Available online at www.sciencedirect.com
ScienceDirect
IFAC PapersOnLine 52-13 (2019) 1145–1149
Exploring Opportunities for Artificial Emotional Intelligence in Service
Production Systems
Marlene Amorim*, Yuval
Cohen**, João Reis***,
Mário Rodrigues****
*DEGEIT & GOVCOPP, Universidade de Aveiro, Campus Universitário
de Santiago, 3810-193, Aveiro, Portugal (Tel: 351
234370200; e-mail: mamorim@ ua.pt).
** Department of Industrial Engineering, Afeka College of Engineering, Tel Aviv Israel (e-mail: yuvalc@afeka.ac.il)
*** Department of Military Science and CINAMIL/CISD, Military Academy, Lisbon, Portugal (email:joao.reis@academiamilitar.pt)
****IEETA & ESTGA, Universidade de Aveiro, Aveiro, Portugal (e-mail: mjfr@ua.pt)
Abstract: This paper offers an exploratory view about opportunities for the integration of Artificial
Emotional Intelligence (AEI) in service production systems. Service systems are conceptualized as
increasingly digitalized and networked co-production environments where employees, customers and
technology engage in rich interactions to deploy outputs and create value. The study builds on a
systematic literature review to identify three dimensions of application of AEI for improving service
efficiency and reliability, namely to augment, to assist and to mimic customer and/or providers
production capabilities. The paper offers a contribution to promoting the dialogue between the
knowledge domains of service systems and AEI.
© 2019, IFAC
: (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
:
Keywords:: artificial intelligence, service production systems, customer interaction, human machine
interaction,
: digital services.
:
1. INTRODUCTION
Services account for a major share of economic value in
modern societies (Vargo, Maglio and Akaka, 2008). The
importance of the service industries for value creation and for
employment in economies is estimated, and expanding,
beyond 70%, while the diversity of service companies, as
well as service functions within manufacturing grows
consistently (Gebauer, Fleisch and Friedli, 2005).
environments
wheresystems
customers’
andasknowledge
are
Service production
can bework
defined
co-production
environments
where customers’
work
and knowledge
are
’s
resources,
implying
that
environments
and
are
environments where
where customers’
customers’ work
work
and knowledge
knowledge
are
’s resources,
implying that
brought together
with the
company’s
resources,
implying that
that
environments
where
customers’
work
and knowledge
are
’s resources,
implying
environments
where
work
andproduction
knowledge
are
customers
always
playcustomers’
an active role
in the
of that
the
’s resources,
implying
resources,
implying2008,
that
service outputs (Zeithaml et al.,’s 2006,
Pinhanez,
Sampson, 2010, Spohrer and Maglio, 2010). Service
production systems operate via an orchestrated set of
interactions, and exchanges, between the company and the
and
even the
customer
(e.g. of
in materials,
the so called
“people
customer,
involving
the self
handling
information,
and
even the
customer
self
(e.g. sectors
in the such
so called
“people
processing
services”,
that
include
as
healthcare,
and even
even the
the
customer
self
(e.g. sectors
in the
the such
so called
called
“people
and
customer
(e.g.
in
so
“people
processing
services”,
thatself
include
as healthcare,
processing
services”,
that
include
sectors
such
as
healthcare,
and
even
the
customer
self
(e.g.
in
the
so
called
“people
processing services”, that include sectors such as healthcare,
and
even the
customer
(e.g.
in the
so coordination
called
“people
transportation,
education,
etc.),
as sectors
well
as such
the
of
processing
services”,
thatself
include
as healthcare,
processing
that include
as healthcare,
production services”,
tasks (Sampson
and sectors
Spring,such
2012).
This coparticipated nature of service production systems, in practice,
i.e.taking
with the
company’s
translates into a number of interactions
place
between
i.e. with the
company’s
customers and the production system, i.e.
i.e. with
with the
the company’s
company’s
employees or with its mediated or automated
interfaces
and
i.e. with the
company’s
i.e. with
the company’s
touch points (Giannakis,2011, Schumann,
Wünderlich
and
Wangenheim, 2012, Larivière, et al. 2017). In this vein,
service production systems have been pioneers in the domain
of experimentation of human computer interactions, and
systems
for in
effective
and efficient
of providers’
particularly
investigations
aimingintegration
for the development
of
systems
for effective
and efficient
integration
of providers’
systems
for
effective
and
efficient
integration
of providers’
providers’
systems for effective and efficient integration of
resources,fori.e.
employees,
technology
and the customers.
