Machine Reading Comprehension: Methods and Trends of Low Resource Languages

Natural language processing (NLP) has been used to establish human-like communication with computers. Machine Reading Comprehension (MRC) has made a significant development; it has been crucial for machines to comprehend substantial language facets, like semantics, syntax, pragmatics, and phonology. Multiple studies have reported MRC models in high-resource languages. However, these models are unable to provide significant performance in MRC models in low resource languages. This is due to the unavailability of large-scale training datasets in low resource languages. Several studies on Machine Reading Comprehension (MRC) have proposed MRC models based on English. Nonetheless, these models provide insignificant performance on low-resource languages. However, limited research has been done using low resource languages, particularly Arabic, Urdu, and Hindi. This study presents a survey on trends and methods of Machine Reading Comprehension (MRC) in these low-resource languages. The survey demonstrates that available MRC models are ineffective in low resource languages, such as Hindi, Arabic, and Urdu, mainly due to large datasets and language structure unavailability. Finally, the study describes open issues in available MRC systems and provides direction for future research.