RECORD
A Study of Different Ways to Use The Conformer Model For Spoken Language Understanding
Computation and Language · Submitted 8 Apr 2022
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim We find that it is not necessarily a choice between two-stage decoding and end-to-end systems which determines the best system for research or application.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
SLU combines ASR and NLU capabilities to accomplish speech-to-intent understanding. In this paper, we compare different ways to combine ASR and NLU, in particular using a single Conformer model with different ways to use its components, to better understand the strengths and weaknesses of each approach. We find that it is not necessarily a choice between two-stage decoding and end-to-end systems which determines the best system for research or application. System optimization still entails carefully improving the performance of each component. It is difficult to prove that one direction is conclusively better than the other. In this paper, we also propose a novel connectionist temporal summarization (CTS) method to reduce the length of acoustic encoding sequences while improving the accuracy and processing speed of end-to-end models. This method achieves the same intent accuracy as the best two-stage SLU recognition with complicated and time-consuming decoding but does so at lower computational cost. This stacked end-to-end SLU system yields an intent accuracy of 93.97% for the SmartLights far-field set, 95.18% for the close-field set, and 99.71% for FluentSpeech.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | SLU combines ASR and NLU capabilities to accomplish speech-to-intent understanding. | background 1.00 | 0.05 | 0.02 | asserted 1.00 |
| s1 | In this paper, we compare different ways to combine ASR and NLU, in particular using a single Conformer model with different ways to use its components, to better understand the strengths and weaknesses of each approach. | contribution 0.95 | 0.79 | 0.05 | asserted 0.85 |
| s2 | We find that it is not necessarily a choice between two-stage decoding and end-to-end systems which determines the best system for research or application. | interpretation 0.67 | 0.94 | 0.56 | asserted 0.91 |
| s3 | System optimization still entails carefully improving the performance of each component. | interpretation 0.91 | 0.42 | 0.17 | asserted 1.00 |
| s4 | It is difficult to prove that one direction is conclusively better than the other. | interpretation 0.79 | 0.64 | 0.18 | asserted 0.94 |
| s5 | In this paper, we also propose a novel connectionist temporal summarization (CTS) method to reduce the length of acoustic encoding sequences while improving the accuracy and processing speed of end-to-end models. | contribution 0.94 | 0.93 | 0.07 | asserted 0.97 |
| s6 | This method achieves the same intent accuracy as the best two-stage SLU recognition with complicated and time-consuming decoding but does so at lower computational cost. | result 0.98 | 0.93 | 0.18 | asserted 1.00 |
| s7 | This stacked end-to-end SLU system yields an intent accuracy of 93.97% for the SmartLights far-field set, 95.18% for the close-field set, and 99.71% for FluentSpeech. | result 1.00 | 0.95 | 0.14 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
SLU combines ASR and NLU capabilities to accomplish speech-to-intent understanding. In this paper, we compare different ways to combine ASR and NLU, in particular using a single Conformer model with different ways to use its components, to better understand the strengths and weaknesses of each approach. We find that it is not necessarily a choice between two-stage decoding and end-to-end systems which determines the best system for research or application. System optimization still entails carefully improving the performance of each component. It is difficult to prove that one direction is conclusively better than the other. In this paper, we also propose a novel connectionist temporal summarization (CTS) method to reduce the length of acoustic encoding sequences while improving the accuracy and processing speed of end-to-end models. This method achieves the same intent accuracy as the best two-stage SLU recognition with complicated and time-consuming decoding but does so at lower computational cost. This stacked end-to-end SLU system yields an intent accuracy of 93.97% for the SmartLights far-field set, 95.18% for the close-field set, and 99.71% for FluentSpeech.
Human review
How the literature uses it
What this paper does with the arXiv papers it cites
What it cites
Uses data or values from 1 paper
-
Snips SmartLights [ref] , another corpus of SLU data for spoken commands to smart homes, contains both close-field and far-field acoustic recording conditions.
Spoken Language Understanding on the Edge · Datasets · 0.97
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.