fizza-rubab / poet-ghazal-dataset

This repository serves as a public data set for shayari from notable Urdu Ghazal poets. The project "Poet Attribution of Urdu Ghazals using Deep Learning" has been published in IEEE.

Poet-Ghazal-Dataset

This repository serves as a public data set for shayari from notable Urdu Ghazal poets.

Source

This dataset has been scraped from the most reliable platform for Urdu Poetry, Rekhta. Rekhta is an Indian literary website devoted to promoting and encouraging Urdu poetry and prose in the subcontinent.

Figures

A corpus of 17,609 couplets was collected from 15 different notable Urdu poets. These couplets (consisting of 2 misras concatenated together).

Acquisition

The data acquisition process from Rekhta involved a blend of manual and automated techniques. We extracted the links to the ghazals of each poet, and visited those links using an automated program. Regular expressions and string processing to extract the couplets of the ghazals from the visited pages, resulting in a dataset of Urdu couplets from various poets.

Characterization

Classwise count of the selected poets is shown below.

No	Poet/Shayar	Couplet Count
1	Ahmed Faraz	926
2	Zafar Iqbal	1104
3	Qateel Shifai	780
4	Parveen Shakir	593
5	Nida Fazli	474
6	Faiz Ahmad Faiz	504
7	Jaun Elia	1470
8	Muneer Niyazi	523
9	Allama Iqbal	797
10	Riyaz Khairabadi	1700
11	Haider Ali Atish	1330
12	Siraj Aurangabadi	860
13	Mir Taqi Mir	2971
14	Nazeer Akbar Abadi	1643
15	Mirza Ghalib	1934

Poet Selection

These 15 poets were selected for the following reasons:

1. Poets [1-8] wrote in the 19th-20th century, while poets [9-15] belonged to the 18th-19th century. This diversity exposes the model to a range of styles in word usage and dialect, facilitating pattern recognition and enhancing performance.
2. The chosen poets in the dataset possess a substantial number of ghazals, ensuring a sufficiently large dataset for training a deep learning model. Each poet has a minimum of 450 couplets, with Mir Taqi Mir having the highest count at 2986.
3. All selected poets hold significant popularity and esteem in Urdu literary circles. This ensures that the analysis and results of this research can be appreciated by readers beyond the realm of computer science.