A
systems
effective
and efficient
integration
of providers’
systems forexample
effectivehas
andbeen
efficient
providers’
prominent
the integration
adoption ofofonline
and
mobile delivery channels, and their integration in cross
channel processes, enabling firms to achieve gains in the
design of service encounters, to meet customer requirements,
notably for access and
service convenience at an
unprecedented pace (Reis, Amorim and Melão, 2018).
Digital technologies find a fertile ground in services, because
they create opportunities for continuous innovation in the
ways that information - that is a core input for service
production systems - can be exchanged, transported and
transformed by different users and contexts (Lusch and
Vargo, 2014). The recent developments and the adoption of–
–
advanced information technologies (e.g. Internet-of-Things ––
IoT, Cyber-Physical Systems, etc.) are triggering promising–
new avenues for service innovation and the emergence of–
smart service systems. The expanding links between the
physical world and networked technologies, including
networked sensors creates a powerful and augmented space
for the interactions and collaboration between service
providers and customers for value creation. The
generalization of the connections between information to
people and people to people, is supporting the connection of
objects, spaces, things etc. As such the context for service
production and interaction between service actors is evolving
towards a reality that is intuitive, context-aware, and with
2405-8963 © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2019.11.350
1146
Marlene Amorim et al. / IFAC PapersOnLine 52-13 (2019) 1145–1149
intelligence capabilities to enable gains in efficiency and
value creation. Two additional concurrent facts, enabled by
the advanced ICT, are colouring this scenario: the amazing
growth in the volume and nature of data that is available,
captured by the smart networked systems, and the
sophistication of the computational capabilities, to deal with
them (Yang et al. 2017). In such contexts, Artificial
Intelligence (AI) allows for the delegation of the complex
tasks of pattern identification, learning, etc. to the capabilities
of computer based processes that can surpass humans in their
performance for capturing, structuring and understanding
large volumes of data (O’Leary, 2013).
One of the emerging areas of research in AI concerns the
development of capabilities to collect process and respond to
emotional states of individuals. The concept of Emotional
Intelligence (EI) refers to the ability of individuals to deal
with his and others peoples’ emotions (Saloyey and Mayer,
1990). The nature of EI involves the activation of several
concurrent skills concerning: i) the ability to identify
emotions; ii) the ability to use emotions; iii) the ability to
understand emotions and iv) the ability to manage them. Over
the years several research efforts have been devoted to
develop methods to support the assessment and measurement
of EI, such as the Mayer-Salovey-Caruso EI Test – MSCEIT
– that reflect the intrincate nature and complexity of EI
(Mayer, Salovey, Caruso, and Sitarenios, 2003).
In this paper we offer an exploratory study to gain
understanding about AEI applications in service production
systems. The study builds on the analysis from a structured
literature review, to identify examples of applications and
experimentation of AEI in service systems, therefore offering
a structured approach to understand the potential of AEI in
this domain, and suggesting paths for future research.
2. STUDY DESIGN AND METHODS
2.1 Process and criteria for selection of publications
In December 2018, a systematic literature review was carried
out resorting to SCOPUS database to search for publications
(i.e. journal articles, extended abstracts, and conference
proceedings) related with artificial emotional intelligence AIE. While no limitations were applied for date, or type of
publication, only articles written in English were taken into
account. The publications resulting from the data base query
went through a selection process involving 2 researchers and
that included the analysis of the title and the abstract, on a
first reduction round, followed by the reading of the complete
content of the manuscripts, to obtain a final set of
publications to be included in the review. The guiding
objective of the publications’ search was to identify research
work describing or discussing the application of technologies
for AIE in the specific context of service business.
Given the novelty of the topic, and in order to limit the
possible exclusion of relevant publications, in the database
search the researchers employed the generic string “artificial
emotional intelligence”, without restricting the query to the
specific field of service companies, given that service
production contexts can be embedded in manufacturing
contexts. Likewise in order to cover an ample spectre of
publications that could refer to the application of artificial
emotional intelligence to the broad field of production
systems, the search considered an ample scope for the
scientific fields, namely, “Engineering”, “Social Sciences”,
and “Business, Management and Accounting”.
