Abstract: A method for translating sentences in a source language to sentences in a target language. A knowledge base includes a plurality of information patterns in the source language which represents substantially all possible information patterns of the source language. A corresponding plurality of information patterns in the target language are contained in the same knowledge base, and are associated with source language information patterns in accordance with a predetermined determination. To translate a source language sentence into a target language sentence, the source language sentence is analyzed and its constituent parts of speech are determined so that a specific information pattern is identified. The knowledge base is consulted, and the corresponding target language information pattern is identified. The source language words are then inserted into the target language information pattern to form a Linguistic Canonical Form which is translated.

Abstract: Sentences from a source language having first rules of grammar are translated into a target language having second rules of grammar; the resulting sentences are meaningful but ungrammatical. The ungrammatical sentences are changed into grammatical sentences by a method that includes identifying grammar markers and grammar marker patterns in the translated put ungrammatical sentence. A first database includes grammar rules and is consulted for each grammar marker that appears in the ungrammatical sentence. A second database includes self-correction rules and is consulted for each grammar marker pattern that appears. The grammar rules and self-correction rules are applied by a correction scheme. The resulting sentences are substantially perfectly grammatical.

Abstract: The character string of a Chinese sentence is input into the memory of a personal computer, and the first six characters of the sentence are taken from that memory and placed in a main buffer. A segment processor fetches the first three of those six characters from the main buffer and puts them into a trio buffer. The segment processor then fetches the first character from the trio buffer to match with the number dictionary. This process is continued until consecutive number characters are identified. If the first character is not a number word, the segment processor uses the three characters in the trio buffer and the segmentation program and logic to search for matched name words and regular words stored in dictionaries within the hard disk memory of the computer. The identified words are treated as candidate words and are successively sent from the computer to the word buffer for further study.