Abstract

The purpose of the proposed project is to develop an open-source sentiment dictionary of African American English (AAE) using crowd-sourced valence assignment of vocabulary derived from tweets using AAE. Sentiment analysis is a widely used method in

research, media, and industry to gauge a target population’s response to events. Yet, the tools used to conduct sentiment analysis employ valence assignments that were derived from linguistic analysis of written text and have been developed from data including exclusively Mainstream American English. As a result, the current dictionaries available for sentiment analysis perform poorly on use of non-standard language or transcribed spoken language, despite efforts to normalize non-standard English (S proat et al., 2001; Plank, 2016; Flint et al., 2017). This results in linguistic discrimination of non-standard variaties, such as AAE; for example, in a study comparing MAE and AAE text, Groenwold et al. (2020) found that a sentiment dictionary assigned more negative sentiment to AAE text despite both texts expressing the same intent. The proposed project aims to fill a gap in text analysis tools available for AAE. The development of a sentiment analysis dictionary will allow researchers to analyze AAE while retaining the linguistic complexity and uniqueness of the language.

Principal Investigators