SFL-MTSC: Leveraging Semantic Frame-Level Multi-Task Self-Consistency for Robust Multi-Intent Spoken Language Understanding

2026-06-24

AI Summary

This work addresses the inconsistency between predicted intents and slot structures in multi-intent spoken language understanding caused by the decoding stochasticity of large language models. To resolve this, the authors propose a semantic frame-level multi-task self-consistency aggregation method that decomposes inputs into intent-specific semantic frames, groups them by domain-intent categories, and clusters associated slots to generate candidate structured outputs. A path support score is introduced to select high-confidence frames, which are then used to reconstruct coherent, structured predictions. This approach achieves multi-task self-consistency at the semantic frame level for the first time, replacing conventional output-level voting mechanisms and significantly enhancing robustness in multi-intent scenarios. Experiments on the MAC-SLU benchmark demonstrate substantial improvements in slot F1 score and overall accuracy while maintaining stable intent recognition performance.

Abstract

Prompt-based spoken language understanding (SLU) with large language models (LLMs) often suffers from inconsistent intent--slot structures due to decoding stochasticity, particularly in multi-intent scenarios. In view of this, we propose Semantic Frame-Level Multi-Task Self-Consistency (SFL-MTSC), a novel structured aggregation framework operating at the semantic frame level. Instead of output-level majority voting, SFL-MTSC decomposes predictions into intent-specific frames, applies domain--intent grouping and slot-level clustering, and evaluates cluster reliability using path support scoring. Reliable frames are retained and re-integrated to form the final prediction. Zero-shot experiments on the MAC-SLU benchmark dataset show improved slot F1 and overall accuracy over single-path inference, while intent accuracy remains largely stable across most settings.

Problem

• spoken language understanding
• multi-intent
• intent-slot inconsistency
• decoding stochasticity
• large language models

Innovation

• Semantic Frame-Level Aggregation
• Multi-Intent SLU
• Self-Consistency
• Slot Clustering
• Path Support Scoring