Machine translation into a target language by interactively and automatically formalizing non-formal source language into formal source language

• US 8,990,067 B2
• Filed: 12/28/2010
• Issued: 03/24/2015
• Est. Priority Date: 12/31/2009
• Status: Active Grant

• Current Assignee: Guangyuan Cheng
• Original Assignee: Guangyuan Cheng
• Inventors: Cheng, Guangyuan
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US13/520,146
• Publication Number:
  US 20120278062A1
• Time in Patent Office: 1,547 Days
• Field of Search: 704/2, 704/3, 704/4, 704/5, 704/277, 704/8
• US Class Current: 704/2
• CPC Class Codes: G06F 40/55 Rule-based translation

Abstract

A machine translation method and system comprises the steps of (a) formalizing a non-formal source language in an interactive or automatic way and (b) transforming the formal source language into a formal or non-formal target language in an automatic way. It eliminates the language barrier between person and person and the language barrier between person and computer: A user translates his/her non-formal native language correctly and without lexical ambiguity into any non-formal foreign language which he/she knows nothing about; a user and a computer exchange information in his/her non-formal native language correctly and without lexical ambiguity. It can be used in network terminal equipment, Internet knowledge bases, knowledge reasoning search engines, expert systems and automatic programming. That formalization of a source language is the common foundation for transformation into various target languages makes it especially suitable for multilingual machine translation.

Claims

1. A method of machine translation comprising steps of:

a) formalizing a non-formal source language by a computer processor; and

b) transforming a formal source language into a target language;

wherein the non-formal source language is formalized in an interactive way or an automatic way;

the target language is a formal target language or a non-formal target language;

the step a) comprises steps of;

identifying one by one fixed language segments of an initial language segment of the non-formal source language and tagging the fixed language segments with meaning marks until the last fixed language segment and then composing level by level non-fixed language segments of the initial language segment and tagging the non-fixed language segments with key component marks and relation marks until the non-fixed language segment constituted by the whole initial language segment with the computer processor;

for transforming the formal source language into the formal target language, the step b) comprises a step of transforming in the automatic way the fixed language segments of the source language into language segments of the target language according to fixed language segment transformation rules stored in a computer memory;

for transforming the formal source language into the non-formal target language, the step b) comprises steps of transforming in the automatic way the fixed language segments of the source language into language segments of the target language according to the fixed language segment transformation rules stored in the computer memory and then transforming level by level in the automatic way the non-fixed language segments of the source language into language segments of the target language according to non-fixed language segment transformation rules stored in the computer memory.

2. The method of machine translation according to claim 1, wherein:

the step a) further comprises a step of pre-processing by means of substitution marks, which specifically comprises steps of;

using the computer processor for separating an initial language segment into a number of sub-segments by means of substitution marks in advance, and then formalizing the sub-segments respectively.

3. The method of machine translation according to claim 1, wherein:

a fixed language segment mode stored in the computer memory contains a fixed language segment and its meaning mark, fixed language segments of the same form and different meanings bearing meaning marks;

in the process of identifying and tagging the fixed language segments in the interactive way, the computer processor judges in turn whether there exists in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the two writing units at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the three writing units at the forefront, and so on;

wherein n is a natural number, if there exists in the computer memory at least one fixed language segment beginning with the n writing unit or units at the forefront and there does not exist in the computer memory at least one fixed language segment beginning with the n+1 writing units at the forefront, the computer processor identifies the n writing unit or units at the forefront as a first fixed language segment of the fixed language segments and tags it with a first meaning mark of the meaning marks according to a fixed language segment mode, and after that a user confirms or revises the mark;

if the computer processor finds out that there does not exist in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront of a remaining language segment, the user identifies the fixed language segment beginning with the one writing unit at the forefront of the remaining language segment and tags it with the first meaning mark;

the process is repeated until the last fixed language segment.

4. The method of machine translation according to claim 1, wherein:

a fixed language segment mode stored in the computer memory contains a fixed language segment and its meaning mark, fixed language segments of the same form and different meanings bearing meaning marks, and contains a grammatical attribute mark and a semantic attribute mark;

in the process of identifying and tagging fixed language segments in the automatic way, the computer processor judges in turn whether there exists in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the two writing units at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the three writing units at the forefront, and so on;

wherein n is a natural number, if there exists in the computer memory at least one fixed language segment beginning with the n writing unit or units at the forefront and there does not exist in the computer memory at least one fixed language segment beginning with the n+1 writing units at the forefront, the computer processor identifies the n writing unit or units at the forefront as a first fixed language segment of the fixed language segments, and then finds all the fixed language segment modes which can be used in the computer memory, chooses one of the fixed language segment modes in the choice order that a mode with a larger total number of use is prior to a mode with a smaller total number of use, and tags the fixed language segment with a first meaning mark of the meaning marks, a grammatical attribute mark and a semantic attribute mark according to the fixed language segment mode;

if the computer processor finds out that there does not exist in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront of a remaining language segment, the computer processor performs backtracking;

the process is repeated until the last fixed language segment.

