Dialogue System Filler Utterance for Continuity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional spoken dialogue systems face difficulties in accurately understanding and responding to users' varied and complex utterances, leading to uncomfortable interactions and potential dialogue failures when they cannot generate appropriate responses.
Innovation Solution
A dialogue method and system that employs a two-step approach: first, presenting a dialogue-establishing utterance that does not include content words, followed by a second utterance associated with the user's previous utterances, to maintain a continuous conversation and reduce user discomfort, even when appropriate responses cannot be generated.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a conventional spoken dialogue system attempts to generate appropriate responses to user utterances, then the system can provide meaningful interaction, but the system may fail to respond appropriately when it cannot accurately understand the user's utterance, leading to dialogue failure
Solution Approach 1:
The response generation is divided into two independent modules: a semantic response generator that handles meaningful responses, and a filler response generator that handles dialogue-maintaining responses. This segmentation allows the system to switch between modules based on understanding confidence, ensuring dialogue continuity without requiring a single complex module to handle all cases perfectly.
Solution Approach 2:
A response selection module acts as an intermediary between the user's utterance and the final response. It evaluates the user's utterance, determines whether meaningful or filler responses are more appropriate, and selects from candidate responses generated by different modules. This intermediary ensures that the system maintains reliable dialogue continuity by preventing inappropriate responses that would cause dialogue failure.
2Reliability
If the system uses only meaningful responses based on accurate understanding, then the interaction quality is high, but the dialogue may fail when understanding is inaccurate
Solution Approach 1:
The system dynamically adjusts its response strategy based on the confidence level of utterance understanding. When understanding confidence is high, it generates meaningful semantic responses; when confidence is low, it switches to generating filler responses that maintain dialogue flow. This dynamic adaptation ensures both response appropriateness and dialogue reliability across varying understanding conditions.
Solution Approach 2:
The system changes the parameter of response type (from semantic to filler) based on the understanding confidence parameter. By monitoring the confidence level and adjusting the response generation strategy accordingly, the system maintains adaptability to different dialogue situations while ensuring response appropriateness through confidence-based selection.
3Duration of action of stationary object
If the system generates filler responses to maintain dialogue, then dialogue continuity is improved, but the interaction becomes less meaningful
Solution Approach 1:
The system periodically alternates between filler responses (for dialogue maintenance) and semantic responses (for information exchange) based on the dialogue state and understanding confidence. This periodic switching ensures that the dialogue maintains continuity through filler responses while still achieving information exchange through semantic responses, balancing dialogue duration with information quality.
Solution Approach 2:
The system ensures continuous useful action by maintaining dialogue flow through filler responses when semantic responses are not feasible, while still pursuing information exchange through semantic responses when possible. The response selection module continuously monitors the dialogue state and switches between response types to maintain both dialogue continuity and information exchange utility.
Data Source
AI summary
The present invention keeps a dialogue continuing for a long time without causing uncomfortable feeling to a user. A dialogue system 10 includes at least an input part 1 that receives a user utterance, which is an utterance from a user and a presentation part 5 that presents the utterance. The input part 1 receives a user utterance performed by the user. A presentation part 5-1 presents a dialogue-establishing utterance which does not include any content words. A presentation part 5-2 presents a second utterance associated with a generation target utterance, which is one or more utterances performed before user utterance that includes at least the user utterance, after the dialogue-establishing utterance.