USER-SPECIFIED PHRASE INPUT LEARNING

• US 20100153091A1
• Filed: 12/11/2008
• Published: 06/17/2010
• Est. Priority Date: 12/11/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kung, Yun-Chiang, Lee, Chin-Nan, Shih, Sheng-Yao, Huang, Yu-Li, Liu, Hsiang-Fu
• Granted Patent: US 9,009,591 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/53 Processing of non-Latin text kana-to-kanji conversion G06F40/129; vowelisation G06F40/232

Abstract

Architecture that enables a user to perform manual word-breaking by phrase input. Phrase input is where the user inserts a phrase-key (or separator) as a delimiter that indicates to an editor application such as an IME (input method editor) the composition of a specific phrase when entering characters (e.g., Asian). The word-breaking is controlled by the user. The conversion quality is improved as the user knows the desired input and ambiguous cases are reduced. A phrase can be specified while the user is composing the characters. By selecting a phrase-key separator, the user can specify the composing characters before the characters are presented as a phrase. Moreover, the architecture includes a phrase prioritization mechanism wherein each phrase can be treated as a single entity and assigned a character identifier (ID), which is related to the sequence of a candidate list.

Claims

1. A computer-implemented language editing system, comprising:

an editor component for receiving and processing a string of language objects of a written language; and

an input component for inserting a user-specified separator into the string of the language objects to break language objects of the string into a phrase recognizable as a grammatical unit.

2. The system of claim 1, wherein the written language is an East Asian language.

3. The system of claim 1, further comprising a decoding component for selecting an optimum path in a language model based on the phrase.

4. The system of claim 1, wherein the separator is manually inserted as the language objects are entered.

5. The system of claim 1, wherein the editor component applies emphasis to the phrase to designate the phrase as disambiguated.

6. The system of claim 1, wherein the editor component applies an associated emphasis to a deconstructed phrase to designate the deconstructed phrase as ambiguated.

7. The system of claim 1, further comprising a learning component for learning the phrase based on multiple uses of the separator to create the phrase.

8. The system of claim 1, further comprising a phrase prioritization component for ranking the phrase among other candidate phrases and automatically suggesting a ranked phrase for insertion into the ambiguous string.

9. A computer-implemented language editing system, comprising:

an editor component for receiving and processing a string of language objects of a written language;

an input component for manually inserting a user-specified separator into the string of the language objects to break language objects of the string into a phrase recognizable as a grammatical unit; and

a learning component for learning the phrase based on multiple uses of the separator to create the phrase.

10. The system of claim 9, further comprising a decoding component for selecting an optimum path in a language model based on the phrase.

11. The system of claim 9, further comprising a phrase prioritization component for ranking the phrase among other candidate phrases and automatically suggesting a ranked phrase for insertion into the ambiguous string.

12. The system of claim 9, wherein the written language is an East Asian language.

13. The system of claim 9, wherein the editor component applies emphasis to the phrase to designate the phrase as disambiguated, and to a deconstructed phrase to designate the deconstructed phrase as ambiguous.

14. A computer-implemented language editing method, comprising:

receiving language objects of a written language into an editor;

processing the language objects as determined and undetermined phrases as the language objects are entered;

processing a manually-entered separator inserted into an undetermined phrase of the language objects as a word-break; and

interpreting language objects relative to the separator as a determined phrase.

15. The method of claim 14, wherein the language objects are part of an East Asian language.

16. The method of claim 14, further comprising:

automatically applying a first emphasis to the undetermined phrase as a visual cue that the phrase is undetermined; and

automatically applying a second emphasis to the determined phrase as a visual cue that the phrase is determined.

17. The method of claim 14, further comprising specifying language objects before the separator as the determined phrase.

18. The method of claim 14, further comprising learning the determined phrase for subsequent interpretation of phrases input as language objects in the editor.

19. The method of claim 14, further comprising automatically toggling between mode typing and modeless typing based on selection of a keyboard key assigned to the separator.

20. The method of claim 14, further comprising fixing an optimum path in a language model runtime using lattice decoding of a phrase input via the editor.