Automated Candidate Training Utterance Selection for AI Assistants
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional AI systems rely on human subject matter experts to update training datasets, which is inefficient due to the high rate of incoming utterances exceeding human supervision capabilities, necessitating an automated method to identify and select candidate training utterances.
Innovation Solution
The method involves generating n-grams from live utterances, assigning unique identifiers, creating binary classification models using SVMs, and determining candidate training utterances by analyzing the difference score between historical live utterance vector representations and the model's line of demarcation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human subject matter experts manually update training datasets, then the training data quality can be maintained, but the system cannot keep up with the high rate of incoming utterances
Solution Approach 1:
The patent introduces an automated intermediary system consisting of n-gram generation, vector representation, and classification models that mediates between incoming utterances and the training dataset. This intermediary automatically processes utterances at high speed while maintaining quality through algorithmic selection criteria, resolving the contradiction between manual quality control and automated processing speed.
Solution Approach 2:
The patent replaces the mechanical human expert review process with an automated computational system. Human experts traditionally manually reviewed and selected training utterances, but this system substitutes that mechanical process with automated n-gram analysis, vector representation, and classification algorithms that can process utterances at the high rates required by modern voice response systems.
2Speed
If AI systems process live incoming utterances at high speed, then the system responsiveness is improved, but the ability of human experts to supervise ongoing learning is exceeded
Solution Approach 1:
The patent enables the AI system to self-service its own training data needs by automatically identifying, processing, and selecting candidate training utterances without human intervention. The system generates n-grams from incoming utterances, creates vector representations, applies classification models, and identifies candidate training data autonomously, allowing high-speed processing while maintaining supervised learning capabilities through automated self-management.
3Productivity
If automated methods are used to identify candidate training utterances, then the processing efficiency is improved, but the complexity of the system increases
Solution Approach 1:
The patent segments the complex task of training data identification into distinct modular components: n-gram generation, vector representation creation, classification model application, and candidate selection. This segmentation allows each component to be independently optimized and managed, improving overall processing efficiency while making the system complexity more manageable through modular architecture.
Data Source
AI summary
A method for creating binary classification models and using the binary classification models to select candidate training utterances from a plurality of live utterances is provided. The method may include receiving a plurality of intents and associated training utterances. The method may include creating, from the training utterances, a binary classification model for each intent. The binary classification model may include a vector representation of a line of demarcation between utterances associated with the intent and utterances disassociated from the intent. The method may also include receiving live utterances. An intent may be determined for each live utterance. The method may include creating a vector representation of the live utterance. The method may include selecting candidate training utterances based on a comparison between the vector representation of the live utterance and the vector representation included in the binary classification model of the intent determined for the live utterance.