EXPLORING THE DIRECTION OF THE
ENGLISH TRANSLATION OF
ENVIRONMENTAL PROTECTION ARTICLES
BASED ON THE ROBOT COGNITIVE-
EMOTIONAL INTERACTION MODEL
Shuai Song*
Shanghai University of Sport, Shanghai, 200000, China
renlilin666@163.com
Reception: 01/12/2022 Acceptance: 16/01/2023 Publication: 05/03/2023
Suggested citation:
S., Shuai. (2023). Exploring the direction of the English translation of
environmental protection articles based on the robot cognitive-
emotional interaction model. 3C TIC. Cuadernos de desarrollo aplicados a
las TIC, 12(1), 222-246. https://doi.org/10.17993/3ctic.2023.121.222-246
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
222
ABSTRACT
To broaden the application area of the cognitive-emotional interaction model for
robots. In this paper, an algorithmic model for the English translation of environmental
articles based on a cognitive-emotional interaction model for robots is used to model
the process of emotion generation using reinforcement learning. Similarly, positivity
and empathy are used to quantify the reward function for emotional state assessment,
and the optimal emotional strategy selection is derived based on the utility function. In
the process of article translation by the robot, Lagrangian factors are introduced to
make the translation probability maximum process transformed into the process of
obtaining the highest value of the auxiliary function at a random state. Finally, the
effectiveness of the robot's cognitive-emotional interaction model in the English
translation of environmental protection articles is verified by the Chinese-English
parallel question-and-answer dataset. The experimental results demonstrate that this
model can not only be used for the English translation of environmental protection
articles but also can give the corresponding English translation work similar to human
emotions, which can better help people understand the meaning of English. It also
provides a basis and direction for the subsequent in-depth application of the robot
cognitive-emotional interaction model in various fields.
KEYWORDS
Robot cognitive model; Emotional interaction model; Optimal emotional strategy;
Emotional state assessment reward function; Reinforcement learning model
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
223
PAPER INDEX
ABSTRACT
KEYWORDS
1. INTRODUCTION
2. A COGNITIVE MODEL OF THE ROBOT WITH EMOTION AND MEMORY
MECHANISM
2.1. Cognitive model structure
2.2. EMOTION GENERATION SYSTEM
2.3. Incentive mechanism
3. GAME MODEL BASED ON COGNITIVE-EMOTIONAL INTERACTION MODEL
OF ROBOT
3.1. Game Model
3.2. Definition of utility functions
3.3. Optimal emotional strategy selection
3.4. Construction of a cognitive-emotional interaction model for robots
4. ENGLISH TRANSLATION MODEL BASED ON ROBOT COGNITIVE-
EMOTIONAL INTERACTION
5. EXPERIMENTAL DESIGN AND ANALYSIS OF RESULTS
5.1. Experimental design
5.2. Emotional Accuracy Analysis
5.3. Retrieving translation validity
5.4. Validation of Interaction Translation
5.5. Model satisfaction assessment
6. CONCLUSION
DATA AVAILABILITY
CONFLICTS OF INTEREST
REFERENCES
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
224
1. INTRODUCTION
In recent years, with the introduction and implementation of the concepts of "smart
home", "smart community" and "smart city", human-computer interaction has become
an indispensable part of the public's daily life [1-4]. People expect robots to have the
cognitive-emotional computing ability to generate advanced anthropomorphic
emotions while satisfying daily interaction needs [5]. At the same time, as the
intersection of psychology, cognitive science, and artificial intelligence has intensified,
researchers have found that robot cognition should be reflected in both "intelligence"
and "emotional intelligence" [6]. Therefore, robot cognition and computation of
emotional interaction models have become a hot topic in the field of intelligent robot
research [7].
Emotional interaction cannot be achieved without the technical means of artificial
intelligence [8]. For nearly two decades, AI researchers have been trying to empower
machines with cognitive abilities to recognize, interpret, and express emotions [9].
Artificial intelligence techniques simulate human emotional cognition and decision-
making processes by correlating, analyzing and reconstructing data containing
emotional information in different scenarios with each other, and eventually
transforming the data into abstracted thoughts that computers can understand [10-12].
