Abstract
The author aims to develop a pun recognition system as pilot research on machineunderstanding techniques for rhetorical expressions in natural language. This report describes several phonemic features of a type of written puns called the “separately located type”. A pun of this type essentially consists of two separately located words called “actual vehicles” which are phonemically distorted from “restored vehicles”. For example, a pun “I am angary with the Hangaryan.” consists of the former actual vehicle “angary” which is distorted from the restored vehicle “angry”, and the latter actual vehicle “Hangaryan” which is distorted from the restored vehicle “Hungarian”. In this report, the following comparisons are performed for each of 203 puns: (1) The lengths of phoneme sequences of the former actual vehicles are compared with those of the latter ones.(2) The phonemes of the former actual vehicles are compared with those of the latter ones.(3) The phonemes of the actual vehicles are compared with those of the restored vehicles. Results of these comparisons suggest that there are relativery few phonemic distortions but they are regular. These results are useful in developing a pun recognition system.
