Unsupervised clustering of dialogs extracted from released application logs

Patent metadata

Field	Value
Publication number	US-9606984-B2
Application number	US-201313969825-A
Country	US
Kind code	B2
Filing date	Aug 19, 2013
Priority date	Aug 19, 2013
Publication date	Mar 28, 2017
Grant date	Mar 28, 2017

Abstract

A natural language understanding system performs automatic unsupervised clustering of dialog data from a natural language dialog application. A log parser automatically extracts structured dialog data from application logs. A dialog generalizing module generalizes the extracted dialog data to generalization identifier vectors. A data clustering module automatically clusters the dialog data based on the generalization identifier vectors using an unsupervised density-based clustering algorithm without a predefined number of clusters and without a predefined distance threshold in an iterative approach based on a hierarchical ordering of the generalization.

First claim

What is claimed is:

1. A natural language understanding system using at least one hardware implemented computer processor for automatic unsupervised clustering of dialog data from a natural language dialog application, the arrangement comprising:
a log parser configured to extract structured dialog data from application logs, the structured dialog data including a transcription of a dialog between a user and the natural language dialog application, the transcription being generated by the natural language dialog application;
a dialog generalizing module configured to automatically generalize the extracted dialog data using different independent generalization methods to produce generalization identifier vectors aggregating the results of the generalization methods used, the generalization identifier vectors indicating descriptive categories of the different independent generalization methods that correspond to statements by the user and statements by the natural language dialog application included in the extracted dialog data; and
a data clustering module configured to automatically cluster the dialog data based on the generalization identifier vectors using an unsupervised density-based clustering algorithm without a predefined number of clusters and without a predefined distance threshold.

2. The system according to claim 1 , further comprising:
a dialog information database configured to store the clustered dialog data.

3. The system according to claim 1 , wherein the generalization identifier vectors include sequences of application state identifiers characterizing internal transition of the state of the dialog application.

4. The system according to claim 1 , wherein the data clustering module further post-processes the clustered dialog data to add additional cluster characteristic information.

5. The system according to claim 1 , wherein the clustering algorithm flattens hierarchic clusters of dialog data.

6. The system according to claim 1 , wherein the clustering algorithm is an iterative clustering algorithm.

7. A computer-implemented method using at least one hardware implemented computer processor for automatic unsupervised clustering of dialog data from a natural language dialog application, the method comprising:
automatically generalizing structured dialog data extracted from application logs using different independent generalization methods to produce generalization identifier vectors aggregating the results of the generalization methods used, the structured dialog data including a transcription of a dialog between a user and the natural language dialog application, the transcription being generated by the natural language dialog application, the generalization identifier vectors indicating descriptive categories of the different independent generalization methods that correspond to statements by the user and statements by the natural language dialog application included in the extracted dialog data; and
automatically clustering the dialog data based on the generalization identifier vectors using an unsupervised density-based clustering algorithm without a predefined number of clusters and without a predefined distance threshold.

8. The method according to claim 7 , further comprising:
storing the clustered dialog data in a dialog information database.

9. The method according to claim 7 , wherein the generalization identifiers include sequences of application state identifiers characterizing internal transition of the state of the dialog application.

10. The method according to claim 7 , post-processing the clustered dialog data to add additional cluster characteristic information.

11. The method according to claim 7 , wherein the clustering algorithm flattens hierarchic clusters of dialog data.

12. The method according to claim 7 , wherein the clustering algorithm is an iterative clustering algorithm.

13. A computer program product encoded in a non-transitory computer-readable medium for automatic unsupervised clustering of dialog data from a natural language dialog application, the product comprising:
program code for automatically generalizing structured dialog data extracted from application logs using different independent generalization methods to produce generalization identifier vectors aggregating the results of the methods used, the structured dialog data including a transcription of a dialog between a user and the natural language dialog application, the transcription being generated by the natural language dialog application, the generalization identifier vectors indicating descriptive categories of the different independent generalization methods that correspond to statements by the user and statements by the natural language dialog application included in the extracted dialog data; and
program code for automatically clustering the dialog data based on the generalization identifier vectors using an unsupervised density-based clustering algorithm without a predefined number of clusters and without a predefined distance threshold.

14. The product according to claim 13 , further comprising:
program code for storing the clustered dialog data in a dialog information database.

15. The product according to claim 13 , wherein the generalization identifiers include sequences of application state identifiers characterizing internal transition of the state of the dialog application.

16. The product according to claim 13 , program code for post-processing the clustered dialog data to add additional cluster characteristic information.

17. The product according to claim 13 , wherein the clustering algorithm flattens hierarchic clusters of dialog data.

18. The product according to claim 13 , wherein clustering algorithm is an iterative clustering algorithm.

Assignees

• Nuance Communications Inc

Inventors

• Lavallée Jean-Francois

Classifications

• G06F40/35 Primary

  Discourse or dialogue representation · CPC title

• G10L2015/0631

  Creating reference templates; Clustering · CPC title

• G10L15/063

  Training · CPC title

• G10L2015/0633

  using lexical or orthographic knowledge sources · CPC title

• G06F17/279 Primary

  Physics · mapped topic