The affective interaction process uses the user's modal data to achieve recognition of
the user's affective state and uses the feedback information from the affect recognition
to perform affective modeling based on cognitive analysis and to guide the interaction
behavior [13-14]. Thus, the sentiment recognition process and the sentiment modeling
process are the two most important steps of sentiment interaction [15].
In recent years, numerous valuable research approaches have emerged in the field
of cognitive and affective interaction modeling for robots. The literature [16] argues
that cognitive-emotional computing is about giving computers the human-like ability to
observe, understand, and generate various emotional states so that they can interact
in a naturally intimate, lively, and interesting way like humans. The literature [17]
proposed an emotional interaction model based on guided cognitive reassessment
strategies GCRs, which can reduce the robot's dependence on external emotional
stimuli and promote positive emotional expression of the robot to some extent. The
literature [18] proposed a personalized emotion model based on PAD and established
a three-level mapping relationship between personality space, mood space and
emotion space to describe the human emotion change pattern. The literature [19]
used electrophysiological techniques and MRI to study the expression of cognition
and emotion in the brain during behavior. In the literature [20], cognitive feelings were
identified as an emotional experience, and this was confirmed by observing changes
in the physiological and behavioral representations of validity and arousal in a
cognitive task. In [21], a willingness-based interpretable and computational emotion
model and CASE, a personality model to measure robot differences, were proposed
to improve the performance of multiple robots in pursuit of tasks by using a willingness
to quantify the effect of emotional factors on task assignment through an emotional
contagion model to compute inter-robot emotional interactions. In the literature [22],
by conducting emotion-cognition-related experiments, it was found that emotional
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
225
states affect the input of cognitive control in the brain and the associated
metacognitive experience. The literature [23] established a multi-emotion dialogue
system MECs by multi-task Seq2Seq learning, and after multi-task learning based on
question-answer datasets, the candidate answer with the most similar emotion to the
input interrogative was selected as the output of the robot, which achieved better
results in a single round of dialogue. The literature [24] proposed an integrated
framework for emotion computation, which firstly considers personality traits, social
content and other factors for the evaluation of external emotional stimuli, secondly
considers mood states, internal memories and other factors for the generation of
emotions, and finally performs the expression of intelligent behaviors based on the
generated emotions. The literature [25] assesses students' interest in learning events
based on a combination of OCC affective modeling and fuzzy reasoning, which is a
means of using affective modeling as an aid to affect recognition prediction. The
literature [26] implements the process of natural change in robot emotions, which for
interactive behavior is still based on discrete rules for emotion mapping. The literature
[27] uses statistical features of skin electrical signals, ECG signals with body
temperature and time-frequency features for global generalized emotion recognition.
The literature [28] uses hierarchical support vector machines for reducing the bias of
robot cognitive training binary classifiers. The literature [29] proposes an emotion
model based on the Pruschik emotional color wheel to enable social robots to mimic
human emotional changes and personality traits in entertainment and education to
talk naturally with people. The literature [30] builds an affective cognitive model that
implements emotion generation based on external event motivation and regulates the
flow of information generated by the generated emotions through the competition
under different drivers to determine the behavioral output.
In this paper, a reinforcement learning-based cognitive emotional interaction model
for robots is proposed [31-32]. First, reinforcement learning is used to model the
emotion generation process, and the one-dimensional emotion model theory is used
as the emotion state space of the robot, which motivates the robot to improve
efficiency in the process of emotional interaction; second, three emotional influencing
factors of similarity, positivity and empathy are considered to quantify as the reward
function for conducting emotional state assessment, and the optimal emotional
strategy selection is derived based on the The optimal emotional strategy selection is
derived based on the effectiveness function to realize the interaction motive of
emotional support, emotional guidance and emotional empathy for the participants;
thirdly, Lagrange factor is introduced in the process of environmental protection
English articles translation by the robot, which makes the process of the highest value
of machine translation probability transform into the process of obtaining the highest
value of the auxiliary function at the random state. The retrieval speed of machine
translation is improved, the efficiency of machine translation is enhanced, and high-
precision translation results can be obtained more effectively. Finally, the Chinese-
English parallel question-and-answer corpus commonly used in environmental
protection articles is used as the experimental data set, and the optimal emotional
state is combined with the optimal emotional state to update the robot's emotional
state transfer probability, to realize the robot's state transfer in the translation process
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
226
and ensure the continuity of the translation process. The experiments validate the
model's effectiveness in terms of accuracy, MAP and MRR.
