Chatbot Intent Disambiguation Using Mapped Confidence Options
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current chatbots struggle to accurately disambiguate user utterances when multiple intents have similar confidence scores, leading to incorrect interpretations and the need for additional clarification questions or user exit from conversations.
Innovation Solution
A chatbot system that presents a list of top matching user intent options to the user for disambiguation, using configurable confidence score thresholds and a user intent mapping table to improve intent recognition through user feedback and autolearning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the chatbot uses natural language understanding to analyze user requests, then the chatbot can respond with coded responses or conversations, but the chatbot struggles to accurately disambiguate user utterances when multiple intents have similar confidence scores
Solution Approach 1:
The system implements feedback by presenting multiple possible intent interpretations to the user and receiving their selection. This feedback loop allows the chatbot to resolve ambiguity by using the user's choice to disambiguate their utterance, thereby improving intent recognition accuracy while maintaining conversation continuity without requiring additional clarification questions.
Solution Approach 2:
The patent introduces an intermediary mechanism that acts as a bridge between the ambiguous user utterance and the chatbot's intent recognition system. By presenting a list of possible intents as an intermediate step, the system allows users to select the correct interpretation, effectively mediating the disambiguation process and resolving the contradiction between accuracy and reliability.
2Measurement precision
If the chatbot presents multiple intent options to the user for disambiguation, then user intent prediction accuracy is enhanced, but the interaction requires additional user input steps
Solution Approach 1:
The system applies preliminary action by proactively presenting multiple intent options to the user before the chatbot needs to respond. This advance presentation of choices allows users to clarify their intent in a single interaction step, preventing the need for multiple clarification turns and reducing overall conversation time while improving intent prediction accuracy.
3Measurement precision
If the chatbot requests additional clarification questions to resolve ambiguous intents, then intent accuracy improves, but user experience deteriorates due to increased interaction complexity
Solution Approach 1:
Instead of the traditional approach where the chatbot asks clarification questions, this patent inverts the process by presenting possible intent interpretations to the user and having them select the correct one. This inversion transforms the clarification process from a question-answering sequence into a selection task, which is more intuitive and easier for users, thereby improving ease of operation while maintaining high intent recognition accuracy.
Data Source
AI summary
Disambiguating user utterances is provided. Disambiguation of an utterance of a user is performed using up to a defined number of user intents in a set of possible user intents having highest confidence scores between a first confidence score threshold level and a second confidence score threshold level in response to determining that each user intent in the set of possible user intents does have a corresponding confidence score less than the second confidence score threshold level. Each of the up to the defined number of user intents is located in a user intent mapping table. A human interpretable label corresponding to each of the up to the defined number of user intents located in the user intent mapping table is extracted. Extracted human interpretable labels are assembled into a set of user intent options. The set of user intent options is sent to a client device of the user.