TAPAS: Training Adapted Personalised Affective Social Skills with - - PowerPoint PPT Presentation
TAPAS: Training Adapted Personalised Affective Social Skills with - - PowerPoint PPT Presentation
ANR-CREST 2109 TAPAS: Training Adapted Personalised Affective Social Skills with Cultural Virtual Agents JP: PI Satoshi Nakamura, Ph.D (NAIST) Hiroki Tanaka, Ph. D (NAIST) Takashi Kudo, M.D (Osaka University) Hidemi Iwasaka, M.D (Nara Medical
Research Topics at AHC-lab, NAIST
Speech Translation Machine Translation Spoken Dialogue Multi-modal Dialogue Why don’t you join our lab! I’m looking for a lab. Data Analytics Caption Generation Multi-language ASR, TTS Machine Speech Chain Deep Neural Network
Affective Computing
①Social Skills Training ②Detection of Early Dementia Natural Language Processing Goal-oriented Dialog Non goal-oriented Dialogue
Brain Analysis
Incongruity measurement Cognitive Load EEG Hyper Scanning
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
2
- 1. Social Skills Training
Training tools for autism spectrum disorders People with autism have good systemizing skills Speaking skills [Tanaka, et al., 2015] and listening skills [Tanaka, et al., 2019]
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
3
Role-play
Analyze behaviors of users and generate avatar actions
4
Features based on [Tanaka et al., 2014]
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
Summary feedback
5
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
After role-play, the system displays feedback Repetitive training until mastery
Training effect
One-way ANOVA: (F[2,24]=4.70, p<.05) Post hoc comparisons with Bonferroni’s method: significantly different between feedback and reading book (p<0.05)
6
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
Human-human and human-agent interaction
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
7
- 2. Detection of early dementia
10 minutes interaction: 12 early stage of dementia and healthy control
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
8
[Tanaka et al., 2017] [Ujiro et al., 2018]
Multimodal-based classification
Answers to fixed questions: e.g. what is the date today? Areas under the ROC curve – 0.90 (SVM), 0.88 (Logistic regression)
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
9
Features: response gap, language, speech, face
ANR-CREST (2019-2024)
TAPAS: Training Adapted Personalised Affective Social Skills with Cultural Virtual Agents
JP: PI Satoshi Nakamura, Ph.D (NAIST) Hiroki Tanaka, Ph. D (NAIST) Takashi Kudo, M.D (Osaka University) Hidemi Iwasaka, M.D (Nara Medical University) FR: PI Catherine Pelachaud, Ph.D (CNRS-ISR, Sorbonne U) Jean-Claude Martin, Ph.D (CNRS-LIMSI)
Research background and objectives
Increase of people who are not good at social communication and social anxiety disorder in schools and workplace (3 to 13%)
11
Social anxiety disorder Public speech: Feel nervous and stress for attentions by others, and bad reputation ↓ Face red, sweat, shaking, pulsation, breathing, stomachache Fear for the symptom Social anxiety, avoidance Nervous, shameful
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
Try to solve by verbal / non-verbal interactive training system by Embodied Conversational Agent
– Target population: general population, depression, autism spectrum disorders, schizophrenia – Behavioral training: apply Social Skills Training (SST) – Cognitive training: apply Cognitive Behavioral Therapy (CBT)
Overall framework
12
Nakamura G NAIST Satoshi Nakamura,Hiroki Tanaka Hirokazu Kato,Yuichiro Fujimoto Iwasaka G Nara Med. Univ. Hidemi Iwasaka,Yasuhiro Matsuda, Kosuke Okazaki Kudo G Osaka Univ. Takashi Kudo,Hiroyoshi Adachi Yukako Sakagami Pelachaud G (France CNRS-ISIR) Catherine Pelachaud Donatella Simonetti MARTIN G (France CNRS-LIMSI) Jean-Claude MARTIN
Elise Prigent Ouriel Grynszpan
France Japan Culture difference
Teenagers, children with ASD
WP4
VR/AR/Display
- ne to one one to multi-party
Behavioral training Cognitive therapy
Agent’s eye gaze, attitude (e.g. cooperative) Multimodal behaviors
Through human-agent interaction
No stress, high self-efficacy Learn appropriate behaviors
Chat-based
AttrakDiff questionnaire Anxiety state inventory Physiological responses
(ISIR) (CNRS-LIMSI) (NAIST) (Nara Med. Univ.)