2. A COGNITIVE MODEL OF THE ROBOT WITH
EMOTION AND MEMORY MECHANISM
2.1. COGNITIVE MODEL STRUCTURE
Different disciplines, such as brain science, cognitive neuroscience, and cognitive
psychology, have researched brain structure and its emotional and cognitive
principles. The results show that emotional and cognitive functional areas of the brain
are mainly concentrated in the thalamus, limbic system, and cerebral cortex. The
thalamus, as the sensory transmission center, is responsible for the transmission of
external sensory information such as visual, auditory and olfactory information, as well
as internal sensory information. The limbic system, as the emotional center, mainly
includes the hippocampus, amygdala, and cingulate gyrus. The hippocampus is
responsible for emotional memory and learning, the amygdala is responsible for
emotion generation, regulation, and recognition, while the anterior lower part of the
cingulate gyrus is related to emotional processing and the posterior upper part is
related to cognitive functions. These structures are linked to the hypothalamus and
the vegetative nervous system and are involved in regulating instinctive responses
and emotional behavior; the cerebral cortex is mainly involved in human brain
activities such as understanding events, making decisions about goals, and managing
the timing of behavior.
Figure 1. Cognitive model structure
Based on the above cross-disciplinary research foundation, this paper proposes a
robot cognitive model, which contains seven parts: receptor, internal state,
environmental state system, emotional system, behavioral decision system, dynamic
Cognitive Systems
Environmental
Status System
Internal
Status Receptors
Dynamic
Knowledge Base
Behavioral
Decision System
Z
-1
Emotional System
Emotional
Memories
Emotion
generation system
Emotional state
Voice Output
External
Environment
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
227
knowledge base, and execution output, and the model structure is shown in Figure 1.
The meanings of each part are as follows:
(1) Receptors: feel all kinds of information from the external environment, and
represent the felt information as a triad:
(1)
Where
is the set of perceptible discrete states,
is the set of optional action subsets corresponding to
discrete states, is the set of maximum internal energy
state replenishment corresponding to discrete states, and 4 is the number of
perceptible discrete states.
(2) Internal state: the internal state information of the robot body, such as the
internal energy robot and the durability of the robot. The internal state of the robot is
the internal energy state , that is:
(2)
Where P is the set of energy states inside the robot body, is the robot task
survival time, t=0 means the robot is at the task start moment; represents the
robot body internal energy state is zero or the task completion moment.
(3) Environmental state system: the robot's external environmental information and
the body's internal state hub station, denoted as:
. Where
is the set of internal energy state gains obtained by the
robot from discrete states, and the internal energy state gains are defined as follows:
(3)
Where is the maximum internal state replenishment corresponding to
the discrete state at the time .
(4) Emotional system: robot emotional state and emotional memory generation
center, expressed as a triad:
(4)
Where is the set of emotional states generated by the
emotion generation system; is the set of inverse
emotional rewards generated by the emotional memory;
is the set of periodic emotional rewards
generated by the emotional memory, is the robot task survival time internal energy
state recharge cycle,
represents the first internal energy state recharge;
represents the robot internal energy state before zero or the completion of the
task maximum energy recharge cycle.
