Speech recognition method

• US 6,073,102 A
• Filed: 04/21/1997
• Issued: 06/06/2000
• Est. Priority Date: 04/19/1996
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Block, Hans-Ulrich
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US08/839,922
• Time in Patent Office: 1,142 Days
• Field of Search: 395/11, 395/12, 395/61, 704/275, 704/270, 704/200, 704/272
• US Class Current: 704/275
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

Proceeding from a command or request that a user (B) gives a speech recognition system, an action is first determined from a prescribable plurality of actions upon employment of action indicators that the action information pertains to. A first set of actions is thereby determined and stored wherein all action indicators coincide with the key concepts. Further, a second set of actions is determined and stored wherein at least a part of the action indicators coincide with the key concepts. The first set of actions and/or the second set of actions is employed as a basis for the unambiguous determination of the action. For the at least one determined action, the action parameters respectfully allocated to the action are determined upon employment of a parameter parser (PP). The sought action is implemented in a last step.

Claims

1. A speech recognition system which identifies an action to be taken upon receipt of a command that is given to the system by a user'"'"'s speech, comprising:

(a) at least one speech recognition unit EM via which the command is received and recognized,

(b) an input memory ESP in which the command is stored,

(c) an action determination unit (AE) which identifies at least one action which is indicated by an action indicator, the action determination unit (AE) is being constructed such that (1) a first set of actions is formed in which the actions are stored wherein all action indicators coincide with at least one part of prescribable key concepts that are allocated to an action, (2) that a second set of actions is formed in which the actions are stored wherein at least a part of the action indicators coincide with at least the part of the prescribable key concepts, and (3) the first set of actions and/or the second set of actions is employed as a basis for the unambiguous determination of the action indicated by the command, and

(d) a parameter parser (PP) with which action parameters prescribable from the command that are allocated to an action and are employed for the execution of the action are identified.

2. The system according to claim 1, further comprising a tactile input unit.

3. The system according to claim 2, whereby a means (GF) for generating questions for the user is provided for determining further action indicators and/or further action parameters.

4. The system according to claim 1 or claim 2,

(a) whereby for that instance wherein the first set contains exactly one action, this action is the determined action,

(b) whereby an action clarification unit (AKE) is provided for determining at least a further action indicator from the user for that case wherein the first group does not contain at exactly one action, and

(c) whereby the action clarification unit (AKE) is constructed such that, when the first group contains no action, the determination of the further action indicator ensues taking the actions contained in the second group into consideration.

5. The system according to claim 1 or 2, whereby a parameter clarification unit (PKE) is provided for determining further action parameters from the user.

6. The system according to claim 1 or 2, whereby a key work editor (SE) is provided for inputting key concepts by the user.

7. The system according to claim 1 or 2, whereby an information data bank (ID) is provided in which information is determined on the basis of the action and of the action parameters.

8. The system according to one claim 1 or 2, whereby at least one output unit (AUS) is provided for the output of information.

9. The system according to claim 1 or 2, whereby a control unit (STE) is provided for controlling a background application (HA) upon employment of the identified information.

10. The system of claim 1 wherein the system is a telephone service.

11. The system of claim 1 wherein the system is a voice mail handling system.

12. A method for recognizing speech and determining for an action to be undertaken upon receipt of a command in digital format that comprises at least one action indicator, using a computer, comprising the steps of:

(a) receiving and recognizing speech of a user,

(b) recognizing a command in said speech,

(c) providing the a command to the computer,

(d) storing the a command in a memory,

(e) selecting from a group of prescribable actions to which a prescribable plurality of action parameters are allocated, at least one action to be undertaken on the basis of the at least one action indicator in the command,

(f) storing a first set of actions whose action indicators coincide with key concepts and a second set of actions wherein at least a part of the action indicators coincide with the key concepts, and

(g) employing the first set of actions and/or the second set of actions as a basis for the unambiguous determination of the action,

(h) determining for the action the action parameters that are allocated to the action and that are employed for the execution of the action, and

(i) implementing the determined action.

13. The method according to claim 12, whereby the information for the action are determined on the basis of the action parameters and of an information data bank (ID).

14. The method according to claim 12 or 13, whereby the command is spoken by the user, and whereby the command is recognized in a speech recognition unit.

15. The method according to claim 12 or 13, whereby the command is input by the user with a tactile input unit.

16. The method according to claim 12 or 13, whereby, when the action indicator is not sufficient for unambiguous determination of an action, further action indicators are determined in a prescribable dialog with the user stored in the computer.

17. The method according to claim 12 or 13, whereby, when the action parameters allocated to the action where not completely identified for the specific action, the remaining action parameters are determined in a prescribable dialog with the user stored in the computer.

18. The method according to claim 12 or 13, whereby the determination of the action ensues in such a way that the at least one action indicator is compared to key concepts that were unambiguously allocated to the actions and the action is determined from the prescribable plurality of actions whose key concepts coincide best with the action indicators with respect to a similarity degree.

19. The method according to one of the claim 12 or 13, whereby the action parameters from the command that were stored in the input memory (ESP) are determined in the parameter parser (PP) in reverse sequence with respect to the input.

20. The method according to claim 12 or 13, whereby a background application (HA) is controlled with the identified information.