Textual analysis of artificial intelligence manuscripts reveals features associated with peer review outcome

Abstract We analyzed a data set of scientific manuscripts that were submitted to various conferences in artificial intelligence. performed combination semantic, lexical, and psycholinguistic analyses the full text compared them with outcome peer review process. found accepted scored lower than rejected on two indicators readability, they also used more intelligence jargon. written words are less frequent, acquired at an older age, abstract manuscripts. The analysis references included revealed subset submissions likely cite same publications. This finding was echoed by pairwise comparisons word content (i.e., indicator semantic similarity), which similar Finally, we predicted their content, related machine learning neural networks positively acceptance, whereas logic, symbolic processing, knowledge-based systems negatively acceptance.