Interactive command recognition enhancement system and method

• US 6,792,408 B2
• Filed: 06/12/2001
• Issued: 09/14/2004
• Est. Priority Date: 06/12/2001
• Status: Expired due to Term

• Current Assignee: Dell Products LP (Dell Technologies Inc.)
• Original Assignee: Dell Products LP (Dell Technologies Inc.)
• Inventors: Sauber, William Frederick
• Primary Examiner(s): ABEBE, DANIEL DEMELASH
• Application Number:
  US09/879,207
• Publication Number:
  US 20020188454A1
• Time in Patent Office: 1,190 Days
• Field of Search: 704/251, 704/235, 704/270, 704/275
• US Class Current: 704/275
• CPC Class Codes: G06V 10/98 Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

An interactive command recognition system and method are described. In a preferred embodiment, responsive to a user inputting a command, or word string, to the interactive command recognition system, a physical recognition portion of the system performs physical recognition functions on the input word string and assigns to a number of candidate matches for each of the individual words of the command, a physical score based on the probability that each word was properly recognized by the system, and then computes an average A of these scores. Similarly, a command recognition portion of the system performs command recognition functions on the input command, assigns to each of the words a score based on the probability that the word is part of a recognized command, and then computes an average B of these scores. These averages A and B can then be used in a variety of manners, depending on the particular implementation of the command recognition system. In one implementation, if B is greater than a predetermined logical threshold, the command is executed. If B is less than the predetermined logical threshold and A is greater than a predetermined physical threshold, indicating that the words were but the command was not understood by the system, the user is advised to rephrase the command. In contrast, if both A and B are less than the respective thresholds, indicating that neither the words nor the command was understood by the system, the user is advised to repeat the command more clearly.

Claims

1. A method of increasing the likelihood that an input string is correctly identified by a system, the method comprising:

determining a probability that each word in the input string was correctly identified by the system;

determining a probability that the input string comprises a recognized command;

determining whether the probability that the words were correctly identified by the system is greater than the probability that the input string comprises a recognized command;

responsive to a determination that the probability that the words were correctly identified by the system is greater than the probability that the input string comprises a recognized command, providing feedback to the user intended to elicit a different input string; and

responsive to a determination that the probability that the words were correctly identified by the system is not greater than the probability that the input string comprises a recognized command, providing feedback to the user intended to elicit the same input string spoken more clearly.

2. The method of claim 1 wherein the determining a probability that each word in the input string was correctly identified by the system comprises, assigning to each word of the input string a physical score representing a probability that the word was correctly identified by the system, and using the physical scores to determine a physical recognition probability, and wherein the determining a probability that the input string comprises a recognized command, comprises assigning to each word in the input string a logical score representing a probability that the words form a recognized command and using the logical scores to determine a logical recognition probability.

3. The method of claim 2 wherein the determining whether the probability that the words were correctly identified by the system is greater than the probability that the input string comprises a recognized command, comprises:

determining whether the logical recognition probability is less than a first predetermined threshold;

determining whether the physical recognition probability is less than a second predetermined threshold;

and wherein the providing feedback to the user comprises;

responsive to a determination that the logical recognition probability is less than the first predetermined threshold and the physical recognition probability is not less than the second predetermined threshold, providing feedback to the user intended to elicit a different input string; and

responsive to a determination that the logical recognition probability is less than the first predetermined threshold and the physical recognition probability is less than the second predetermined threshold, providing feedback to the user intended to elicit the same input string spoken more clearly.

4. The method of claim 3 further comprising, if the logical recognition probability is not less than the first predetermined threshold, processing the input string.

5. The method of claim 2 wherein the using the logical scores to determine a logical recognition probability comprises, a technique selected from a group consisting of using the highest logical score as the logical recognition probability;

discarding the highest and lowest logical scores and computing the average of the remaining logical scores;

calculating an average of logical scores for words occurring in a selected command;

calculating an average of logical scores for words occurring in the selected command after discarding the highest and lowest of those logical scores;

computing the mean of the logical scores;

or calculating a difference between the highest and lowest logical scores.

6. The method of claim 5 wherein the using the physical scores to determine a physical recognition probability comprises, a technique selected from a group consisting of calculating an average of physical scores for words occurring in the selected command;

and calculating an average of physical scores for words occurring in the selected command after discarding physical scores for words that are articles.

7. The method of claim 2 wherein the using the physical scores to determine a physical recognition probability comprises, using a technique selected from a group consisting of using the highest physical score as the physical recognition probability;

computing an average of the physical scores;

discarding the highest and lowest physical scores and computing the average of the remaining physical scores; and

computing the mean of the physical scores.

8. The method of claim 2 wherein the determining whether the probability that the words were correctly identified by the system is greater than the probability that the input string comprises a recognized command comprises:

computing a sum of the logical recognition probability and the physical recognition probability;

determining whether the sum is less than a predetermined threshold; and

responsive to a determination that sum is less than the predetermined threshold, comparing the logical recognition probability with the physical recognition probability; and

wherein the providing feedback to the user comprises;

if the logical recognition probability is less than the physical recognition probability, providing feedback to the user intended to elicit a different input string; and

if the logical recognition probability is not less than the physical recognition probability, providing feedback to the user intended to elicit the same input string spoken more clearly.

