Machine translation split between front end and back end processors

• US 8,886,516 B2
• Filed: 03/01/2012
• Issued: 11/11/2014
• Est. Priority Date: 05/01/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Menezes, Arul A., Quirk, Christopher B., Rashid, Richard F., Moore, Robert C.
• Primary Examiner(s): He, Jialong
• Application Number:
  US13/409,419
• Publication Number:
  US 20120179450A1
• Time in Patent Office: 985 Days
• Field of Search: 704 2- 8
• US Class Current: 704/2
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation

Abstract

A method of translation includes uploading a source text portion to a back end processor. The back end processor identifies a subset of translation knowledge associated with the source text portion. The back end processor downloads the subset to a front end processor. A translation engine runs on the front end processor. The translation engine generates a translation of the source text portion as a function of the subset.

Claims

1. A method of supporting translation of a source text in a source language, the method comprising:

receiving, from a client device, a source text portion in the source language to be translated into a target language;

splitting the source text portion into minimal independent translation segments, each minimal independent translation segment being translatable independent of the other minimal independent translation segments;

for each minimal independent translation segment of the source text portion, identifying a corresponding knowledge statement comprising information that relates words or combinations of words in the minimal independent translation segment, that are in the source language, to corresponding words or combinations of words in the target language;

accessing a stored record that identifies one or more knowledge statements for another source text portion, other than the received source text portion, that have been previously transmitted to the client device;

using a computer having a processor to assemble the identified knowledge statements for the minimal independent translation segments of the source text portion together to form a subset of translation knowledge specific to the source text portion, wherein the subset of translation knowledge is formed based on the stored record; and

remotely transmitting, from the computer, the subset of translation knowledge to the client device to support a translation on the client device, remote from the computer, of the source text portion into the target language as a function of the subset.

2. The method of claim 1, wherein identifying comprises utilizing a phrase-based statistical machine translation system to limit the subset to a limited number of words and phrases.

3. The method of claim 1, wherein identifying comprises limiting the subset to words and phrases identified as being potentially in the source text portion, and wherein limiting includes maintaining a word order of possible translations in the target language.

4. The method of claim 1, wherein receiving comprises utilizing a filter to remove formatting from the source text portion and automatically identifying the source language of the source text portion.

5. The method of claim 1, wherein the source text portion is received by a web service provider.

6. The method of claim 1, wherein receiving the source text portion comprises:

utilizing multiple input filters that convert input information from a non-textual format to a textual format and remove non-translatable formatting from the source text portion.

7. The method of claim 6, wherein the multiple input filters includes one or more of a filter that performs optical character recognition on graphics to generate text and a filter that performs voice-to-text conversion on audio to generate text.

8. The method of claim 6, wherein the corresponding knowledge statement is identified based on a user-specified context.

9. The method of claim 1, wherein the source text comprises a plurality of source text portions received from the client device, the method comprising:

receiving, from the client device, the plurality of source text portions of the source text in the source language to be translated into the target language; and

for each source text portion of the plurality of source text portions,

splitting the source text portion into minimal independent translation segments;

identifying, for each minimal independent translation segment of the source text portion, a corresponding knowledge statement;

assembling the identified knowledge statements for the minimal independent translation segments of the source text portion together to form a subset of translation knowledge specific to the source text portion; and

remotely transmitting the subset of translation knowledge so that it can be utilized to support a translation of the source text portion into the target language as a function of the subset.

10. The method of claim 9, wherein multiple ones of the plurality of source text portions received from the client device are substantially simultaneously processed to form the subsets of translation knowledge for the multiple source text portions.

11. The method of claim 1, wherein the subset of translation knowledge is formed based on the stored record to prevent a duplicate knowledge statement from being transmitted to the client device.

12. The method of claim 11, wherein assembling the identified knowledge statements for the minimal independent translation segments of the source text portion together to form the subset of translation knowledge comprises:

identifying a particular one of the knowledge statements for a minimal independent translation segment of the source text portion as being substantially similar to a knowledge statement identified in the stored record; and

omitting the particular knowledge statement from the subset of translation knowledge.

13. The method of claim 12, and further comprising:

prior to receiving the source text portion,

receiving the other source text portion in the source language to be translated into the target language;

splitting the other source text portion into minimal independent translation segments;

identifying corresponding knowledge statements for the minimal independent translation segments of the other source text portion;

storing the record to identify the corresponding knowledge statements for the minimal independent translation segments or the other source text portion;

forming a subset of translation knowledge specific to the other source text portion; and

remotely transmitting the subset of translation knowledge to the client device to support a translation on the client device of the other source text portion into the target language.

14. A computer readable storage medium storing instructions which, when executed by a computer, perform a method of supporting translation of a source text in a source language, the method comprising:

receiving a source text portion in the source language to be translated into a target language;

splitting the source text portion into a plurality of minimal independent translation segments, each minimal independent translation segment being translatable independent of the other minimal independent translation segments;

accessing a translation knowledge store to identify, for each minimal independent translation segment (MITS) of the source text portion, corresponding translation knowledge specifically associated with the MITS that relates words in the MITS, that are in the source language, to corresponding words in the target language, the translation knowledge specifically associated with the MITS comprising;

phrase translation pairs and their associated scores obtained from the translation knowledge store; and

target language word N-grams and their associated scores obtained from the translation knowledge store;

forming a set of translation knowledge specific to the source text portion based on the translation knowledge identified for each MITS; and

remotely transmitting, over a network, the set of translation knowledge from the computer to support a translation, remote from the computer, of the source text portion into the target language.

15. The computer readable storage medium of claim 14, wherein the translation knowledge for a particular MITS comprises:

all of the phrase translation pairs identified in the translation knowledge store whose source language side matches a substring occurring in the MITS; and

all word N-grams occurring in the translation knowledge store that can be constructed from the target language sides of the phrase translation pairs for the MITS while respecting the word order of those target language sides of the pairs.

16. A mobile device comprising:

one or more user input mechanisms that receive a user input indicative of a source text in a source language to be translated into a target language;

a translation component that splits the source text into a plurality of source text portions, transmits a first one of the source text portions over a network to a remote translation knowledge store, receives a first set of translation knowledge from the remote translation knowledge store that is specific to the first source text portion, and performs a translation process that translates the first source text portion into the target language using the first set of translation knowledge, wherein the translation component transmits a second one of the source text portions over the network to the remote translation knowledge store while performing the translation process on the first source text portion; and

at least one computer processor, being a functional component of the mobile device and activated by the translation component to facilitate transmitting the first source text portion, receiving the first set of translation knowledge, and performing the translation.

17. The mobile device 16, wherein the network comprises a wide area network, and the mobile device comprises:

a local translation knowledge store that stores translation knowledge that is not specific to the source text; and

wherein, to translate the source text, the translation component uses translation knowledge from both the local translation knowledge store and the remote translation knowledge store.

18. The mobile device 17, wherein, for a second one of the source text portions, the translation component determines whether the local translation knowledge store includes translation knowledge for translating the second source text portion, and, if not, receives a second set of translation knowledge from the remote translation knowledge store that is specific to the second source text portion and uses the second set of translation knowledge to translate the second source text portion.

19. The mobile device 16, wherein the translation component communicates with the remote translation knowledge store by interleaving the plurality of source text portions with sets of translation knowledge received from the remote translation knowledge store.