Employing a stepwise approach the each researcher
independently assessed the relevance of the publications, by
reading the title and the abstract, using a rating scale, from 0 not relevant – to 2 – very relevant. Publications rated with a 0
were excluded from the sample, and those with at least 2
points were selected. In a following step the full text was
read, allowing, if necessary for the revision of the decision
about the inclusion of the publication.
2.2 Characterization of selected publications
In the first phase of the query, a total of 511 documents were
identified, by applying the aforementioned filtering
conditions. The researchers’ assessment about the relevance
of the publications led to a sample number of 67 publications,
that was further restricted to 61 after full text reading. Most
of the publications were excluded because they focused in the
advancement in the technical development of artificial
intelligence functions and capabilities, without analysing a
specific industry or service scenario for its application. As
expected, it was possible to observe an upward trend in the
publication volume (Fig. 1).
Regarding the domain and type of publication, a substantial
number of the selected articles were obtained from indexed
conference proceedings, something that can be explained by
the emergent nature of the topic. Predominant domains
included artificial intelligence scientific meetings as well as
computer science, robotics and automation. Nevertheless, the
selected sample suggested that the topic is gaining
momentum among industry and production systems
academics, as per some of the articles appeared in outlets
such as: International Journal of Modern Manufacturing
Technologies, International Journal of Production,
International Journal of Cognitive Research in Science,
Engineering and Education, Computers and Electrical
Engineering, Journal of Cognitive Engineering and Decision.
Fig. 1. Yearly evolution of relevant publications
3. AIE AND SERVICE PRODUCTION SYSTEMS
Marlene Amorim et al. / IFAC PapersOnLine 52-13 (2019) 1145–1149
Articles were read in full, and analysed in order to address
three questions: i) who is the subject of emotion
identification?; ii) what kind of inputs are being used as
sources data to understand the emotions; iii) what is the
[production system] purpose of understanding the emotion?
The researchers classified the manuscripts content into
categories, to address each of the proposed questions
independently, and this information was then compiled from
the individual reading notes of each researcher. In order to
provide a common framework for the classification of the
information extracted from the publications, the researchers
followed the service interaction classifications prevalent in
the literature and that distinguish; face-to-face service
interactions and face to screen service interactions. In this
service typology proposed by Froehle and Roth (2004), faceto-screen services can be (usefully) subdivided in to
technology mediated contact (e.g. phone service interactions,
online calls, etc.) and technology generated customer contact
(i.e. self-service).
In what concerns the subjects considered for emotion
identification, it was possible to observe articles focusing on
two distinct actors of service systems: service employees, and
service customers. The works addressing the emotions of
employees, understood as key resources for the performance
and quality of the service system outputs, were outnumbered
by research work focusing on the emotions of customers.
Examples of advanced work addressing the emotions of
employees include: the work of Ribak et al. (2017) who
propose an “intelligent control room” for monitoring and
assessing the condition of employees (e.g. stress) from the
collection and interpretation of voice and facial expressions;
the work of Eboli, Mazzulla and Pungillo (2017) that
explores the relationship between driving risk and the
drivers’ physical and emotional conditions; the work of
Thatcher and Kilingaru (2012) that build on iformation on
eye movement to understand emotions from airline
employees and identify abnormal flight conditons; the work
of Baowman and Rogers (2016) that explores the influence of
emotional issues for the decision making capabilities of direct
care workers in ambient assisted living; and, in the same line
of concern, the work of Coroiu (2015) that propose an
emotional intelligent agent to support, and improve decision
making in manufacturing contexts. Overall the extant work is
acknowledging that the emotional conditions of employees
are determinant for their effectiveness and for the quality of
their work, and that their emotions are subject to variations
(e.g. derived from personal contexts as well as from workload
conditions). The development of AEI models and
applications in this domain is advancing in ameliorating the
capabilities for emotion recognition and in the development
of tools that can assist and improve employee performance,
by creating awareness of emotional conditions and providing
tools and information to mitigate any undesirable
consequences.
