Corpus for Emotion Detection on Roman Urdu

Abstract

Language assets, like corpora, are essential for different natural language processing tasks. There are many useful applications of emotion analysis of text such as dialog systems, smart agents, mental disorder clinical diagnoses. In this research, an emotion labeled corpus for Roman Urdu is presented for evaluation of emotions in short text. We collected a sizable corpus with 10,000 manually annotated sentences in Roman Scripted Urdu to facilitate the development and evaluation of emotion detection systems for Roman Urdu. This corpus is the first of its kind to be created for the Roman Urdu.