A Computational Framework for Slang Generation

Slang is a common type of informal language, but its flexible nature and paucity data resources present challenges for existing natural language systems. We take an initial step toward machine generation slang by developing framework that models the speaker's word choice in context. Our encodes novel meaning relating conventional senses while incorporating syntactic contextual knowledge usage. construct using combination probabilistic inference neural contrastive learning. perform rigorous evaluations on three dictionaries show our approach not only outperforms state-of-the-art models, also better predicts historical emergence usages from 1960s to 2000s. interpret proposed find contrastively learned semantic space sensitive similarities between words. work creates opportunities automated interpretation language.