A new critical survey on arXiv examines how natural language processing (NLP) technologies are being developed and applied, with a focus on queer communities. The paper notes that NLP is rapidly reshaping how language is created, processed, and interpreted, and that these systems are already used in consequential domains such as hiring, legal decisions, and healthcare.
The authors identify significant gaps in the existing literature, including a lack of attention to queer identities and experiences in both training data and evaluation practices. They also point to persistent biases that can lead to harmful outcomes when NLP systems are deployed without adequate consideration of diverse gender and sexual identities.
The survey reviews current trends and argues for a more critical and inclusive research agenda. It emphasizes the need to move beyond token inclusion and to address structural issues in how NLP models are built, tested, and used in real-world settings.