Effectiveness of Text, Acoustic, and Lattice-Based Representations in Spoken Language Understanding Tasks

Effectiveness of Text, Acoustic, and Lattice-Based Representations in Spoken Language Understanding Tasks

Stať ve sborníku v databázi WoS či Scopus

In this paper, we perform an exhaustive evaluation of different representations to address the intent classification problem in a Spoken Language Understanding (SLU) setup. We benchmark three types of systems to perform the SLU intent detection task: 1) text-based, 2) lattice-based, and a novel 3) multimodal approach. Our work provides a comprehensive analysis of what could be the achievable performance of different state-of-the-art SLU systems under different circumstances, e.g., automatically- vs. manuallygenerated transcripts. We evaluate the systems on the publicly available SLURP spoken language resource corpus. Our results indicate that using richer forms of Automatic Speech Recognition (ASR) outputs, namely word-consensus-networks, allows the SLU system to improve in comparison to the 1-best setup (5.5% relative improvement). However, crossmodal approaches, i.e., learning from acoustic and text embeddings, obtains performance similar to the oracle setup, a relative improvement of 17.8% over the 1-best configuration, being a recommended alternative to overcome the limitations of working with automatically generated transcripts.

In this paper, we perform an exhaustive evaluation of different representations to address the intent classification problem in a Spoken Language Understanding (SLU) setup. We benchmark three types of systems to perform the SLU intent detection task: 1) text-based, 2) lattice-based, and a novel 3) multimodal approach. Our work provides a comprehensive analysis of what could be the achievable performance of different state-of-the-art SLU systems under different circumstances, e.g., automatically- vs. manuallygenerated transcripts. We evaluate the systems on the publicly available SLURP spoken language resource corpus. Our results indicate that using richer forms of Automatic Speech Recognition (ASR) outputs, namely word-consensus-networks, allows the SLU system to improve in comparison to the 1-best setup (5.5% relative improvement). However, crossmodal approaches, i.e., learning from acoustic and text embeddings, obtains performance similar to the oracle setup, a relative improvement of 17.8% over the 1-best configuration, being a recommended alternative to overcome the limitations of working with automatically generated transcripts.

Speech Recognition, Human-computer Interaction, Spoken Language Understanding, Word Consensus Networks, Cross-modal Attention

Speech Recognition, Human-computer Interaction, Spoken Language Understanding, Word Consensus Networks, Cross-modal Attention

VILLATORO-TELLO, E.; MADIKERI, S.; ZULUAGA-GOMEZ, J.; SHARMA, B.; SARFJOO, S.; NIGMATULINA, I.; MOTLÍČEK, P.; IVANOV, V.; GANAPATHIRAJU, A.

2024

04.06.2023

IEEE Signal Processing Society

Rhodes Island

978-1-7281-6327-7

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

1

5

5

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10095168

villatoro-tello_icassp2023_10095168