Many individuals with autism spectrum disorder (ASD) face challenges in communicating effectively due to difficulties in interpreting and empathizing with emotions. This issue is further exacerbated in the virtual realm, where comprehending others' emotions during conversations can be particularly challenging. This study investigates the potential of artificial intelligence (AI) to enhance virtual communication for individuals with ASD by training two distinct sentiment analysis machine learning models: one that includes neutral sentiments and another that excludes them. Both models are trained using human-tagged sentiments from a variety of texts. The ultimate goal of this study is to evaluate the feasibility of AI-generated sentiment indicators in supporting individuals with ASD in their virtual communication.