Correction device to mark parts of a recognized text

• US 20030110030A1
• Filed: 10/09/2002
• Published: 06/12/2003
• Est. Priority Date: 10/12/2001
• Status: Active Grant

• Current Assignee: Nuance Communications Austria Gmbh (Microsoft Corporation)
• Original Assignee: Koninklijke Philips Electronics N.V. (Koninklijke Philips N.V.)
• Inventors: Rajic, Kresimir, Gschwendtner, Wolfgang
• Granted Patent: US 6,708,148 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

In a correction device (3, 4) for correcting a text (ET) recognized by a speech recognition device (2) from a spoken text (GT), marking means (12, 15, 17) are provided for storing marking information (WMI) in memory means (8), which marking information (WMI) flags those parts of the recognized text (ET) and/or of the spoken text (GT) that were reproduced at least once by reproduction means (13) in a synchronous reproduction mode. In this way, parts of the recognized text (ET) and the spoken text (GT) that have already been reproduced once can be acoustically or visually marked, and different parts can be reproduced depending on whether reproduction takes place for the first time or is repeated.

Claims

1. A correction device (3, 4) for correcting a text (ET) recognized by a speech recognition device (2) from a spoken text (GT), wherein an item of link information (LI) for each part of the spoken text (GT) flags the associated recognized text (ET),

having memory means (8) for storing at least the spoken text (GT) and the recognized text (ET), and

having reproduction means (13) for acoustically reproducing the spoken text (GT) and synchronously visually marking the associated recognized text (ET) flagged by the item of link information (LI) when a synchronous reproduction mode is activated in the correction device (3, 4), and

having marking means (12, 15, 17) for storing marking information (WMI) in the memory means (8), which marking information (WMI) flags those parts of the recognized text (ET) and/or the spoken text (GI) that were reproduced at least once by the reproduction means (13) when the synchronous reproduction mode was activated.

2. A correction device (3, 4) as claimed in claim 1, wherein the reproduction means (13) are arranged to suppress the acoustic reproduction of unwanted parts of the spoken text (GT) when the synchronous reproduction mode is activated, which unwanted parts are flagged by suppression information (PMI, HMI, GMI, DMI, RMI) stored in the memory means (8).

3. A correction device (3, 4) as claimed in claim 2, wherein the reproduction means (13) are arranged to acoustically reproduce the unwanted parts of the spoken text (GT) flagged by the suppression information (PMI, HMI, GMI, DMI, RMI), if these unwanted parts are also flagged by the marking information (WMI) when the synchronous reproduction mode is activated.

4. A correction device (3, 4) as claimed in claim 2, wherein unwanted parts of the spoken text (GT) flagged by the suppression information (PMI, HMI, RMI) are formed by a pause in the spoken text (GT), repeated words or a hesitating sound.

5. A correction device (3, 4) as claimed in claim 1, wherein the reproduction means (13) are arranged for the visual marking of those parts of the recognized text (ET) that are flagged by the marking information (WMI).

6. A correction device (3, 4) as claimed in claim 1, wherein positioning means (18) are provided for positioning a text cursor (P2) identifying a text input position when the synchronous reproduction mode is interrupted to correct an incorrectly recognized word, and wherein the positioning means (18) are arranged to position the text cursor N words upstream of the word of the recognized text (ET) that was marked at the time of interruption of the synchronous reproduction mode if this part of the recognized text (ET) is flagged by the marking information (WMI), and wherein the positioning means (18) are arranged to position the text cursor M words upstream of the word of the recognized text (ET) that was marked at the time of interruption of the synchronous reproduction mode if this part of the recognized text (ET) is not flagged by any marking information (WMI).

7. A method of correcting a text (ET) recognized by a speech recognition device (2) from a spoken text (GT), wherein an item of link information (LI) for each part of the spoken text (GT) flags the associated recognized text (ET) and wherein the following steps are performed:

storage of at least the spoken text (GT) and the recognized text (ET);

when the synchronous reproduction mode is activated, acoustic reproduction of the spoken text (GT) and synchronous visual marking of the associated recognized text (ET) flagged by the link information (LI);

storage of marking information (WMI), which marking information (WMI) flags those parts of the recognized text (ET) and/or of the spoken text (GT) that were reproduced at least once before when the synchronous reproduction mode was activated.

8. A method of correction as claimed in claim 7, wherein the following further steps are performed:

suppression of the acoustic reproduction of unwanted parts of the spoken text (GT) when the synchronous reproduction mode is activated, which unwanted parts are flagged by stored suppression information (PMI, HMI, GMI, DMI, RMI).

9. A method of correction as claimed in claim 8, wherein the following further steps are performed:

acoustic reproduction of the unwanted parts of the spoken text (GI) flagged by the suppression information (PMI, HMI, GMI, DMI, RMI) when the synchronous reproduction mode is activated if such parts are additionally flagged by the marking information (WMI).