In the domain of emotions understating for customers, there
is a remarkable concentration of efforts in the areas of health
and care services, education and online purchasing and
service delivery. Illustrative examples include: the work of
1147
Amifra et al. (2018) that integrates emotion recognition – for
home patients - into monitoring services for chronic disease;
and, in the same vein the work of Myakala et al. (2017) that
identify emotions from children cry-detection, in order to
assist and inform care personnel. The early work of
Marcelino et al. (2015) explored the same domain of service
opportunity, i.e. elderly care by incorporating emotion
understanding in an eService platform for pervasive case, to
support the daily monitoring and detection of danger
scenarios. In the domain of education, online learning
contexts are a rich field for experimenting with AEI
applications, given their ease in capturing and tracking the
user information. Subramainan et al. (2017), Xu et al. (2017),
among others, are including capabilities for students’ emotion
recognition (e.g. anger, anxiety, boredom) into the
monitoring capabilities for an online learning system in order
to emit alerts and allow educators to adjust their teaching
strategies. In the domain of online service, notably in retail
and customer assistance, the pioneer examples stand out,
such as the work of Povoda et al. (2015) that focused on
understanding customer emotions building on data from
helpdesk messages to devise priority rules for adequate
service responses.
The data being used to inform emotion-understanding
models, is primarily driven from two types of sources. One
domain concerns behavioural sources, such as facial
expressions, voice tones, speech characteristics, eye
movement, lip movement and so on. Collecting such data
requires the adoption of capturing and recording technologies
such as, eye tracking devices (Thatcher and Kilingaru, 2012);
biometric mouses (Kaklauskas et al., 2008); computer and
mobile cameras (e.g. Xu et al. 2017), among others. One of
the key advantages of capturing behavioural data that is
frequently pointed out is the fact that many of such devices
are generalized technology, hold by most of the population
and therefore being easily accessible and with a relatively
low cost. Other area of data that can allow for the
understanding of emotions is performance data, i.e. data that
offers evidence on the tasks conducted by employees and
customers, and whose characteristics (e.g. duration, delays,
interruptions, failures) can allow for inferring about the
emotional condition of individuals. This data is extracted
from records and outputs of production processes (e.g. text,
e-mail messages, log times, etc.). The digitalization of service
processes comes in favour for the abundance of such data.
With the proliferation of mediated services, self-service
systems, online services and alike the traceability of
customers and employees interactions and task results is
largely increased (as compared to traditional face.to-face
services where, most often, the service encounters were not
recorded or tracked).
The purposes of the reported adoption of AIE in service
systems, are related with the objectives of improving the
efficiency, the reliability and the overall performance of the
production system. This is set to be achieved by the
augmenting and/or complementary role that AEI has towards
the capabilities of employers, customers and technology.
Anchoring the analysis on the typologies for service
interaction, that involve the elements of the service
1148
Marlene Amorim et al. / IFAC PapersOnLine 52-13 (2019) 1145–1149
production system – employees, technology and customers –
it was possible to observe that the AIE experimentations are
contributing to enable service capabilities around three broad
forms, as illustrated in Fig.1.
infused with some emotion recognition and adequate
response capabilities that humans hold. Examples of this
stream of work include, Tuyen, Jeong and Chong (2017), that
have developed a robot for the delivery of ambient assisted
living services, or Loghmani, Rovetta and Venture (2017).
4. CONCLUSIONS
Fig. 1 Subjects, Data Sources and Capabilities associated
with AEI in service systems
The application of AIE described in service systems is
allowing for the improvement of service system capabilities
in the following manners:
i) “augment” the capabilities of employees and customers in
face to face service interactions. In such service settings AI
algorithms and models that are being advanced are, for
instance providing employees with information about their
own emotions, in order to allow them to adjust their
behaviour (e.g. the aforementioned work of Ribak et al.
(2017) that aims to gather knowledge about the emotional
conditions of employees in a intelligent control room). In
addition to providing such, own “awareness” information, we
could also observe examples of the possibilities of AI for
providing knowledge to employees about the emotional state
of the customers, i.e., providing insights about the emotional
sate of the “other” (e.g. the aforementioned work of Povoda
et al. (2015) that offered to the company knowledge about
customer emotions derived from the content of helpdesk
messages). This perspective allows for a tipification of the
functional purposes of AEI in face to face service settings as
follows: augmenting service capabilities by offering self
and/others’ emotion understanding;
ii) “assist” the capabilities of the service production system in
technology mediated interactions. In such settings some of
the richness of the face to face contact between employers
and customers is lost (e.g. call center services). As such some
of the AEI proposals that are being advanced are supporting
the work of the employees offering complementary
information that they cannot get by personal contact (e.g. see
for example the referenced works o AEI applications in
online learning settings, were the system is capturing data to
understand the users’ attention, motivation, and allow for
pedagogical adjustments - Subramainan et al. (2017), Xu et
al. (2017)).
