Contextualized Attention-based Knowledge Transfer for Spoken Conversational Question Answering

Chenyu You; Nuo Chen; Yuexian Zou

RECORD

Contextualized Attention-based Knowledge Transfer for Spoken Conversational Question Answering

Chenyu You, Nuo Chen, Yuexian Zou

Computation and Language · Submitted 21 Oct 2020

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim We conduct extensive experiments on the Spoken-CoQA dataset and demonstrate that our approach achieves remarkable performance in this task.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2010.11066 (version 4)

Pith IDMN7CI5FA, Pith's permanent record of this paper and its review

CitationChenyu You, Nuo Chen, Yuexian Zou. "Contextualized Attention-based Knowledge Transfer for Spoken Conversational Question Answering." arXiv:2010.11066 (2020). https://pith.science/paper/MN7CI5FA

Claim paper

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

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 4,963 input tokens.

What the abstract itself states, as probabilities: claims first 0.09 · states limitation 0.09 · human participants 0.08 · controlled comparison 0.18 · randomized assignment 0.01 · states quantitative result 0.05 · compares with prior results 0.24 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v4. Results and the main claim are underlined; show every role.

Spoken conversational question answering (SCQA) requires machines to model complex dialogue flow given the speech utterances and text corpora. Different from traditional text question answering (QA) tasks, SCQA involves audio signal processing, passage comprehension, and contextual understanding. However, ASR systems introduce unexpected noisy signals to the transcriptions, which result in performance degradation on SCQA. To overcome the problem, we propose CADNet, a novel contextualized attention-based distillation approach, which applies both cross-attention and self-attention to obtain ASR-robust contextualized embedding representations of the passage and dialogue history for performance improvements. We also introduce the spoken conventional knowledge distillation framework to distill the ASR-robust knowledge from the estimated probabilities of the teacher model to the student. We conduct extensive experiments on the Spoken-CoQA dataset and demonstrate that our approach achieves remarkable performance in this task.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	Spoken conversational question answering (SCQA) requires machines to model complex dialogue flow given the speech utterances and text corpora.	background 1.00	0.06	0.02	asserted 1.00
s1	Different from traditional text question answering (QA) tasks, SCQA involves audio signal processing, passage comprehension, and contextual understanding.	background 1.00	0.06	0.02	asserted 1.00
s2	However, ASR systems introduce unexpected noisy signals to the transcriptions, which result in performance degradation on SCQA.	background 0.91	0.12	0.02	asserted 1.00
s3	To overcome the problem, we propose CADNet, a novel contextualized attention-based distillation approach, which applies both cross-attention and self-attention to obtain ASR-robust contextualized embedding representations of the passage and dialogue history for performance improvements.	method 0.65	0.91	0.13	asserted 0.99
s4	We also introduce the spoken conventional knowledge distillation framework to distill the ASR-robust knowledge from the estimated probabilities of the teacher model to the student.	method 0.76	0.90	0.04	asserted 0.99
s5	We conduct extensive experiments on the Spoken-CoQA dataset and demonstrate that our approach achieves remarkable performance in this task.	result 0.98	0.97	0.86	asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 1.00.

What was examined, and what was not

sources read

arXiv listing abstract v4

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper

Original abstract

Spoken conversational question answering (SCQA) requires machines to model complex dialogue flow given the speech utterances and text corpora. Different from traditional text question answering (QA) tasks, SCQA involves audio signal processing, passage comprehension, and contextual understanding. However, ASR systems introduce unexpected noisy signals to the transcriptions, which result in performance degradation on SCQA. To overcome the problem, we propose CADNet, a novel contextualized attention-based distillation approach, which applies both cross-attention and self-attention to obtain ASR-robust contextualized embedding representations of the passage and dialogue history for performance improvements. We also introduce the spoken conventional knowledge distillation framework to distill the ASR-robust knowledge from the estimated probabilities of the teacher model to the student. We conduct extensive experiments on the Spoken-CoQA dataset and demonstrate that our approach achieves remarkable performance in this task.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

10 arXiv papers cite this in 12 sentences

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

Papers citing it · 6 sentences without a confident label

Cites it as background 6 papers

Recently, the attention-enhanced multi-layer encoder, e.g., Transformer [ref] , ALBERT [ref] , RoBERTa [ref] , and XLNet [ref] , which is based solely on attention mechanisms [ref] and eliminates recurrence entirely, has been proposed and has established the state-of-the-art in multiple challenging MRC datasets [ref] .

Adaptive Bi-directional Attention: Exploring Multi-Granularity Representations for Machine Reading Comprehension · Introduction · 0.90
Prior SCQA systems focus on utilizing pre-trained language models (PLMs) as the backbone models and achieve promising results [ref] .

Self-supervised Dialogue Learning for Spoken Conversational Question Answering · Introduction · 0.82
Spoken language question answering (SQA) [ref] has become increasingly pervasive in our day-to-day lives as we frequently talk to digital assistants on devices including our phones, and smart watches, asking them to retrieve factual information or take actions to help us.

HeySQuAD: A Spoken Question Answering Dataset · Introduction · 0.97

What it cites · 5 sentences without a confident label

Builds on 1 paper

In this paper, based on our previous work [ref] on the spoken conversational question answering task, we present CADNet, a contextual attention-based data distillation network.

Towards Data Distillation for End-to-end Spoken Conversational Question Answering · Introduction · 0.96

Uses the method or tool of 1 paper

Inspired by the previous work [ref] that trains a student model to match the full softmax distribution of the teacher model, we adopt a teacher-student paradigm to distill ASR-robust knowledge into a single student CMRC model.

Distilling the Knowledge in a Neural Network · Knowledge Distillation · 0.93

Uses data or values from 1 paper

We use the listening comprehension benchmark dataset Spoken-CoQA [ref] corpus.

Towards Data Distillation for End-to-end Spoken Conversational Question Answering · Dataset · 1.00

Compares with 3 papers

In this study, we choose four state-of-the-art CMRC models (FlowQA [ref] , SDNet [ref] , BERT-base [ref] , ALBERT-base [ref] ).

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding · Results · 0.84
In this study, we choose four state-of-the-art CMRC models (FlowQA [ref] , SDNet [ref] , BERT-base [ref] , ALBERT-base [ref] ).

FlowQA: Grasping Flow in History for Conversational Machine Comprehension · Results · 0.84
In this study, we choose four state-of-the-art CMRC models (FlowQA [ref] , SDNet [ref] , BERT-base [ref] , ALBERT-base [ref] ).

SDNet: Contextualized Attention-based Deep Network for Conversational Question Answering · Results · 0.84

Cites as background 3 papers

Spoken-SQuAD [ref] is one of the typical benchmarks, which uses CMU Sphinx to generated auto-transcribed text given the Text-SQuAD [ref] dataset.

SQuAD: 100,000+ Questions for Machine Comprehension of Text · Introduction · 0.92
Spoken-SQuAD [ref] is one of the typical benchmarks, which uses CMU Sphinx to generated auto-transcribed text given the Text-SQuAD [ref] dataset.

Spoken SQuAD: A Study of Mitigating the Impact of Speech Recognition Errors on Listening Comprehension · Introduction · 0.92
BERT-base [ref] is a remarkable breakthrough in the natural language processing community, which achieves state-of-the-art performances in many down-stream natural language processing tasks.

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding · Baseline Models · 0.87

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.