Machine translation method for PDF file

• US 20090030671A1
• Filed: 05/29/2008
• Published: 01/29/2009
• Est. Priority Date: 07/27/2007
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Yin, Chang-Hao, Park, Eun jin, Hwang, Young-Sook, Lee, Ki Young, Kwon, Oh Woog, Kim, Young Kil, Kim, Chang Hyun, Choi, Sung Kwon, Seo, Young-Ae, Yang, Seong Il, Roh, Yoon-Hyung
• Granted Patent: US 8,108,202 B2
• Time in Patent Office: Days
• US Class Current: 704/2
• CPC Class Codes: G06F 40/103 Formatting, i.e. changing of presentation of documents automatic justification G06F40/189; automatic line break hyphenation G06F40/191
• CPC Class Codes: G06F 40/143 Markup, e.g. Standard Generalized Markup Language [SGML] or Document Type Definition [DTD]
• CPC Class Codes: G06F 40/42 Data-driven translation

Abstract

Disclosed is a machine translation method for a PDF file. A machine translation device extracts source language text and non-text from the input source language PDF file through image transformation, corrects the extracted source language text by using the source language text extracted from text information, restores a part that is contextually separated by the non-text from among the extracted source language text, generates a source language XML/HTML file by rearranging the extracted text and non-text so as to satisfy the contextual flow of the source language PDF file, separates source language text from a tag of the source language XML/HTML file, generates target language text by using translation knowledge and a transformation engine specified for the technical field corresponding to the source language PDF file, inserts the translated target language text other than source language text into XML/HTML file, and transforms the generated target language XML/HTML file into a target language PDF file to be output.

Claims

1. A machine translating method for a source language PDF file in a machine translation device, the method comprising:

extracting source language text from the source language PDF file;

restoring the source language text by combining a contextually divided body paragraph in the source language text; and

translating the restored source language text into target language text.

2. The method of claim 1, wherein

the step of extracting includes:

transforming the source language PDF file into an image;

extracting first source language text from the source language PDF file image;

extracting second source language text from the source language PDF file when the source language PDF file is an electronic document text file; and

outputting the source language text by correcting an error through comparison of the first source language text and the second source language text when the source language PDF file is an electronic document text file.

3. The method of claim 2, wherein

the step of outputting includes:

selecting the first source language text as the source language text when the error is a character repetition error; and

selecting the second source language text as the source language text when the error is not a character repetition error.

4. The method of claim 1, wherein

the step of restoring includes:

analyzing the source language text and classifying it into at least one text paragraph;

classifying the at least one text paragraph into at least one body paragraph and at least one non-body paragraph;

combining a numerical formula and the body paragraph that is provided before the numerical formula from the at least one body paragraph when the numerical formula is provided in the source language PDF file;

determining whether all body paragraphs including the body paragraph combined with the numerical formula are contextually separated;

combining contextually separated body paragraphs when there are contextually separated body paragraphs; and

restoring the source language text by rearranging the combined body paragraph.

5. The method of claim 4, wherein

the step of classifying into at least one body paragraph and at least one non-body paragraph includes

classifying a text paragraph corresponding to a preface, a postface, a footnote, or a caption from among the at least one text paragraph into the at least one non-body paragraph, and

the step of classifying into at least one non-body paragraph includes:

determining an upper text paragraph that digresses from the page layout of the source language PDF file as the preface;

determining a lower text paragraph that digresses from the page layout of the source language PDF file as the postface;

determining a text paragraph that is identified by a random line over the postface, which is provided at the bottom of the page, and in which a string start character is a subscript, to be the footnote; and

determining a text paragraph that is provided below or above a figure or a table and that starts with a string for designating the figure or the table to be the caption.

6. The method of claim 4, wherein

the step of determining includes:

selecting a body paragraph that is not contextually separated from among the at least one body paragraph based on font information and paragraph information;

testing a paragraph of the body paragraph other than the body paragraph that is not contextually separated from among the at least one body paragraph, and extracting start and end information of the paragraph; and

selecting the contextually separated body paragraph based on the start and end information.

7. The method of claim 1, wherein

the step of translating includes:

classifying the source language text into at least one sentence and at least one word included in the at least one sentence;

generating a source language syntactic tree through a morpheme analysis for the at least one sentence;

transforming the source language syntactic tree into a target language syntactic tree by using translation knowledge and transformation patterns corresponding to the technical field that corresponds to the source language PDF file;

transforming the source language vocabulary included in the at least one sentence into target language vocabulary; and

generating the target language text by arranging the target language vocabulary based on the target language syntactic tree.

8. The method of claim 1, further comprising:

generating a source language markup language file by rearranging the restored source language text; and

inserting the target language text other than the restored source language text into the source language markup language file to generate a target language markup language file.

9. A recording medium for recording a program for performing the method disclosed in claim 1.

10. A method for mechanically translating a source language PDF file in a machine translation device, the method comprising:

transforming the source language PDF file into an image by using print information of the source language PDF file;

extracting first source language text from the resulting image;

extracting second source language text from text information of the source language PDF file when the source language PDF file is an electronic document text file;

comparing the first source language text and the second source language text and correcting the first source language text when the source language PDF file is an electronic document text file; and

translating the first source language text into target language text.

11. The method of claim 10, wherein

the step of correcting includes

correcting a character of the first source language text with a character of the second source language text when another character is generated according to the result of comparison between the first source language text and the second source language text and there is no repetition error at the character of the second source language text.

12. The method of claim 10, wherein

the step of translating includes:

restoring the first source language text by combining a paragraph that is separated by at least one of non-text, postface, preface, footnote, and caption from the first source language text; and

transforming the first source language text into the target language text by using translation knowledge and transformation patterns specified for the technical field that corresponds to the source language PDF file.

13. A method for mechanically translating a source language PDF file in a machine translation device, the method comprising:

extracting source language text and non-text from the source language PDF file;

restoring the source language text by combining a contextually separated paragraph from among the source language text;

generating a source language markup language file by rearranging the restored source language text and the non-text based on the contextual flow of the source language PDF file; and

generating a target language markup language file by translating the restored source language text into target language text.

14. The method of claim 13, wherein

the step of extracting comprises:

extracting the source language text from the source language PDF file through image transformation; and

when the source language electronic document file is an electronic document text file, comparing the source language text and the text that is extracted from text information included in the source language PDF file, and correcting the source language text when an error occurs except a character repetition error of the text that is extracted from the text information.

15. The method of claim 13, wherein

the step of restoring includes:

analyzing the source language text, and classifying it into at least one text paragraph;

classifying the at least one text paragraph into at least one body paragraph and at least one non-body paragraph;

combining the body paragraph separated by the non-text from among the at least one body paragraph and at least one of the at least one non-body paragraph; and

restoring the source language text by rearranging the combined body paragraph.

16. The method of claim 13, wherein

the step of translating into the target language text includes:

transforming the source language text into a target language syntactic tree by using a transformation pattern specified for the technical field that corresponds to the source language PDF file;

transforming vocabulary of the source language text into target language vocabulary by using a translation dictionary specified for the technical field that corresponds to the source language PDF file; and

generating the target language text by arranging the target language vocabulary based on the target language syntactic tree.