iii) “mimic” the capabilities of employees in automated
contact, or self service systems. In such service settings the
state of the art of the existing AEI proposals concerns the
utilization of customers’ data to train and “qualify”
automated systems, including robots, so that they can be
The purpose of this study is to advance the understanding of
the opportunities for the application artificial emotional
intelligence in service production systems. This purpose is
achieved by analysing and synthesizing the most recent
research and applications of artificial emotional intelligence
in service contexts through a systematic review of 61 articles
on the topic. Given the emergent nature of the topic, to the
best of our knowledge no previous attempts have been made
to synthesize AEI on this domain. The main contribution of
this paper is to provide a structured characterization of
existing empirical and conceptual research on AEI in the
context of services, structuring the roles of AEI developments
for improving the efficiency and the reliability of service
systems, by augmenting, assisting and mimicking the roles of
employees and customers in production. In particular, this
paper brings forth evidence of the existence of an unbalance
in the number of examples and contribution in the domains of
intervention for AEI in services, therefore calling for future
research on and opens areas for the development of
applications. Overall the study contributes to enrich the
dialogue about opportunities for digital innovation in services
by approximating the bodies of knowledge of service systems
and AI.
REFERENCES
Amira T., Dan I., Az-Eddine B., Ngo H.H., Said G.,
Katarzyna W.-W.,"Monitoring chronic disease at home
using connected devices",2018,"2018 13th System of
Systems Engineering Conference, SoSE 2018"
Bowman, S. E., & Rogers, W. A. (2016). Understanding
decision making among direct care workers in assisted
living. Journal of Cognitive Engineering and Decision
Making, 10(4), 369-390.
Coroiu, A. M. (2015). Emotional Intelligent Agent in
Decision-Making Process with Implications in
Manufacturing. International
Journal
of
Modern
Manufacturing Technologies, 7(2), 44.
Eboli L., Mazzulla G., Pungillo G.,"The relationship between
driving risk levels and drivers’ personality traits,
physical and emotional conditions",2017,"Transport
Infrastructure and Systems - Proceedings of the AIIT
International Congress on Transport Infrastructure and
Systems, TIS 2017.
Froehle, C. M., and Roth, A. V. (2004). New measurement
scales for evaluating perceptions of the technologymediated customer service experience. Journal of
Operations Management, 22(1), 1-21.
Gebauer, H., Fleisch, E., and Friedli, T. (2005). Overcoming
the
service
paradox
in
manufacturing
companies. European management journal, 23(1), 14-26.
Marlene Amorim et al. / IFAC PapersOnLine 52-13 (2019) 1145–1149
Giannakis, M. (2011). Conceptualizing and managing service
supply chains. The Service Industries Journal, 31(11),
1809-1823.
Kaklauskas, A., Zavadskas, E. K., Seniut, M., Krutinis, M.,
Dzemyda, G., Ivanikovas, S., and Jaruševičius, A.
(2008). Web-based biometric mouse decision support
system for user’s emotional and labour productivity
analysis.
Larivière, B., Bowen, D., Andreassen, T. W., Kunz, W.,
Sirianni, N. J., Voss, C., & De Keyser, A. (2017).
“Service Encounter 2.0”: An investigation into the roles
of technology, employees and customers. Journal of
Business Research, 79, 238-246.
Loghmani, M. R., Rovetta, S., and Venture, G. (2017, May).
Emotional intelligence in robots: Recognizing human
emotions from daily-life gestures. In Robotics and
Automation (ICRA), 2017 IEEE International
Conference on (pp. 1677-1684). IEEE.
Lusch, R. F., and Vargo, S. L. 2014. Service-Dominant
Logic: Premises, Perspectives, Possibilities, Cambridge,
UK: Cambridge University Press.
Vargo, S. L., Maglio, P. P., Akaka, M. A. (2008). On value
and value co-creation: A service systems and service
logic perspective. European management journal, 26(3),
145-152.
