Interactive task-sensitive assistant

• US 7,890,336 B2
• Filed: 01/13/2003
• Issued: 02/15/2011
• Est. Priority Date: 01/13/2003
• Status: Active Grant

• Current Assignee: Northwestern University
• Original Assignee: Northwestern University
• Inventors: Hammond, Kristian, Birnbaum, Lawrence, Chen, Leonard
• Primary Examiner(s): Ouellette; Jonathan
• Application Number:
  US10/341,154
• Publication Number:
  US 20040138899A1
• Time in Patent Office: 2,955 Days
• Field of Search: 705/1, 705/1.1
• US Class Current: 705/1.1
• CPC Class Codes: G06Q 10/10 Office automation; Time management

Abstract

A system and method in accordance with the present invention include means for providing interactive assistance for the performance of a set of predefined steps, including selecting the set of predefined steps and automatically generating a step-sensitive grammar for each step. Generating the step-sensitive grammar includes generating a set of navigation commands related to each step and generating a set of rules to recognize potential queries related to each step. A recognizer is configured for determining if a received utterance forms one of the navigation commands or one of the potential queries, within a context of the current step. Form this determination, provided are navigation to a different step if the utterance was a navigation command or a response if the utterance was a query.

Claims

1. A machine-executed method of providing interactive assistance for the performance of a set of predefined steps, said method comprising:

A. accessing data related to the set of predefined steps, and for each of one or more steps from said set of predefined steps, automatically generating a grammar, including;

1) generating a set of navigation commands related to the step, based on a context of the step; and

2) generating a set of rules to recognize potential queries related to the step, based on the context of the step;

B. providing a recognizer configured for determining if a received vocal utterance corresponds to one of said set of generated navigation commands or one of said potential queries related to a current step according to the generated rules, within a context of said current step; and

C. navigating to a different step from the current step if said vocal utterance is a navigation command corresponding to at least one of the generated navigation commands or providing a response if said vocal utterance is a query corresponding to one of said potential queries according to the generated rules.

2. A machine-executed method of generating step-sensitive grammars for a recipe, including:

A. parsing the recipe into a set of steps;

B. selecting a recipe step from the set of steps; and

C. generating a grammar comprising a set of rules, including;

1) determining if the recipe step includes an ingredient reference and, if so, adding to said grammar a rule that recognizes the ingredient in a query;

2) determining if the recipe step includes a technique reference and , if so, adding to said grammar a rule that recognizes the technique in said query;

3) determining if the recipe step includes a tool reference and, if so, adding to said grammar a rule that recognizes the tool in said query; and

4) repeating

     1),

     2), and

     3) for each ingredient, technique, and tool referenced in said recipe step;

D. repeating steps B and C for each step in said recipe.

3. The method of claim 2 further comprising:

E. providing a recognizer configured for determining if a received vocal utterance corresponds to one of the generated set of rules related to a current step, within a context of said current step; and

F. if said vocal utterance corresponds to one of the generated rules related to the current step, providing a response to the vocal utterance based on a type of the rule.