Labelled data set of human-written and AI-generated text (DET12)

This data set contains labelled text excerpts of detective fiction for the classification of text as either human-written or AI-generated. The texts are preprocessed and tokenized such that each text excerpt can be split on whitespaces to obtain the tokenized text. The data is split into training, validation, and test sets.

Each excerpt is labelled as 0 (human) or 1 (AI-generated). The human-written texts (label 0) are sourced from 12 copyright-free detective novels, and the AI-generated texts (label 1) are generated with GPT4.1. A detailed description of the data generation process is given in the associated publication.