Marcelino, I., Laza, R., Domingues, P., Gómez-Meire, S.,
and Pereira, A. (2015). eServices-Service Platform for
Pervasive Elderly Care. In Ambient IntelligenceSoftware and Applications(pp. 203-211). Springer,
Cham.
Mayer, J., Salovey, P., Caruso, D. and Sitarenios, G. (2003).
Measuring emotional intelligence with the MSCEIT V
2.0. Emotion, 3(1): 97–105.
Myakala, P. R., Nalumachu, R., Sharma, S., and Mittal, V. K.
(2017, November). A low cost intelligent smart system
for real time infant monitoring and cry detection.
In Region 10 Conference, TENCON 2017-2017
IEEE (pp. 2795-2800). IEEE.
O'Leary, D. E. (2013). Artificial intelligence and big
data. IEEE Intelligent Systems, 28(2), 96-99.
Pinhanez, C. (2008). Service systems as customer-intensive
systems and its implications for service science and
engineering. In Hawaii International Conference on
System Sciences, Proceedings of the 41st Annual (pp.
117-117). IEEE.
Povoda, L., Arora, A., Singh, S., Burget, R., and Dutta, M. K.
(2015, October). Emotion recognition from helpdesk
messages. In Ultra Modern Telecommunications and
Control Systems and Workshops (ICUMT), 2015 7th
International Congress on (pp. 310-313). IEEE.
Reis, J., Amorim, M., and Melão, N. (2018, September).
Omni-Channel Service Architectures in a TechnologyBased Business Network: An Empirical Insight.
In International Conference on Exploring Service
Science (pp. 31-44). Springer, Cham.
Rybak, N., Hassall, M., Parsa, K., & Angus, D. J. (2017,
October). New real-time methods for operator situational
awareness retrieval and higher process safety in the
control room. In Systems Engineering Symposium
(ISSE), 2017 IEEE International (pp. 1-7). IEEE.
1149
Salovey, P. and Mayer, D. 1990. Emotional intelligence.
Imagination, Cognition and Personality, 9(3): 185–211.
Sampson, S. E. (2010). The unified service theory.
In Handbook of service science (pp. 107-131). Springer,
Boston, MA.
Sampson, S. E., and Spring, M. (2012). Customer roles in
service supply chains and opportunities for
innovation. Journal of Supply Chain Management, 48(4),
30-50.
Schumann, J. H., Wünderlich, N. V., and Wangenheim, F.
(2012). Technology mediation in service delivery: A new
typology and an agenda for managers and
academics. Technovation, 32(2), 133-143.
Spohrer, J. C., & Maglio, P. P. (2010). Toward a science of
service systems. In Handbook of service science (pp.
157-194). Springer, Boston, MA.
Subramainan, L., Mahmoud, M. A., Ahmad, M. S., &
Yusoff, M. Z. M. (2017, June). A Simulator’s
Specifications for Studying Students’ Engagement in a
Classroom. In International Symposium on Distributed
Computing and Artificial Intelligence(pp. 206-214).
Springer, Cham.
Thatcher, S., and Kilingaru, K. (2012). Intelligent monitoring
of flight crew situation awareness. In Advanced
Materials Research (Vol. 433, pp. 6693-6701). Trans
Tech Publications.
Tuyen, N. T. V., Jeong, S., and Chong, N. Y. (2017, June).
Learning human behavior for emotional body expression
in socially assistive robotics. In Ubiquitous Robots and
Ambient Intelligence (URAI), 2017 14th International
Conference on (pp. 45-50). IEEE.
Wilson, A., Zeithaml, V. A., Bitner, M. J., & Gremler, D. D.
(2012). Services marketing: Integrating customer focus
across the firm(No. 2nd Eu). McGraw Hill.
Yang, C., Huang, Q., Li, Z., Liu, K., & Hu, F. (2017). Big
Data and cloud computing: innovation opportunities and
challenges. International Journal of Digital Earth, 10(1),
13-53.
Xu, J., Huang, Z., Shi, M., & Jiang, M. (2017, September).
Emotion Detection in E-learning Using ExpectationMaximization
Deep
Spatial-Temporal
Inference
Network. In UK Workshop on Computational
Intelligence (pp. 245-252). Springer, Cham.