9. The method of claim 8 further comprising, prior to the determining, weighting the logical recognition probability and the physical recognition probability with first and second weighting constants, respectively.

10. The method of claim 8 further comprising, prior to the comparing, weighting the logical recognition probability and the physical recognition probability with first and second weighting constants, respectively.

11. A method of increasing the likelihood that an input word string comprising a search request was correctly understood by a command recognition system comprising a search engine, the method comprising:

determining whether a match quality for all matches is less than a match quality threshold (“

MQT”

);

if the match quality for all matches is not less than the MQT, outputting to the user the matches with standard advice;

if the match quality for all matches is less than the MQT, determining a probability that the system failed to identify a word in the input word string;

responsive to a determination that the probability that the system failed to identify a word in the input word string is higher than a predetermined recognition threshold, outputting to the user the matches and advising the user to repeat the input word string more clearly; and

responsive to a determination that the probability that the system failed to identify a word in the input word string is not higher than a predetermined recognition threshold, outputting to the user the matches with standard advice.

12. A method of increasing the likelihood that an input string is accurately identified by a system, the method comprising:

evaluating the input string;

determining, based on the evaluating, a likelihood that words comprising the input string were properly recognized by the system;

determining, based on the evaluating, a likelihood that the input string corresponds to a recognized command;

responsive to a determination that the likelihood of the words comprising the input string were properly recognized by the system is lower than a first predetermined threshold, advising a user to repeat the input word string more clearly; and

responsive to a determination that the likelihood that the input string corresponds to a recognized command is lower than a second predetermined threshold, advising the user to rephrase the input word string.

13. The method of claim 12 wherein the input string is input to the system in a form selected from a group consisting of handwritten text and audio.

14. A computer comprising a processor, a memory device connected to the processor, and an enhanced command recognition system, the enhanced command recognition system comprising:

a physical recognition portion for performing physical recognition functions on a word string input to the system and determining a likelihood that the words were properly identified by the physical recognition portion;

a logical recognition portion for performing logical recognition functions on the input word string and determining a likelihood that input the word string forms a recognized command;

means for comparing the likelihood that the words comprising the input word string were properly identified by the system with a first predetermined threshold and for comparing the likelihood that the input word string corresponds to a recognized command with a second predetermined threshold;

means, responsive to a determination that the likelihood of the words comprising the input word string were properly identified by the system is lower than a first predetermined threshold, for advising a user to repeat the input word string more clearly; and

means, responsive to a determination that the likelihood that the input word string corresponds to a recognized command is lower than a second predetermined threshold, for advising the user to rephrase the input word string.

15. The computer of claim 14 wherein the input word string is input to the system in a form selected from a group consisting of handwritten text, typewritten text, and audio.

16. A method of increasing the likelihood that an input string is correctly identified by a system, the method comprising:

determining whether the input string comprises a recognized command;

responsive to a determination that the input string does not comprise a recognized command, determining a probability that each word in the input string was correctly identified by the system;

comparing the probability that each word in the input string was correctly identified by the system with a predetermined threshold;

responsive to a determination that the probability that each word in the input string was correctly identified by the system is greater than the predetermined threshold, providing feedback to the user intended to elicit a different input string; and

responsive to a determination that the probability that each word in the input string was correctly identified by the system is not greater than the predetermined threshold, providing feedback to the user intended to elicit the same input string spoken more clearly.

17. The method of claim 16 wherein the determining a probability that each word in the input string was correctly identified by the system comprises, assigning to each word of the input string a physical score representing a probability that the word was correctly identified by the system, and using the physical scores to determine a physical recognition probability.

18. The method of claim 17 wherein the using the physical scores to determine a physical recognition probability comprises, using a technique selected from a group consisting of using the highest physical score as the physical recognition probability;

computing an average of the physical scores;

discarding the highest and lowest physical scores and computing the average of the remaining physical scores; and

computing the mean of the physical scores.

19. The method of claim 16 further comprising, responsive to a determination that the input string comprises a recognized command, processing the input string.

20. A system for increasing the likelihood that an input string is correctly identified by the system, the system comprising:

means for determining whether the input string comprises a recognized command;

means, responsive to a determination that the input string does not comprise a recognized command, for determining a probability that each word in the input string was correctly identified by the system;

means for comparing the probability that each word in the input string was correctly identified by the system with a predetermined threshold;

means, responsive to a determination that the probability that each word in the input string was correctly identified by the system is greater than the predetermined threshold for providing feedback to the user, for eliciting a different input string; and

means, responsive to a determination that the probability that each word in the input string was correctly identified by the system is not greater than the predetermined threshold for providing feedback to the user, for eliciting the same input string spoken more clearly.

21. The system of claim 20 wherein the means for determining a probability that each word in the input string was correctly identified by the system comprises, means for assigning to each word of the input string a physical score representing a probability that the word was correctly identified by the system, and using the physical scores to determine a physical recognition probability.

22. The system of claim 21 wherein the means for assigning the physical scores to determine a physical recognition probability is selected from a group consisting of means for using the highest physical score as the physical recognition probability;

means for computing an average of the physical scores;

means for discarding the highest and lowest physical scores and computing the average of the remaining physical scores; and

means for computing the mean of the physical scores.

23. The system of claim 20 further comprising means responsive to a determination that the input string comprises a recognized command for processing the input string.