(5) Behavior decision system: Based on the output of the environment state system
and the emotion system, we combine the dynamic knowledge base to realize the
PER _ORG =⟨S,A,Ga⟩
S={Si|i= 1,2,⋯,ns}
A={Ai|i= 1,2,⋯,ns}
Ga ={Gai|i= 1,2,⋯,ns}
P(t)
P={P(t)|t= 1,2,⋯,nt}
nt
t=nt
⟨PER _ORG,P,G⟩
G={G(t)|t= 1,2,⋯,nt}
G
(t) =
{P(t)−P(t−1) Ga(t)≠0
G(t−1) Ga(t)=0
Ga(t)
Gai
t
EMO_SYS =⟨E,Remo,Rmen⟩
E={E(t)|t= 1,2,⋯,nt}
Remo ={Remo(t)|t= 1,2,⋯,nt}
Rmem ={Rmem(T)|T= 1,2,⋯,nT}
nT
T= 1
T=nT
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
228
robot's behavior decision. It is expressed as a binary group:
, where
is the set of behavioral decisions, is the number of
behaviors of the translation robot, is the set of translation
word selection, and
is the number of actions of the translation robot. For the
English translation task of environmental protection articles requiring "energy
replenishment", the robot's behaviors are divided into the search, energy
replenishment, and search actions for the selection of
directions at each node of the article.
(6) Dynamic knowledge base: Knowledge base of English words for robotics and
environment, with knowledge elements represented as five-tuples:
(5)
is the set of discrete states corresponding to the best
action of energy replenishment; is the set of word
search states; is the set of environment search states;
is the set of state-English word memory,
records the sequence of states
encountered and English word selection during the cycle,
represents the total
number of discrete states encountered during the cycle, records the last state
encountered during the cycle and the sequence of English word selection for that
state, , is the state-energy memory set,
which records the discrete states encountered during the cycle and the internal
energy states required to return to the energy recharge point , is the number of
discrete states encountered during the search cycle, and
reflect the memory function of the robot cognitive model.
(7) Execution output: the robot article translation output actuator and the action
actuator are represented as a binary group:
. Where
is the first article translation output set;
is the correction article translation output set.
2.2. EMOTION GENERATION SYSTEM
To study robots with emotional mechanisms, first of all, artificial emotions have to
be generated, which requires modeling emotions. In this paper, based on the theory of
the one-dimensional emotion model, an emotion interaction model is designed for the
emotion generation system, which can generate six emotions: happy, surprised,
disgusted, angry, fearful and sad, in the following form:
(6)
(7)
⟨π,a⟩
π
=
{
πj|j= 1,2,⋯,nj
}
nj
am={m= 1,2,⋯,nm}
nm
{nor th,south,east,west}
DYN_K NO =⟨A′ ,EL,D,STA _EGW,STA _PWO⟩
A′ = {A′
i|i= 1,2,⋯,ns}
EL = {EL(T)|T= 1,2,⋯,nT}
D= {D(T)|t= 1,2,⋯,nt}
STA _EGW =⟨(Y,U), (Y′ ,U′ )⟩= {(Yk,Uk), (Y′
c,U′
c)|k= 1,2,⋯,nk,c= 1,2,⋯,nc}
(Y,U)
nk
(Y′ ,U′ )
nc
STA _PWO = {(Y′ ′
z,Bz)|z= 1,2,⋯,nz}
Bz
nz
STA _EGW
STA _PWO
⟨V1,V2⟩
V1={V1m|m= 1,2,⋯,nm}
V2={V2m|m= 1,2,⋯,nm}
( )
2
1 3
( ) ( ) ( )
( ) [arctan( ( ) ) ]
D t
G t P t B t
E t P t k e
k k
−
−
=−+⋅
4
5
2 ( ) ( ) ( )
6
( ) ( )
P t B t k D t
k
E t G t e k
− − −
=⋅ −
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
229
Where is the internal energy state value
required to return to the energy
recharge point corresponding to the discrete state at the time
. The emotional
intensity is positively related to the internal state gain
obtained by the
robot. When
, (6) produces four emotions: happy, surprised, sad, and
fearful, with positive happy emotion at
and fearful emotion at
. When , (7) produces angry emotion at and
disgustful emotion at . is the emotion model parameters.
