Third language text generating algorithm by multi-lingual text inputting and device and program therefor

• US 20040254783A1
• Filed: 04/26/2004
• Published: 12/16/2004
• Est. Priority Date: 08/10/2001
• Status: Abandoned Application

• Current Assignee: Communications Research Laboratory Independent Administrative Institution
• Original Assignee: Communications Research Laboratory Independent Administrative Institution
• Inventors: Isahara, Hitsohi
• Application Number:
  US10/486,087
• Publication Number:
  US 20040254783A1
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/42 Data-driven translation
• CPC Class Codes: G06F 40/55 Rule-based translation
• CPC Class Codes: G06F 40/56 Natural language generation

Abstract

Provided is a technique which includes inputting a plurality of multi-lingual texts and using multi-lingual text corpora, thereby generating a higher-accuracy third language text as compared to the input of a unilingual text which has heretofore taken place. After inputting the texts, the processes for analyzing, converting and generating are performed, and a target language document text is outputted. The generation of target language document text does not require a large-scale corpus because information characteristic of the language can be automatically acquired.

Claims

1. A third language text generating algorithm, for use in computer-based language processing, for generating anew third language text by using a plurality of multi-lingual texts, the algorithm including the steps of:

inputting two or more multi-lingual texts written in different languages including a first language which serves as a source language and at least a second language into which the first language is translated;

performing language analysis including at least dependency analysis and semantic analysis, on each of the mufti-lingual texts in the form of each language or a combination of any two or more languages, thereby obtaining language information on at least a dependency structure and semantic representation; and

generating a third language text,

wherein the generating step generates a third language text by using the language information obtained by the analyzing step, or

the algorithm further including the step of performing language conversion based on the results of analysis obtained by the analyzing step or based on the results of analysis and conversion knowledge characteristic of a third language, the converting step following the analyzing step,

wherein the generating step generates a third language text by using at least either the language information obtained by the analyzing step or the results of conversion obtained by the converting step.

2. A third language text generating algorithm according to claim 1, wherein the analyzing step includes:

an associating process for performing associating to determine the correspondence between words constructing the multi-lingual texts, the correspondence between phrases constructing the multi-lingual texts, and the correspondence between sentences constructing the multi-lingual texts;

an analyzing process for analyzing at least the first language text by using an analysis module previously prepared; and

a merging process for analyzing parts in at least the second language text corresponding to the first language text, based on the results of associating, by using an analysis module previously prepared, and merging the results of analysis.

3. A third language text generating algorithm according to claim 1 or 2, wherein at least one of the analyzing, converting and generating steps uses rule-based information containing at least either dictionary information or grammar information on each language, and empirical information based on the results of learning obtained from actual data in corpora.

4. A third language text generating algorithm according to claim 1, wherein the generating step includes:

automatically acquiring part or all of information on at least either third language syntax structure information or third language word usage information from an existing third language corpus; and

generating a third language text based on the automatically acquired information characteristic of the third language.

5. A third language text generating device, for use in language processing, for generating a new third language text by using a plurality of languages, the device including:

inputting means for inputting two or more mufti-lingual texts written in different languages including a first language which serves as a source language and at least a second language into which the first language is translated;

analyzing means for performing language analysis including at least dependency analysis and semantic analysis, on each of the mufti-lingual texts in the form of each language or a combination of any two or more languages, thereby obtaining language information on at least a dependency structure and semantic representation;

generating means for generating a third language text; and

outputting means capable of outputting the third language text generated by the generating means,

wherein the generating means generates the third language text by using the language information obtained by the analyzing means, or

the device further including converting means for performing language conversion based on the results of analysis obtained by the analyzing means or based on the results of analysis and conversion knowledge characteristic of a third language,

wherein the generating means generates the third language text by using at least either the language information obtained by the analyzing means or the results of conversion obtained by the converting means.

6. A third language text generating device according to claim 5, wherein the analyzing means includes:

an associating portion which performs associating to determine the correspondence between words constructing the multi-lingual texts, the correspondence between phrases constructing the multi-lingual texts, and the correspondence between sentences constructing the multi-lingual texts;

an analysis module which analyzes at least the first language text; and

a merging portion which analyzes parts in at least the second language text corresponding to the first language text, based on the results of associating, by using an analysis module previously prepared, and merges the results of analysis.

7. A third language text generating device according to claim 5 or 6, further including information storing means for storing rule-based information containing at least either dictionary information or grammar information on each language, and empirical information based on the results of learning obtained from actual data in corpora,

wherein at least one of the analyzing means, the converting means and the generating means performs analysis based on the rule-based information and the empirical information stored by the information storing means.

8. A third language text generating device according to claim 5, further including at least either third language information acquiring means for automatically acquiring part or all of information on at least either third language syntax structure information or third language word usage information from an existing third language corpus, or third language information storing means capable of holding the previously automatically acquired information characteristic of the third language,

wherein the generating means generates a third language text based on the information characteristic of the third language.

9. A third language text generating device according to claim 5, wherein the inputting means can input to the device at least one of computer data converted by document text capturing/converting means for converting a document text, such as a sheet of paper or a book, into an electromagnetic record;

computer data read out from an electromagnetic recording device such as a hard disk or an optical storage or memory; and

computer data obtainable from an electromagnetic storage device on a network such as the Internet.

10. A third language text generating program, for use in computer-based language processing, for generating anew third language text by using a plurality of multi-lingual texts, the program including:

an inputting portion which obtains two or more multi-lingual texts written in different languages including a first language which serves as a source language and at least a second language into which the first language is translated, from a storage device or an input device of a computer;

an analyzing portion which performs language analysis including at least dependency analysis and semantic analysis, on each of the obtained multi-lingual texts in the form of each language or a combination of any two or more languages, and obtains language information on at least a dependency structure and semantic representation by arithmetic operation using an arithmetic unit and a storage device of a computer;

a generating portion which generates a third language text by arithmetic operation using the arithmetic unit and the storage device of the computer; and

an outputting portion which outputs the third language text generated by the generating portion by using the storage device or an output device of the computer,

wherein the generating portion generates the third language text by using the language information obtained by the analyzing portion, or

the program further including a converting portion which performs language conversion based on the results of analysis obtained by the analyzing portion or based on the results of analysis and conversion knowledge characteristic of a third language,

wherein the generating portion generates the third language text by using at least either the language information obtained by the analyzing portion or the results of conversion obtained by the converting portion.

11. A third language text generating program according to claim 10, wherein the analyzing portion includes:

an associating routine which performs associating to determine the correspondence between words constructing the multi-lingual texts, the correspondence between phrases constructing the multi-lingual texts, and the correspondence between sentences constructing the multi-lingual texts;

an analysis routine which analyzes at least the first language text; and

a merging routine which analyzes parts in at least the second language text corresponding to the first language text, based on the results of associating, by using an analysis routine, and merges the results of analysis.

12. A third language text generating program according to claim 10 or 11, wherein at least one of the analyzing, converting and generating portions uses rule-based information containing at least either dictionary information or grammar information on each language, and empirical information based on the results of learning obtained from actual data in corpora.

13. A third language text generating program according to claim 10, further including a third language information reading routine which reads out information characteristic of a third language obtained by automatically acquiring part or all of information on at least either third language syntax structure information or third language word usage information from an existing third language corpus,

wherein the generating portion generates a third language text based on the information characteristic of the third language.