).

---

## 📁 File Formats

The corpus is distributed in several machine-readable formats to support a range of NLP tasks:

### **1. CSV (primary format)**  
Contains all metadata fields and verse-level text in UTF-8.

### **2. Annotated TXT**  
Each song is provided as a structured, UTF-8 annotated document with lightweight XML-style tags identifying metadata and verse boundaries.

### **3. JSON & XML**  
Converted directly from the CSV, preserving all metadata and text segmentation.

---

## 🔧 Recommended Use-Cases

Researchers may use Habibi to explore:

- Arabic **dialect identification** (sentence- or song-level)
- **Country-of-origin classification**
- **Authorship attribution** (singers, songwriters)
- **Linguistic variation** in modern Arabic music
- **Embeddings training** using domain-specific lyrics
- **Sociolinguistic analysis** of themes, vocabulary, and regional patterns

The dataset has demonstrated strong performance across classical and deep-learning models, including CNN, LSTM, CLSTM, BiGRU, and BiLSTM architectures.

---

## 📝 Data Split
This dataset is provided with standard train/validation/test splits created using an 80/10/10 random split from the master file `habibi.csv`.

```
import pandas as pd
from sklearn.model_selection import train_test_split

df = pd.read_csv("habibi.csv")

train, temp = train_test_split(df, test_size=0.2, random_state=42)
val, test = train_test_split(temp, test_size=0.5, random_state=42)

train.to_csv("train.csv", index=False)
val.to_csv("validation.csv", index=False)
test.to_csv("test.csv", index=False)

```
---

## 📊 Benchmarks

Extensive experiments were conducted on:

- **Binary dialect classification**  
- **Multi-class dialect classification (3–6 classes)**
- **Binary country-of-origin classification**
- **Multi-class country classification (3–18 classes)**

Models tested include:

- Naïve Bayes, Logistic Regression, SVM  
- CNN, LSTM, BiLSTM, CLSTM, BiGRU  
- FastText Arabic word embeddings  
- In-house **Habibi CBOW embeddings** (300-dimensional)

**The word-level CNN model** achieved the strongest deep-learning results, while **Naïve Bayes** consistently performed best among classical models on multi-class tasks.

---

## 📦 Download

All formats (CSV, TXT, JSON, XML) as well as **Habibi’s in-house word embeddings** are available for free research use.

Original repository:  
http://ucrel-web.lancaster.ac.uk/habibi/

---

## 📝 Citation

If you use the Habibi corpus, please cite:

**El-Haj, M. (2020).**  
*Habibi – a multi-dialect multi-national Arabic song lyrics corpus.*  
In **Proceedings of the Twelfth Language Resources and Evaluation Conference (LREC 2020)**, pp. 1318–1326.  
URL: https://www.lancaster.ac.uk/staff/elhaj/docs/habibi.pdf

### BibTeX
```bibtex
@inproceedings{elhaj2020habibi,
  title={Habibi--a multi dialect multi national Arabic song lyrics corpus},
  author={El-Haj, Mahmoud},
  booktitle={Proceedings of the Twelfth Language Resources and Evaluation Conference},
  pages={1318--1326},
  year={2020}
}