5. The method of machine translation according to claim 1, wherein:

a non-fixed language segment mode stored in the computer memory contains grammatical attribute marks and semantic attribute marks of component segments and contains a combination mark, a key component mark, a relation mark, a grammatical attribute mark and a semantic attribute mark of a composed segment;

in the process of composing and tagging non-fixed language segments in the automatic way, the computer processor finds all the non-fixed language segment modes which can be used in the computer memory, chooses one of the non-fixed language segment modes in the choice order that a mode with a larger total number of use is prior to a mode with a smaller total number of use, composes a non-fixed language segment and tags it with a key component mark, a relation mark, a grammatical attribute mark and a semantic attribute mark according to the non-fixed language segment mode;

if the computer processor finds out that there does not exist a at least one non-fixed language segment mode which can be used in the computer memory, the computer processor performs backtracking;

the process is repeated until the non-fixed language segment constituted by the whole initial language segment.

6. The method of machine translation according to claim 1, further comprising steps of:

c) after a user clicks on a meaning mark, using a screen for displaying a meaning represented by the meaning mark, and, d) after the user clicks on a relation mark, using the screen for displaying a relation represented by the relation mark.

7. The method of machine translation according to claim 1, wherein:

the non-fixed language segment transformation rule stored in the computer memory is a rule which forms a non-fixed language segment of the target language with translations of components of a non-fixed language segment of the source language and a relation word of the target language according to a key component mark and a relation mark or relation word of the non-fixed language segment of the source language;

the computer processor searches a list of non-fixed language segment transformation rules for a matching non-fixed language segment transformation rule and transforms a first non-fixed language segment of the non-fixed language segments of the source language into a first non-fixed language segment of the non-fixed language segments of the target language according to the matching non-fixed language segment transformation rule, repeating the step recursively until all the non-fixed language segments are transformed level by level, in which process, concerning the current non-fixed language segment of the source language, first, all the components of the current non-fixed language segment are transformed into the target language respectively, and then, the current non-fixed language segment is transformed into the target language according to the matching non-fixed language segment transformation rule, the result of the transformation of the current non-fixed language segment being returned to be used by the non-fixed language segment at the higher level, until the non-fixed language segment as the initial data is transformed into the target language.

8. A system of machine translation comprising:

a computer processor for formalizing a non-formal source language and a module for transforming a formal source language into a target language which is connected to the computer processor for formalizing the non-formal source language before it;

wherein the non-formal source language is formalized in an interactive way or an automatic way;

the module for transforming the formal source language into the target language has two target languages comprising a formal target language and a non-formal target language;

in the computer processor, a process of formalizing the non-formal source language performed by the computer processor is first identifying one by one fixed language segments of an initial language segment of the non-formal source language and tagging the fixed language segments with meaning marks until the last fixed language segment and then composing level by level the non-fixed language segments of the initial language segment and tagging the non-fixed language segments with key component marks and relation marks until the non-fixed language segment constituted by the whole initial language segment;

in the module for transforming the formal source language into the target language, a process of transforming the formal source language into the formal target language performed by the computer processor is transforming in the automatic way the fixed language segments of the source language into language segments of the target language according to fixed language segment transformation rules stored in a computer memory;

in the module for transforming the formal source language into the target language, a process of transforming the formal source language into the non-formal target language performed by the computer processor is first transforming in the automatic way the fixed language segments of the source language into language segments of the target language according to fixed language segment transformation rules stored in the computer memory and then transforming level by level in the automatic way the non-fixed language segments of the source language into language segments of the target language according to non-fixed language segment transformation rules stored in the computer memory.

9. The system of machine translation according to claim 8, wherein:

the system includes a substitution module which is connected to the computer processor, for pre-processing by means of substitution marks performed by the computer processor, which comprises separating the initial language segment into a number of sub-segments by means of substitution marks in advance, and then formalizing the sub-segments respectively.

