Domain State Tracking for a Simplified Dialogue System
This addresses efficiency issues in task-oriented dialogue systems for developers and users, though it is incremental.
The paper tackles the problem of high memory and computational costs in neural dialogue systems by proposing DoTS, which uses a simplified input context instead of the entire dialogue history, improving inform and success rates by 1.09 and 1.24 points on MultiWOZ.
Task-oriented dialogue systems aim to help users achieve their goals in specific domains. Recent neural dialogue systems use the entire dialogue history for abundant contextual information accumulated over multiple conversational turns. However, the dialogue history becomes increasingly longer as the number of turns increases, thereby increasing memory usage and computational costs. In this paper, we present DoTS (Domain State Tracking for a Simplified Dialogue System), a task-oriented dialogue system that uses a simplified input context instead of the entire dialogue history. However, neglecting the dialogue history can result in a loss of contextual information from previous conversational turns. To address this issue, DoTS tracks the domain state in addition to the belief state and uses it for the input context. Using this simplified input, DoTS improves the inform rate and success rate by 1.09 points and 1.24 points, respectively, compared to the previous state-of-the-art model on MultiWOZ, which is a well-known benchmark.