JUMLA-QSL-22: Creation and Annotation of a Qatari Sign Language Corpus for Sign Language Processing
Achraf Othman, Oussama El Ghoul, Maryam Aziz, Khansa Chemnad, Sammy Sedrati, Amira Dhouib · 2023
Sign language processing (SLP) is essential to creating digital assistive technology for deaf people and requires the availability of large-scale corpus for accurate systems. However, large-scale publicly available datasets for sign languages, particularly Qatari Sign language, are limited. The aim of this paper is to present the development of the first Qatari Sign Language (QSL) corpus, named JUMLA-QSL-22, and the annotation process used to facilitate the SLP research. The corpus supplies a large sample of Arabic sign language in the Qatari dialect, with 6300 records collected over a period of five months. The study involved seven participants, including five hearing-impaired individuals and two sign language interpreters. The participants were recorded signing a sentence from a list of 900 sentences classified into 22 intents at a time. Furthermore, a new tool is proposed for sign language annotation, called “Jumla Annotation Tool” to build a Qatar Sign Bank. The videos were annotated using four types of annotations, namely glosses, translation, signer identity, and signer location. A consensus-based annotation approach was adopted to ensure the accuracy of annotations. This paper also discusses the challenges faced during the corpus development and annotation process, as well as the proposed solutions.