Systems and methods for generating and implementing an interactive man-machine web interface based on natural language processing and avatar virtual agent based character

• US 8,156,060 B2
• Filed: 02/27/2008
• Issued: 04/10/2012
• Est. Priority Date: 02/27/2008
• Status: Active Grant

• Current Assignee: Inteliwise SP ZOO
• Original Assignee: Inteliwise SP ZOO
• Inventors: Borzestowski, Marek, Trojanowicz, Marek, Strzalkowski, Marcin
• Primary Examiner(s): Vincent, David R
• Assistant Examiner(s): Kim, David H
• Application Number:
  US12/038,628
• Publication Number:
  US 20090216691A1
• Time in Patent Office: 1,504 Days
• Field of Search: 706/11, 706/45, 715/706
• US Class Current: 706/45
• CPC Class Codes: G06F 16/3332 Query translation
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

An interactive virtual agent (avatar) system includes a plurality of databases configured to store at least one of client specific information and non-client specific information, internal resources configured to provide access to information sources, the character interface configured to receive information from a user, and respond to the user with information based on the received information, a processing engine coupled with the client knowledge database, the basic knowledge database, the internal resources, and the character interface.

Claims

1. An interactive virtual agent (avatar) system, comprising:

a plurality of databases configured to store at least one of client specific information and non-client specific information;

internal resources configured to provide access to information sources;

a character interface configured to receive information from a user, and respond to the user with information based on the received information;

a knowledge model configured to define relations and connections between a plurality of knowledge bases, which comprise a plurality of knowledge layers that define paths from inputs to responses, wherein the plurality of knowledge bases comprise a combination of a general use knowledge base comprising knowledge layers that are used to determine the composition of a received input and a path leading from a question to answer, a corrective action knowledge base configured to correct mistakes in a received input, a generalization knowledge database configured to draw conclusions based on the parsed input, and a multiplicative knowledge base configured to weight relations between the outputs of various knowledge layers;

a processing engine comprising a computer processor that can be coupled with the client knowledge database, the basic knowledge database, the internal resources, and the character interface, the processing engine configured to;

receive input from the character interface;

parse the input using the knowledge model to determine the content of the input;

break the parsed input into sentences and words for matching into known notions and words using the knowledge model;

correct spelling and other mistakes for unmatched words and to weight possible variants of words in the parsed input;

use relations to transform words in the parsed input into notions and notions into other notions and then degrading the weighting of the transformed notions

determine relations that apply to the parsed content;

perform at least one of deduction, generalization, context expansion and synonym finding on the content using the determined relations;

using the knowledge model, and at least one of the plurality of databases, the internal resources, and external resources, determine an appropriate response to the input; and

provide the response to the character input to be output to the user.

2. The interactive virtual agent (avatar) system of claim 1, wherein the knowledge layers included in the general use knowledge base comprise a combination of a word knowledge layer configured to determine the words included in a received input, a notion knowledge layer configured to determine the notions included in the received input, a patterns knowledge layer configured to control a process of deduction engaged in by the processing engine, a response group know ledge layer configured to group responses into units, collect different versions of responses, a response knowledge layer configured to manage response statements, and a response component knowledge layer configured to manage reactions related to the response statements that can be portrayed to a user via the character interface.

3. The interactive virtual agent (avatar) system of claim 1, wherein the knowledge layers included in the corrective knowledge base comprise a combination of a additional word knowledge layer configured to identify misspelled or incorrect words and a set of relations configured to map misspelled or incorrect words to correct words.

4. The interactive virtual agent (avatar) system of claim 3, wherein mapping misspelled words comprises mapping misspelled or incorrect words onto a basic form and mapping a basic form onto many variants.

5. The interactive virtual agent (avatar) system of claim 1, wherein the knowledge layers included in the generalization knowledge base comprise a combination of relations including that allow the processing engine to draw conclusions and determine notions including synonym, anonym, superior clause, and subordinate clause.

6. The interactive virtual agent (avatar) system of claim 1, wherein the processing engine is further configured to match the notions to responses based on the weightings.

7. The interactive virtual agent (avatar) system of claim 6, wherein the processing engine is further configured to perform a net weighting of a selected response based on the individual weighting of the individual component responses.

8. The interactive virtual agent (avatar) system of claim 7, wherein the processing engine is further configured to select a response from a group of candidate responses based on the associated response net weighting.

9. The interactive virtual agent (avatar) system of claim 1, wherein the plurality of databases comprises a client knowledge database that is configured to store information specific to one or more clients.

10. The interactive virtual agent (avatar) system of claim 1, wherein the plurality of databases comprises a basic knowledge database that is configured to store information that is not specific to one or more clients.

11. The interactive virtual agent (avatar) system of claim 1, wherein the internal resources include at least one of a website, Really Simple Syndication (RSS) feeds, and information from an encyclopedic website.

12. The interactive virtual agent (avatar) system of claim 1, wherein the character interface comprises a text to speech processor, and wherein the character interface is configured to provide the output in a spoken language format using the text-to-speech processor.

13. The interactive virtual agent (avatar) system of claim 12, wherein the character interface is a visual character interface configured to respond to spoken input with spoken output.

14. The interactive virtual agent (avatar) system of claim 13, wherein the visual interface is a video character interface.

15. The interactive virtual agent (avatar) system of claim 13, wherein the visual interface is a cartoon character interface.

16. The interactive virtual agent (avatar) system of claim 1, wherein the character interface is configured to receive spoken input, and wherein the processing engine comprises a natural language processor configured to receive the spoken input from the character interface and process the spoken input.