Speech recognition method with a replace command

• US 20020046032A1
• Filed: 09/04/2001
• Published: 04/18/2002
• Est. Priority Date: 09/08/2000
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Koninklijke Philips Electronics N.V. (Koninklijke Philips N.V.)
• Inventors: Wutte, Heribert
• Granted Patent: US 7,027,985 B2
• Time in Patent Office: Days
• US Class Current: 704/270
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A speech recognition method (4) with a text editing method of editing a text information part (TW) erroneously recognized in first text information (TI1) with a four-word spoken command (SB) recognized in second text information (TI2), the speech recognition method comprising the following steps:

receiving speech information (SI) and recognizing the first text information (TI1) and the second text information (TI2);

checking whether a word at the first position in the second text information (TI2) matches a first spoken command part (SB1) of the four-word spoken command (SB) and checking whether a word at the third or further position in the second text information (TI2) matches a third spoken command part (SB3) of the four-word spoken command (SB) while, in the event of a positive result of the checks, the next step of the text editing method is executed;

checking whether the first text information (TI1) contains the erroneously recognized text information part (TW) which is featured by a second spoken command part (SB2) in the second text information (TI2) between the first spoken command part (SB1) and the third spoken command (SB3) while, in the event of a positive result of the check, the next step of the text editing method is executed;

replacing the erroneously recognized text information part (TW) in the first text information TI1 with new text information part (TN) which the second text information (TI2) contains as a fourth spoken command part (SB4) of the four-word spoken command (SB) after the third spoken command part (SB3).

Claims

1. A speech recognition method (4) with a text editing method for editing a text information part (TW) erroneously recognized in first text information (TI1) with a four-word spoken command (SB) recognized in second text information (TI2), the speech recognition method (4) comprising the following steps:

receiving speech information (SI) and recognizing the first text information (TI1) and the second text information (TI2);

checking whether a word at the first position in the second text information (TI2) matches a first spoken command part (SB1) of the four-word spoken command (SB) and checking whether a word at the third or further position in the second text information (TI2) matches a third spoken command part (SB3) of the four-word spoken command (SB) while, in the event of a positive result of the checks, the next step of the text editing method is executed;

checking whether the first text information (TI1) contains the erroneously recognized text information part (TW) which is featured by a second spoken command part (SB2) in the second text information (TI2) between the first spoken command part (SB1) and the third spoken command (SB3) while, in the event of a positive result of the check, the next step of the text editing method is executed;

replacing the erroneously recognized text information part (TW) in the first text information TI1 with new text information part (TN) which the second text information (T12) contains as a fourth spoken command part (SB4) of the four-word spoken command (SB) after the third spoken command part (SB3).

2. A speech recognition method (4) as claimed in claim 1 in which after the replacement of the erroneously recognized text information part (TW), a cursor (C) featuring the entry position for further text information (TI) recognized by the speech recognition method (4) is positioned at the end of the most recently recognized first text information (TI1).

3. A speech recognition method (4) as claimed in claim 1, in which the four-word spoken command (SB) has the structure “

REPLACE A WITH B” and

in which the second spoken command part “

A”

features an erroneous recognized word and the fourth spoken command part “

B”

features a word actually to be recognized instead of the erroneously recognized word.

4. A speech recognition method as claimed in claim 1, in which the following steps are provided:

determining alternative text information parts (AT1) for each spoken information part (STI) of the received spoken information (SI) when the second text information (TI2) is recognized, while the alternative text information parts (ATI) contained in the second text information (TI2) have the highest overall probability of a correct recognition of the second text information (TI2),

checking whether the first text information (TI1) contains the erroneously recognized text information part (TW) that is characterized by one of the alternative text information parts (ATI) of the second spoken command part (SB2) not contained in the second text information (TI2), and if the check produces a positive result, the erroneously recognized text information part (TW) is replaced with the new text information part (TN).

5. A speech recognition method as claimed in claim 1, in which the following further steps are provided:

determining alternative text information parts (ATI) for each spoken information part (STI) of the received spoken information (SI) when the second text information (TI2) is recognized, while the alternative text information parts (ATI) contained in the second text information (TI2) have the highest overall probability of a correct recognition of the second text information (TI2);

checking whether in the erroneously recognized text information part (TW) contained in the first text information (TI1) matches with the new text information (TN) contained in the second text information (TI2) while, in the event of a positive result of the check, an alternative text information part (ATI) determined for the new text information (TN) is stored as new text information (TN).

6. A computer program product which can be loaded directly into the internal memory of a digital computer (1) and comprises software code sections, while the computer (1) executes the steps of the speech recognition method (4) as claimed in claim 1 when the product runs on the computer (1).

7. A computer program product as claimed in claim 6, with the computer program product being stored on a computer readable medium.