10. The system of machine translation according to claim 8, wherein:

a fixed language segment mode of the computer processor contains a fixed language segment and its meaning mark, fixed language segments of the same form and different meanings bearing meaning marks;

in the process of identifying and tagging fixed language segments in the interactive way, the computer processor judges in turn whether there exists in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the two writing units at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the three writing units at the forefront, and so on;

wherein n is a natural number, if there exists in the computer memory at least one fixed language segment beginning with the n writing unit or units at the forefront and there does not exist in the computer memory at least one fixed language segment beginning with the n+1 writing units at the forefront, the computer processor identifies the n writing unit or units at the forefront as a first fixed language segment of the fixed language segments and tags it with a first meaning mark of the meaning marks according to the fixed language segment mode, and after that a user confirms or revises the mark;

if the computer processor finds out that there does not exist in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront of a remaining language segment, the user identifies the fixed language segment beginning with the one writing unit at the forefront of the remaining language segment and tags it with the first meaning mark;

the process is repeated until the last fixed language segment.

11. The system of machine translation according to claim 8, wherein:

a fixed language segment mode of the computer processor contains a fixed language segment and its meaning mark;

fixed language segments of the same form and different meanings bearing meaning marks, and contains a grammatical attribute mark and a semantic attribute mark;

in the process of identifying and tagging fixed language segments in the automatic way, the computer processor judges in turn whether there exists in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the two writing units at the forefront, whether there exists in the computer memory at least one fixed language segment beginning with the three writing units at the forefront, and so on;

wherein n is a natural number, if there exists in the computer memory at least one fixed language segment beginning with the n writing unit or units at the forefront and there does not exist in the computer memory at least one fixed language segment beginning with the n+1 writing units at the forefront, the computer processor identifies the n writing unit or units at the forefront as a first fixed language segment of the fixed language segments, and then finds all the fixed language segment modes which can be used in the computer memory, chooses one of the fixed language segment modes in the choice order that a mode with a larger total number of use is prior to a mode with a smaller total number of use, and tags the fixed language segment with a first meaning mark of the meaning marks, a grammatical attribute mark and a semantic attribute mark according to the fixed language segment mode;

if the computer processor finds out that there does not exist in the computer memory at least one fixed language segment beginning with the one writing unit at the forefront of a remaining language segment, the computer processor performs backtracking;

the process is repeated until the last fixed language segment.

12. The system of machine translation according to claim 8, wherein:

a non-fixed language segment mode of the computer processor contains grammatical attribute marks and semantic attribute marks of component segments and contains a combination mark, a key component mark, a relation mark, a grammatical attribute mark and a semantic attribute mark of a composed segment;

in the process of composing and tagging non-fixed language segments in the automatic way, the computer processor finds all the non-fixed language segment modes which can be used in the computer memory, chooses one of the non-fixed language segment modes in the choice order that a mode with a larger total number of use is prior to a mode with a smaller total number of use, composes a non-fixed language segment and tags it with a key component mark, a relation mark, a grammatical attribute mark and a semantic attribute mark according to the non-fixed language segment mode;

if the computer processor finds out that there does not exist at least one non-fixed language segment mode which can be used in the computer memory, the computer processor performs backtracking;

the process is repeated until the non-fixed language segment constituted by the whole initial language segment.

13. The system of machine translation according to claim 8, wherein:

after a user clicks on a meaning mark, a screen displays a meaning represented by the meaning mark, and, after the user clicks on a relation mark, a screen displays a relation represented by the relation mark.

14. The system of machine translation according to claim 8, wherein:

the non-fixed language segment transformation rule of the module for transforming the formal source language into the target language stored in the computer memory is a rule which forms a non-fixed language segment of the target language with translations of components of the non-fixed language segment of the source language and a relation word of the target language according to a-one of said key component marks and a-one of said relation marks or relation word of the non-fixed language segment of the source language;

the computer processor searches a list of non-fixed language segment transformation rules for a matching non-fixed language segment transformation rule, transforms the non-fixed language segment of the source language into the non-fixed language segment of the target language according to the matching non-fixed language segment transformation rule, repeating the step recursively until all the non-fixed language segments are transformed level by level, in which process, concerning a current non-fixed language segment of the source language, first, all the components of the current non-fixed language segment are transformed into the target language respectively, and then, the current non-fixed language segment is transformed into the target language according to the matching non-fixed language segment transformation rule, the result of the transformation of the current non-fixed language segment being returned to be used by the non-fixed language segment at the higher level, until the non-fixed language segment as the initial data is transformed into the target language.