Voice input word processor

• US 6,338,035 B1
• Filed: 12/28/1998
• Issued: 01/08/2002
• Est. Priority Date: 12/26/1997
• Status: Expired due to Term

• Current Assignee: Acer, Inc.
• Original Assignee: NEC Corporation
• Inventors: Mori, Ichiro
• Primary Examiner(s): {haeck over (S)}mits, Ta̅livaldis Ivars
• Application Number:
  US09/220,769
• Time in Patent Office: 1,107 Days
• Field of Search: 704/231, 704/251, 704/235
• US Class Current: 704/235
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A voice-input document creation system which reduces correction times to a wrongly-recognized input to ensure input efficiency. The system has a determination means for comparing the feature content of a newest voice input extracted by a feature extracting module with a feature content of an immediately preceding voice input to determine if the newest voice input is a correction to the immediately preceding voice input. When a first-time correction is received, the system displays a list of all output candidates for the immediately preceding voice input stored in a second memory. When a second-time correction is received, the system stores the output candidates for the newest voice input into a third memory and, at the same time, displays the output candidates; at this time, the system neither displays nor stores in the third memory those output candidates displayed upon the first-time correction.

Claims

1. A voice input document creation system having a speech recognition module for recognizing a plurality of output candidates in response to a voice input entered on a clause or word basis the system comprising:

means for comparing a feature content of a newest voice input with a feature content of an immediately preceding voice input to determine if the newest voice input is a correction to the immediately preceding voice input,

wherein, upon a first-time correction, a list of all output candidates for the immediately preceding voice input is displayed,

and wherein, upon a second-time correction with the list of all output candidates displayed, a list of output candidates for the newest voice input is displayed, said list of output candidates for the second-time correction excluding output candidates displayed upon the first-time correction.

2. A voice input document creation system having a speech recognition module for recognizing a plurality of output candidates in response to a voice input entered on a clause or word basis, the system comprising:

means for comparing a feature content of a newest voice input with a feature content of an immediately preceding voice input to determine if the newest voice input is a correction to the immediately preceding voice input; and

means for displaying a list of all output candidates for the immediately preceding voice input when the newest voice input is determined to be the correction to the immediately preceding voice input on the assumption that the input was retried, and when the same voice input is entered again with the list of all output candidates displayed, for displaying the list of output candidates for the newest voice input excluding output candidates from the first displayed list of output candidates.

3. A voice input document creation system comprising:

input means for receiving speech;

feature extracting means for extracting a feature content of a voice input received via the input means;

a first memory in which the feature content of the voice input is saved;

a second memory in which at least one output candidate for a newest voice input is saved;

determination means for comparing the feature content of the newest voice input with the feature content of a voice input immediately preceding the newest voice input saved in the first memory and for determining if the newest voice input is a correction to a voice input immediately preceding the newest;

comparison means for comparing a feature content of the newest voice input with a feature content of each clause or word stored in a recognition dictionary to select at least one output candidate;

a third memory in which said at least one output candidate for a second-time correction is saved; and

display means,

wherein, when the determination means determine that the newest voice input is the correction to the voice input immediately preceding the newest and when a list of output candidates is not yet displayed, all recognized results of the voice input immediately preceding the newest are displayed on the display means as a list of output candidates, said recognized results being saved in the secondary memory, and

wherein, when the determination means determine that the newest voice input is the correction and when the list of output candidates is already displayed, a determination is made that the displayed list of output candidates does not include a desired clause or word and the list of output candidates is replaced by a list of output candidates for the newest voice input, and

wherein, upon a first-time correction, a list of all output candidates for the voice input immediately before the newest is displayed, said output candidates being saved in the second memory, and

wherein, upon a second-time correction, output candidates for the newest voice input are saved in the third memory and a list thereof is displayed, the output candidates displayed upon the first-time correction not being included in said list, nor being saved in said third memory.

4. A voice input document creation system as defined by claim 3 wherein, when said determination means determines that the voice input is not a correction, the feature content of said newest voice input is saved in said first memory for use upon entering of next voice input and, at the same time, all recognized results of the newest voice input are saved in said second memory and wherein a top-priority output candidate is displayed at a sentence input location on the display means.

5. A voice input document creation system as defined by claim 3 wherein, when said determination means determines that the voice input is not a correction and the list of output candidates is already displayed, a determination is made that of output candidates does not include a desired clause or word and, when the list of output candidates is replaced by the list of the newest voice input, the recognized results already saved as the recognized results of the original voice input are not displayed.

6. A recording medium having stored therein a program which causes a computer to execute a voice input document creation system having a speech recognition module for recognizing a plurality of output candidates in response to a voice input entered on a clause or word basis, the program comprising instructions for:

(a) comparing a feature content of a newest voice input with a feature content of a voice input immediately before the newest to determine if the newest voice input is a correction to the voice input immediately before the newest; and

(b) displaying a list of all output candidates for the voice input immediately before the newest when the newest voice input is determined to be a correction to the voice input immediately before the newest on the assumption that the input be retired; and

when the same voice input is subsequently entered again with the list of all output candidates displayed, displaying the list of output candidates for the newest voice input excluding output candidates from the first displayed list of output candidates.