Salam Khalifa

The lack of specialized language technologies for Arabic in both NLP and Computational Linguistics is the main motivator to pursue this line of research. However, creating high quality technologies is costly, therefore, Salam found a personal interest in exploring the idea of modeling aspects of Arabic computationally using cognitively motivated models of child language acquisition.

This nexus of linguistics and NLP is rarely explored, especially for Arabic given its inherent challenges of being a morphologically rich language, and has a number of understudied dialects and linguistic domains.

 This general simplified model is then used to automatically learn different aspects of Arabic morphology using as little data as possible. Since Arabic is expressed in different but related dialectal varieties, it is expected that generalizations that apply to one dialect will apply to closely related dialect to an extent. This fact is then leveraged, and the learned generalized model is applied to neighboring dialects and then examined for the potential exceptions which are then studied in order to improve upon the model and enhance its generalizability. This model is then used to create deployable language technologies that are readily usable in NLP pipelines and in other aspects of Computational Linguistics.