2.3. INCENTIVE MECHANISM
In everyday life, rewards are usually sparse. The performer often needs to go
through a series of attempts during a task until the task is completed to obtain a
reward, and no reward is manifested during the process. However, the human brain
possesses a reflective mechanism that can establish a relevant connection between
the temporal situation and the target thing and obtain a reward based on the memory
mechanism. Therefore, in this paper, while considering affective cognition, we
combine memory cognition, based on the framework of reinforcement learning theory,
to integrate intra-emotional reward with memory. A reward mechanism is proposed,
which consists of environmental reward, reversed affective reward and periodic
affective reward. Among them, the reversed affective reward and the periodic affective
reward are internal rewards.
Figure 2. Schematic diagram of the robot cognitive reward mechanism
Figure 2 shows the process of obtaining a search reward for an article to be
translated by the article English translation robot (schematic diagram). The robot
consumes internal energy during the search process, and translates the word
selection according to the optional action subset
of nodes, assuming that node 1 is the energy replenishment point and node X
indicates the unsearched point. As shown in Figure 2, the robot search trajectory is:
B(t)
Bz
t
|E(t)|
G(t)
G(t)>0
E(t)>0
k7<E(t)<0
G(t)<0
k6<E(t)<k5
k7<E(t)<k6
k1∼k7
X
Paths
54 6
X32X
1
Environmental Rewards
Reverse Emotional Reward
Periodic Emotional Rewards
am
Ai∈{nor th,south,east,west}
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
230
, if it reaches node 6, the robot must go back to node 1 to
replenish the internal energy state to maintain the next search, the return trajectory is:
, the robot from the energy replenishment point to search
and then back to the energy replenishment point. At this point, the article English
translation robot completes a cycle of the search for the article to be translated. The
reward mechanism in this process generates rewards
, and
each reward is shown in Table 1:
Table 1. Schematic diagram of the robot cognitive reward mechanism
Where, process to node
, 2 is the external environment reward for
node 1 to obtain action toward node 2;
is the reward for node 2 to obtain
action toward node 1, i.e., the reverse emotional reward.
the process to node 1, is the reward for node 6, node 5, node 4,
node 3, and node 2 to obtain the tendency to node 1 action step by step, i.e., the
periodic affective reward.
The environment bonus is set according to (8). The energy replenishment point is
the node where the robot replenishes internal energy during the article translation
process; the dead-end node is the node where only the "return" action can be
selected when there is a single word that cannot be searched; the trap point is the
node where the robot loses additional internal energy at this node; and the normal
node is the node with internal state gain and is not a dead-end node. The
Q value is updated by (9), s is the current state, a is the current action state of the
selected English word, is the learning rate, and
is the maximum
gain of the next state after the current state selected action.
(8)
(9)
The reverse sentiment reward is set as in equation (10), and the Q value is
updated as in equation (11), with
the reverse direction at the time of entering this
node state.
1→2→3→4→5→6
6→5→4→3→2→1
R=⟨Renv,Remo,Rmem⟩
Reward type Reward collection
Renv1→2,Renv2→3,Renv3→4,Renv4→5,Renv5→6
Remo
Renv
Remo1←2,Remo2←3,Remo3←4,Remo4←5,Remo5←6
Rmem6→5→4→3→2→1
Renv6→5,Renv5→4,Renv4→3,Renv3→2,Renv2→1
Rmem
1→2
Renv1→2
Remo1←2
6→5→4→3→2→1
Rmem6→5→4→3→2→1
Ga(t)=0
α
ma x Q(s′ ,a′ )
e
100 Energy recharge point
0 General Node
R ( ) Dead end nodes
Trap point
nv t
⎧
⎪
⎪
=⎨
⎪
⎪
⎩
- 5
-100
e
( , ) (1 ) ( , ) [R max ( , )]
nv
Q s a Q s a Q s aα α ʹ ʹ
=−+ +
a′ ′
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
231
(10)
(11)
The cycle sentiment reward is set according to equation (12), where
is the
sentiment state generated at the moment of completion of the T
th cycle, and the Q
value is updated according to equation (13); this reward is obtained only when the
cycle is completed.
