An Error-Controlling Scheme Based on Different Importance of Segments of a Natural Language

Taroh SASAKI  Ryuji KOHNO  Hideki IMAI  

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E75-A   No.9   pp.1076-1086
Publication Date: 1992/09/25
Online ISSN: 
Print ISSN: 0916-8508
Type of Manuscript: Special Section PAPER (Special Section on Information Theory and Its Applications)
intelligent communication,  error-correcting codes,  natural language,  

Full Text: PDF(834.7KB)>>
Buy this Article

Although individual segments of a natural language such as words have different importance on human interpretation of the meaning, every segment has been uniformly protected by an error-correcting code. If the importance of individual segments is defined by considering their meaning in the sentence, we can adaptively control the level of error-protection for each segment according to its importance in order to reduce errors on human interpretation of the meaning. In this paper, we propose an error-control scheme based on the varying importance of each word. We first introduce a method which determines the importance of each word and then propose an error-control scheme in which several error-correcting codes are alternately used to protect each word according to its importance. Probablity of semantic errors, that is, errors on human interpretation of the meaning, is defined and used as a criterion in mapping error-correcting codes to words possessing different importance. We theoretically formalize the problem of obtaining an optimum mapping which minimizes the probability of semantic errors under some constraint. Given a certain probability distribution of the importance of words and set of error-correcting codes, we can derive the optimum mapping. The proposed error-controlling scheme is theoretically evaluated by comparing its probability of semantic errors with that of a conventional scheme in which every word is uniformly protected by a single error-correcting code. Results show that the proposed scheme can considerably raduce the probability of semantic errors while retaining the same average transmission rate or redundancy.