Contact Center Intent Matching With Confidence-Driven Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training an intent matching engine for contact centers is costly in person-hours due to the manual generation of queries and associated intents, necessitating more efficient approaches.
Innovation Solution
A method involving a contact center server that generates a training dataset by manually reviewing a subset of queries based on confidence scores and timestamps, selecting queries 'almost' matched to intents, and using online learning techniques to refine the engine.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual generation of queries and intents is used for training the intent matching engine, then the training data can be accurately labeled, but the process becomes costly in person-hours
Solution Approach 1:
The system uses the intent matching engine itself to generate training data by automatically labeling queries with intents based on confidence scores, allowing the system to train itself without requiring manual annotation for all data points
Solution Approach 2:
The system performs preliminary automated labeling of queries using the intent matching engine before manual review, so that only a subset of queries requiring manual attention needs to be processed by human annotators, reducing overall manual effort
2Measurement precision
If all queries are manually reviewed for training, then training data quality is maximized, but the manual effort required increases significantly
Solution Approach 1:
The system applies different levels of review to different queries based on their confidence scores - high confidence queries are automatically accepted while low confidence queries undergo manual review, so that manual effort is concentrated only where needed
Solution Approach 2:
Instead of manually reviewing all queries, the system performs partial manual review only on queries that fall into specific confidence score ranges or categories, achieving sufficient training quality without the excessive manual effort of complete review
3Device complexity
If the intent matching engine uses fixed training data, then the training process is simple, but the engine cannot adapt to real-world changes
Solution Approach 1:
The system implements dynamic training where the intent matching engine continuously learns from new queries and feedback in real-time, allowing the model to adapt to changing user behavior and language patterns without requiring complete retraining
Solution Approach 2:
The system incorporates feedback loops where the intent matching engine receives feedback on its predictions and automatically adjusts its parameters, enabling continuous improvement and adaptation to real-world conditions
Data Source
AI summary
Respective confidence scores for a query with respect to respective intents are determined. An indication associated with the query is provided to a client device for review based on a determination that the query does not match any of the respective intents based on the respective confidence scores. Data representing an intent of the respective intents that matches the query is received from the client device. An intent matching engine is trained to associate the query with the intent based on the received data.