(12)
(13)
3. GAME MODEL BASED ON COGNITIVE-EMOTIONAL
INTERACTION MODEL OF ROBOT
3.1. GAME MODEL
Modeling the emotion generation process of the participant and the robot during
human-robot interaction, the emotional cognitive interaction model tries to get the
optimal emotional response of the robot based on the previous historical emotion and
the current interaction input emotion of the participant, which leads to a more natural
and harmonious human-robot interaction, i.e., the optimal
is obtained by
knowing
, as shown in Figure 3 (R denotes the robot and H denotes
the participant object).
Figure 3. Human-machine interaction process
To facilitate theoretical analysis, as mentioned above, this paper unifies and
normalizes the emotional strategies of the participating object and the robot in the
human-robot interaction process into six basic emotions. After the robot is
stimulated by the external emotion of the participant
, it then selects the
optimal emotion strategy from the six basic emotions. In the process of making
the optimal emotion selection, the robot needs to perform the emotion trend
prediction for each emotion selection. The prediction of the emotion
that
will be generated by the participant in
interactions and the emotion that
e
( ) E(t)>0
R ( ) 1 E(t)>0
| ( ) |
mo
E t
t
E t
⎧
⎪
=⎨
⎪
⎩
e
( , ) R
mo
Q s aʹʹ =
E(T)
( ) T>0
( ) 0 T=0
mem
E T
R T ⎧
=⎨
⎩
( , ) (1 ) ( , ) [ max ( , )]
mem
Q s a Q s a R Q s aα α ʹ ʹ
=−+ +
Ek+1
RH
El
HR(1 ≤l≤k)
H H H
RRR
k
HR
E
1k
HR
E
+
2k
HR
E
+
3k
HR
E
+
4k
HR
E
+
Ek
HR
Ek+2
HR
k+ 2
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
232
may be replied to by the participant in sessions . The emotion strategy
selection process of the robot is shown in Figure 4.
Figure 4. The emotional strategy selection process of the robot
The game model should be judged by 3 elements: the participant, the strategy
combination, and the game gain. The participant and the robot constitute the two
objects of the game model, and both parties make different strategy choices around
subjective satisfaction, and different combinations of these strategies will produce
different game outcomes. Considering the human-robot interaction, both the
participant and the robot start from their subjective satisfaction, which is a non-
cooperative game. In human-robot interaction, emotions are bidirectional, i.e., the
subjective satisfaction of the participant is influenced by the robot's reply emotions,
and the subjective satisfaction of the robot is also influenced by the participant's reply
emotions, and the robot does not know what kind of emotional response it will get
from the participant for
sessions, which is an incomplete information game.
Therefore, this paper uses an embedding game based on the robot cognitive-
emotional interaction model to model the emotion generation process of the
participant and the robot.
3.2. DEFINITION OF UTILITY FUNCTIONS
The utility is the subjective satisfaction obtained by the interacting parties during
human-computer interaction without loss of generality. Consider the definition of the
participant's utility function: denotes
session
participant interactions with an input sentiment of
. Assuming that the sentiment
value of the session robot response is
, predict the value of subjective
satisfaction obtained by the participant if the sentiment value of the
session
participant is and the sentiment value of the session robot is .
In this paper, we define the utility function of a participant based on whether the
robot can adjust its self-friendliness according to the change in participant friendliness
k+ 3
Ek+3
HR
H
R
k
HR
E
1k
HR
E
+
2k
HR
E
+
3k
HR
E
+
HH H
Strategy 1 Strategy 2 Strategy 6
Ā
Ā
Incomplete information game
R
Strategy 2
Strategy 1 Strategy 6
Ā
HH H
Strategy 6Strategy 1 Strategy 2 Ā
Ā
Ā
k+ 1
UH(Ek
RH,Ek+1
RH ,Ek+2
HR ,Ek+3
HR )
k
Ek
RH
k+ 1
Ek+1
RH
k+ 2
Ek+2
HR
k+ 3
Ek+3
HR
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
233
and define the utility function of a participant based on whether the value of emotional
empathy between the participant and the robot keeps increasing.
is defined as:
(14)
where
denotes the ratio of the amplitude of change in participant and
robot friendliness, and tends to 1 when the amplitude of change between
participants and robots is essentially the same, and vice versa, tends to 0; for
and are defined as:
(15)
denote the empathy values between session participant emotions and
session bot emotions, and session participant emotions and
session bot emotions, respectively. The multiplication term 10 is to ensure that the
values are taken in the range [0,10].