(Image: psychology toolbox)
Public speaking training Technical approach
(Image: the verge)
WP1 WP2 WP3 Gathering therapy data WP4 (Osaka Univ.) WP5 Situation & immersion Virtual agent platform Dialog modeling Online measuring & feedback
Language processing
Facial expression Body movement
Evaluation Agents integration, translation
SST CBT
Speech recognition
New scenarios System improvement
Theories of social interaction
Data collection
quantitative qualitative
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
Social skills training (SST)
Medically established method for behavior training for autism, and schizophrenia. Flow: Situation setting -> Modeling -> Role play -> Feedback -> Homework 1 period: 30 minutes or more x 10 times [Nara Med. Univ. 2018]
13
Trainer Participant
Language: words, expression, consistency, coherence, causality, etc. Speech: pitch, speaking speed, filler, voice quality, amplitude etc. Image: facial expression, gesture, posture, eye gaze etc.
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
Cognitive behavioral therapy (CBT)
Psychotherapy-based instructional method to change the cognitive schema Cognitive Reconstruction, Behavioral Activation, Situation Analysis, Problem Solving, Progressive Relaxation, Assertion, Schema Modification This study uses cognitive reconstruction
– Situation setting -> Initial mood -> Automatic thinking
- > Disapproval -> Adaptive thinking -> Change in
mood -> Analysis (homework) – 1 period: 30 minutes or more (usually 4-50 minutes) x 16 times
14
e.g. thinking note (column method)
Cog
- gnitive Th
Therapy Reco cord rd ECA A Dialog
- gue
Situation
- n
Poi
- inted ou
- ut my
my sm small voi
- ice in
prese sent ntation
- n, and w
weakn kness ss log
- gicality by the a
audienc nce. e. Situation
- n breakd
kdow
- wn
In Initial sc scor
- re Disc
scou
- urage
ged(80)、Sad Sad(90) Desc scribe be emot
- tion
- n
st state Ne Nega gative thinki king I I am not
- t a
able to
- prese
sent nt in fron
- nt
- f
- f audience.
- nce. I
I am not
- t go
good
- d for
- r this
s job
- b.
Ext xtract tendency ency Evidence ce for
- r
this s thinki king My voi
- ice i
is s sm small in ge genera
- ral. I
I am not
- t able t
to
- exp
xplain l log
- gically.
Ext xtract evidenc nce Al Alternative thinki king Th There are m many peop
- ple who g
- give
a go good
- d prese
sent ntation
- n in sm
small voi
- ice.
It It is s pos
- ssi
sible to to prese sent nt lo logi gically if I I prepare e well in a advance. Prov
- vide infor
- rmation
- n
to
- gu
guide for
- r
ext xtracting g cou
- unt
nter er thinki king. g. Change ge thou
- ugh
ghts I I can gi give a go good
- d prese
sent ntation
- n
even voi
- ice i
is s sm
- small. I
I will prepare e mor
- re in a
advance Ext xtract alternative thinki king sc scor
- re
Disc scou
- urage
ged(20)、Mot
- tivated(8
0) Reco cogn gnize ze chang nge e of
- f
thou
- ugh
ghts
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
Summary
Social Skill Training – SST for Autism Detection of early dementia New project
– ANR-CREST: TAPAS project !
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
15
Continue to J-C Martin!
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
16
Children with ASD
Communicate to unfamiliar person Pre-post comparison(p = 0.002, Cohen's d = 1.17)
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
17
Cognitive reconstruction data
Collecting interaction data between doctors and people with depression e.g. Beck & Beck cognitive therapy live session
2019/12/02 ANR-CREST Symposium 2019 Dec. 2nd Satoshi Nakamura, AHC, NAIST
18