Computerized Assistant Constraint Updates Without Full Data Annotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computerized assistants face challenges in efficiently updating constraints for actions, particularly due to the time-consuming and technically challenging process of annotating training data for machine learning models.
Innovation Solution
A method is introduced to adapt a computerized assistant program by maintaining a dialogue history, recognizing initial and new constraints, and using a revision function statement to generate a revised program that satisfies the new constraint.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning models are used to provide assistive functionality, then the computerized assistant can respond to user utterances with suitable actions, but training requires large quantities of annotated training data which is extremely time consuming and technically challenging
Solution Approach 1:
The patent uses copying by having human annotators provide simplified revision function statements rather than complete annotated training data. The revision function statements copy the essential constraint modification logic without requiring full program annotations, significantly reducing the annotation effort while still enabling the machine learning model to learn constraint updates effectively
Solution Approach 2:
The patent extracts only the essential constraint modification information from complete training data by using revision function statements. Instead of annotating entire training programs, the system extracts and annotates only the constraint revision portions, which are the critical components for teaching the model to update actions according to new constraints
2Reliability
If human annotators annotate training data for machine learning models, then the model can be trained to satisfy constraints, but the process is extremely time consuming and technically challenging
Solution Approach 1:
The patent simplifies the annotation process by having human annotators create revision function statements that copy the constraint modification pattern rather than creating complete annotated training examples. This reduces the technical challenge and time required while maintaining the reliability of constraint satisfaction through the structured revision function format
Solution Approach 2:
The patent changes the parameter of annotation granularity from complete training program annotations to constraint revision level annotations. By operating at the constraint revision level rather than the full program level, the system reduces the complexity and time required for annotation while still achieving reliable constraint satisfaction through the revision function statements
Data Source
AI summary
A method of adapting a computerized assistant program to satisfy an updated constraint. The method comprises maintaining a dialogue history including a first utterance that indicates an initial constraint. The method further comprises receiving a second utterance indicating a new constraint that conflicts with the initial constraint. The method further comprises recognizing a revision function statement parametrized by a reference to an initial computerized assistant program configured to satisfy the initial constraint, and a reference to the new constraint. The method further comprises executing instructions derived from the revision function statement to return a revised computerized assistant program that is configured to satisfy the new constraint.