| 摘要 |
PURPOSE:To attain a language processing method having satisfactory working efficiency by storing a function calculated once and showing the relation matching degree and comparing the basic arithmetic with addition of read numbers. CONSTITUTION:When the sentence lattices and the numeric values showing the sureness degrees of paragraphs are received, an optimum paragraph string is selected based on the relation matching degree between two paragraphs and the sureness degree of each paragraph. Then an optimum syntax structure is decided in response to said selected paragraph string. The propriety degree of said syntax structure is calculated by defining the character positions decided by two given natural numbers as the KANA (Japanese syllabary) description initiation and termination positions respectively. Then a paragraph string having the highest propriety degree is selected as a Japanese phrase out of the entire paragraph strings where the last paragraph is fixed. This selected paragraph is stored and then utilized when the similar calculation is carried out to a longer KANA character section that includes such a KANA character section that is decided by both the initiation and the termination. Thus it is possible to avoid systematically such a case where the calculations which are partly equal to each other are repeated.
|