(16)
The definition of the robot utility function
takes into
account, on the one hand, the change in the participant's friendliness, and if the
participant's friendliness increases, then the robot's utility value increases; on the
other hand, the robot's utility value decreases. On the other hand, the robot's
emotional resonance with the participant is considered based on the "principle of
increase and decrease in interpersonal attraction", which is similar to the definition of
the participant's utility function. Thus, the utility function of the robot is defined as:
(17)
3.3. OPTIMAL EMOTIONAL STRATEGY SELECTION
Based on the definition of the utility function, the optimal emotional choice strategy
of the robot is obtained for the emotional interaction input of the participant with the
help of the game model:
(1) The emotional stimuli of
session participants, each emotional strategy of
session robots, 6 emotional strategies of session participants, and 6
emotional strategies of session robots form a game matrix, and since
session robots share 6 emotional strategies, there are 6 game matrices in total;
(2) Assume that the emotional choice strategy of the robot for sessions is s.
Predict the emotional choice strategies of the participants and the robot by finding a
UH(Ek
RH,Ek+1
RH ,Ek+2
HR ,Ek+3
HR )
1 2 3 min
2 1
max
( , , , ) 10 0.5 0.5( )
k k k k
RH RH HR HR
F
UH E E E E R R
F
+ + + ⎧ ⎫
=⋅+−
⎨ ⎬
⎩ ⎭
Fmin /Fmax
Fmin /Fmax
Fmin
Fmax
min
max
min( ( 2) ( ), ( 3) ( 1)
max( ( 2) ( ), ( 3) ( 1)
F F k F k F k F k
F F k F k F k F k
= + −+−+⎫
⎬
= + −+−+⎭
R1,R2
k
k+ 1
k+ 2
k+ 3
1
1
2 3
2
( , )
( , )
k k
RH RH
k k
HR HR
R R E E
R R E E
+
+ +
⎫
=⎪
⎬
=⎪
⎭
UH(Ek
RH,Ek+1
RH ,Ek+2
HR ,Ek+3
HR )
{ }
1 2 3
2 1
( , , , ) 10 0.5[ ( 3) ( 1)] 0.5( )
k k k k
RH RH HR HR
UH E E E E F k F k R R
+ + +
=⋅+−+ + −
k
k+ 1
k+ 2
k+ 3
k+ 1
k+ 1
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
234
pure strategy Nash equilibrium for the game matrix formed by sessions,
i.e.:
(18)
(3) Solving for the optimal affective choice strategy s using the cis-induction
method.
(19)
The solution of the static embedded game Nash equilibrium is mainly for the
prediction of the next session sentiment trend from the subjective satisfaction of
participants and robot self, and the sub-game perfect equilibrium of the embedded
game is mainly for obtaining the optimal sentiment selection strategy of the sub-
session robot from maximizing the subjective satisfaction of the robot by using the
parsimonious induction method to simplify the Nash equilibrium.
