African American Language (AAL), a set of marginalized language varieties spoken by African Americans, has been studied widely by linguists. Yet, the availability of accessible sources of AAL speech data is greatly limited. Studies have described the stigmatization of AAL perpetuated by many Americans, negatively impacting job prospects, legal decisions, educational development, and health outcomes for African Americans who speak AAL.
The effects of this gap in speech data extend to emerging technologies in natural language processing (NLP). These technologies are used in personal device virtual assistants, social media auto-captioning services, and typing assistance tools. As text and speech recognition expands into everyday life, racial disparities in their NLP performance could have dire consequences for African Americans.
SPOHP’s Joel Buchanan Archive of linguistically annotated AAL audio and transcripts provide a much-needed boost in African American speech data needed to create more equitable Natural Speech Processing (NLP) systems. Given the archive’s size, this could have a significant impact on the development of NLP. Additionally, this project will provide a tool for activists and researchers to use in holding major natural language processing developers accountable for addressing their racially biased systems.