Browse Prior Art Database

Synthesizing a new translation using existing translations.

IP.com Disclosure Number: IPCOM000231491D
Original Publication Date: 2013-Oct-02
Included in the Prior Art Database: 2013-Oct-02
Document File: 2 page(s) / 8K

Publishing Venue

Lenovo

Related People

Spencer Guy: INVENTOR [+6]

Abstract

The quality of a computer-generated language translation can vary wildly, and is subject to double meanings in the text. This technique describes how software could improve the quality of computer-generated translations when the text already exists in multiple languages.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 67% of the total text.

Page 01 of 2

Synthesizing a new translation using existing translations

Disclosed is a method for producing a more accurate computer-generated translation using a base text that already exists in other languages.

For projects of smaller scopes, hiring a translator may be too expensive, and simply putting one phrase through an online translator may create confusion where double meanings are involved.

While some software is already translated into multiple languages, using a standard translating service will only take into account one base statement. For example, say some software already supports English, Spanish, and French, and its developers want to support Japanese. One way to get this translation would be to translate the English text into Japanese using an online translator. However, the quality of this translation could vary wildly due to heteronyms (same spelling, different pronunciation and meaning) in the text.

Synthesizing a new translation using existing translations could bypass some of these double meanings. Instead of translating a single language's text into the target language, software could translate every available string using an existing online translation service (English to Japanese, Spanish to Japanese, French to Japanese). The output will be a series of strings in the target language, which can be synthesized to output a more educated guess as to the best translation. The most common translation is then selected, ideally weeding out any one-off trans...