Computer-implemented method for automatic training of a dialogue system, and dialogue system for generating semantic annotations

• US 9,679,557 B2
• Filed: 04/24/2014
• Issued: 06/13/2017
• Est. Priority Date: 04/25/2013
• Status: Active Grant

• Current Assignee: Elektrobit Automotive GmbH (Continental AG)
• Original Assignee: Elektrobit Automotive GmbH (Continental AG)
• Inventors: Weilhammer, Karl, Goronzy-Thomae, Silke
• Primary Examiner(s): Siddo, Ibrahim
• Application Number:
  US14/260,824
• Publication Number:
  US 20140324429A1
• Time in Patent Office: 1,146 Days
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding

Abstract

An adaptive dialogue system and also a computer-implemented method for semantic training of a dialogue system are disclosed. In this connection, semantic annotations are generated automatically on the basis of received speech inputs, the semantic annotations being intended for controlling instruments or for communication with a user. For this purpose, at least one speech input is received in the course of an interaction with a user. A sense content of the speech input is registered and appraised, by the speech input being classified on the basis of a trainable semantic model, in order to make a semantic annotation available for the speech input. Further user information connected with the speech input is taken into account if the registered sense content is appraised erroneously, incompletely and/or as untrustworthy. The sense content of the speech input is learned automatically on the basis of the additional user information.

Claims

1. A computer-implemented method for automatic training of a dialogue system, in order to generate semantic annotations automatically on the basis of a received speech input, the method comprising the following steps:

receiving, by the dialogue system, at least one speech input in the course of an interaction with a user;

generating, by the dialogue system, a symbolic representation from the speech input by performing recognition on the received speech input;

registering and appraising, by the dialogue system, a sense content of the symbolic representation of the received speech input, by the symbolic representation being classified on the basis of a trainable semantic model, in order to make a semantic annotation available automatically for the received speech input;

controlling, by the dialogue system, a vehicle component based on the semantic annotation;

receiving, by the dialogue system, additional user information representing a function selection of a vehicle component by the user;

automatic learning of the sense content of the received speech input, by the dialogue system, on the basis of the additional user information, if a connection between the reception of the speech input and the additional user information is determined on the basis of a temporal and semantic correlation, or semantic correlation,

repeating iteratively, by the dialogue system, the receiving of additional user information and automatic learning of the sense content of the received speech input until a termination condition is satisfied.

2. The method according to claim 1, further including:

training the semantic model on the basis of at least one of the received speech input, the learned sense content of the received speech input and a corresponding transcription into a semantic annotation.

3. The method according to claim 1, wherein the learning step includes at least one of the following steps:

establishing the sense content of the additional user information;

comparing the received speech input with the sense content of the additional user information; and

successively deriving or completing the sense content of the received speech input on the basis of a known sense content of the additional user information.

4. The method according to claim 1, wherein the learning includes at least one of the following procedures:

completing an incompletely classified received speech input;

correcting a misinterpreted part of the received speech input; and

reclassifying an incorrectly classified or non-classified received speech input.

5. The method according to claim 1, wherein the receiving the additional user information includes deriving the additional user information from at least one user input that immediately follows the reception of the speech input.

6. The method according to claim 5, wherein the at least one user input includes at least one correcting control input which corrects a preceding control command made available on the basis of the speech input.

7. The method according to claim 5, wherein the user inputs are based on acoustic or haptic actions of the user.

8. A non-transitory computer readable medium storing a program causing a computer to execute the method according to at least one of the preceding claims.

9. An adaptive dialogue system for use in a vehicle, which has been designed to generate semantic annotations automatically on the basis of received speech inputs, comprising a processor executing a program to function as:

an input interlace which is designed to receive at least one speech input in the course of an interaction with a user;

recognition device which is designed to generate a symbolic representation from the speech input by performing recognition on the received speech input; and

a semantic classifying device which is designed to register and classify the sense content of the symbolic representation of the received speech input on the basis of a trainable semantic model, to make a semantic annotation available automatically for the received speech input on the basis of the classification,

an output interface which is designed to output a signal based on the semantic annotation to a vehicle component in order to control the vehicle component, wherein the input interface is further designed to receive additional user information representing a function selection of a vehicle component by the user,

to automatically learn the sense content of the received speech input on the basis of the additional user information, if a connection between the reception of the speech input and the additional user information is determined on the basis of a temporal and semantic correlation or a semantic correlation, and

to repeat iteratively receiving of additional user information and automatic learning of the sense content of the received speech input until a termination condition is satisfied.

10. The adaptive dialogue system according to claim 9, further including a training device which is provided to train the semantic model on the basis of at least one of the learned sense content of a received speech input, the received speech input, and a transcription into a semantic annotation.

11. The adaptive dialogue system according to claim 9, further including a storage device which is provided to store, entirely or partly, at least one of the received speech input, the learned sense content of the received speech input and a transcription into a semantic annotation.

12. The adaptive dialogue system according to claim 9, wherein the semantic model has been designed in the form of a neural network or a decision tree.

13. The adaptive dialogue system according to claim 9, further including a communications device which is provided to send to a server or to retrieve from a server the received speech input together with at least one of the learned sense content of the received speech input and with the transcription.

14. The adaptive dialogue system according to claim 9, wherein the adaptive dialogue system is part of a human/machine interface.

15. A control device for a vehicle, including the dialogue system according to claim 9.