Sensing Puns is its Own Reword: Automatic Detection of Paronomasia

Date and time: 
Tuesday, November 28, 2017 - 15:00
200 Deschutes
Nisansa deSilva
University of Oregon
  • Dejing Dou (Chair)
  • Christopher Wilson
  • Stephen Fickas

Since the early stages of human language, puns have been an important component of it. The proof for this is the fact that Sumerian cuneiform and Egyptian hieroglyphs, which are two of the earliest human writing systems were based on punning systems. In contemporaneity applications, puns are abundant in literature, advertising and social media. However, despite the importance of puns in human language understanding and its long history, computational detection of puns is scarce. This shortcoming has become a major challenge to machine translation, human-computer interaction, and other applications of computational natural language processing (NLP). In building algorithms to detect puns in free text for the purpose of remedying this situation, it is prudent to analyse how state of the art systems have utilized underlying linguistic properties such as lexical polysemy and phonetics. Further, this study surveys various other innovations in NLP that have yet to be utilized in the task of automatic detection of paronomasia such as data embedding and ontology based error detection in text.