This paper proposes a method to detect and cor,rect erroneous characters which locate so closely in Japanese text that they interfere each other. Texts input through OCR or speech recognition devices usually contain three kind of erroneous characters such as wrongly substituted, wrongly deleted and wrongly inserted characters. Recently, selective error correction method has been proposed for these three kind of errors and it was shown that the method is very effective for the errors of isolated type which are located at so different places from each other that there in no mutual interference. However, the method does not work well for the error characters located closely each other, In such a case, it has been very difficult not only to distinguish the types of errors but also to determine the location and the length of erroneous character parts. In order to resolve this problem, this paper proposed a two tiered methods. First it finds an error section which includes one or more error character8 located closely each other. Second the error section is analyzed in detail to find and correct error characters. To perform these proceeeea, rkip type Markov chain model was newly introduced. The experimental results showed that the proposed method was very effective for finding and correcting error charactera cloaely located each other in Japanese textr. kmission to make digital or hard Copier Of all Or put Of 1hiS Wd fa perao~i w classroom use is gmtcd witbout fee provided that copia are not nude or distributed for proffi 01 CO~MICVA advantage and that copies bear this notice and the full citation on the lint page. To copy otherwise, lo republish, to post on servers or to redistribute to lists. requires prior specific permission and/or a fee. SAC 99. San Anlonio, Texas 01998 ACM l-581 13-08649910001 S5.00

Orthography and Word Recognition in Reading

Detection and correction of mutually interfered erroneous characters in Japanese texts

Segmentation in models of reading

Research on reading has regained the attention of experimental psychologists in the last two decades (Tzeng, 1981). The new field is markedly interdisciplinary, and addresses questions of practical application as readily as questions of pure theory or knowledge. These questions have been attacked from a number of different perspectives, using various experimental paradigms. Some studies record patterns of eye movements and relate them to various phases of information pickup during reading comprehension (Thibadeau, Just, & Carpenter, 1982). Others look into specific characteristics of word perception, e.g., the word-superiority effect, and attempt to build interactive models of human information processing (Rumelhart, 1985). A third type of study examines the consequences of becoming literate in more than one writing system. For example, differences as subtle as phonemic awareness and as general as problem-solving strategies have been observed in such comparative reading studies (Scribner & Cole, 1978; Liberman, Liberman, Mattingly, & Shankweiler, 1980; Tzeng & Hung, 1981).

Orthography and Word Recognition in Reading

Citations

Detection and correction of mutually interfered erroneous characters in Japanese texts

Segmentation in models of reading

Orthography, Reading, and Cerebral Functions

Knowing and knowing you know: Better methods or better models?

Visual word processing: Procedures, representations, and routes

Related Papers (5)

The Organization and Activation of Orthographic Knowledge in Reading Aloud.

An interactive activation model of context effects in letter perception: I. An account of basic findings.

Interaction of information in word recognition.

Lexical access in simple reading tasks

Patterns of paralexia: A psycholinguistic approach