3.4. CONSTRUCTION OF A COGNITIVE-EMOTIONAL
INTERACTION MODEL FOR ROBOTS
The game model is used to model the emotion generation process of the
participant and the robot during human-robot interaction, and the optimal emotional
response of the robot is obtained based on the previous historical emotions and the
interaction input emotions of the participant. The model construction process is as
follows:
Step 1 Input: post-session friendliness update values
and the
probability of transferring the sentiment state of the robot ,
sessions of
participant interaction input sentiment ;
Step 2 Output: the sentiment value of the robot at sessions ;
Repeat:
Step 3 Participant input interaction emotion ;
Step 4 Calculate the utility values of the participant and robot under each sentiment
strategy choice for the sessions robot, predicted sessions participant for
each sentiment strategy choice, and sessions robot for each sentiment strategy
according to Eqs. (14)-(17);
k+ 2, k+ 3
1 2 3
1 2 3
2 3 2
1 2 3
1 2 3
2
( , ( ), ( ), ( ))
( , ( ), ( ), ( )),
( ), ( ) , ( )
( , ( ), ( ), ( ))
( , ( ), ( ), ( )),
k k k k
RH RH HR HR
k k k k
RH RH HR HR
k k k
HR HR l HR l
k k k k
RH RH HR HR
k k k k
RH RH HR HR
k
HR
UH E E s E E
UH E E s E i E
E E E E i E
UH E E s E E
UH E E s E E j
E
+ + +
+ + +
+ + +
+ + +
+ + +
+
⋅ ⋅
≥ ⋅
∃ ⋅ ⋅ ∈ ∀ ∈
⋅ ⋅
≥ ⋅
∃3 2
( ), ( ) , ( )
k k
HR l HR l
E E E j E
+ +
⎫
⎪
⎪
⎪
⎪
⎬
⎪
⎪
⎪
⎪
⋅ ⋅ ∈ ∀ ∈ ⎭
1 2 3 1 2 3
( , ( ), ( ), ( )) ( , ( ), ( ), ( )), ( )
k k k k k k k k
RH RH HR HR RH RH HR HR l
UH E E s E E UH E E s E i E s E
+ + + + + +
⋅ ⋅ ≥ ⋅ ∃ ⋅ ∈
k−1
F(k−1)
Pr(k−1)
k
Ek
RH
k+ 1
Ek+1
RH
Ek
RH
k+ 1
k+ 2
k+ 3
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
235
Step 5
Solve the emotional choice strategy s of the cognitive model according to
Eqs. (18)-(19);
Step 6 The probability of transferring the emotional state of the robot is updated by
the optimal emotional strategy s; the human-robot interaction friendliness is updated
such that ;
Step 7 Until the participant stops entering interactive emotions;
Step 8 End of the HCI session.
During each round of human-computer interaction, the cognitive interaction model
is mainly a matrix operation with a time complexity of constant order O(1) in the
process of participant interaction input emotion evaluation and robot optimal emotion
strategy selection. Assuming that the number of human-computer interaction rounds is
n, the time complexity of the model is O(n), which ensures that the response time of
the model during human-computer interaction is acceptable.
4. ENGLISH TRANSLATION MODEL BASED ON
ROBOT COGNITIVE-EMOTIONAL INTERACTION
Setting any Chinese matrix f and an English sentence e, the probability of e being
machine translated into f is
. The problem of machine translation of f into e
can be viewed as the process of solving equation (20):
(20)
If the lengths of the English string e, as well as the Chinese string, are
and m, respectively, then we have 1. The alignment can describe
the positions of the words within the Chinese sentence corresponding to the words in
the English sentence by the presence of the position information of each value, then
we have
. Where the value interval of each value is [0,1], then
we have :
(21)
In this process, we generate Chinese sentences and alignment process based on
English sentences, obtain Chinese sentence length based on English sentences,
obtain the link position of the first Chinese word string, and then obtain the first word
of Chinese sentences based on English sentences, Chinese sentence length and the
position of the English sentence related to the first Chinese word, and loop the
process to obtain the overall Chinese sentences.
The English translation model based on robot cognitive-emotional interaction can
be implemented to simplify equation (21) and then give the model the ability of
emotional interaction, which makes the text more colorful in translation and can keep
the semantic and emotional color of the original text. The prerequisites set at the
same time are:
k=k+ 2
P(e|f)
e=arg ma x P(e|f)
f=fm
1=f1f2⋯fm
a=am
1=a1a2⋯am
1 1 1
1 1 1 1
1
( , | ) ( | ) ( | , , , ) ( | , , , )
m
j j j j
j
j
P f A e p m e p a f m e p fj a f m eα− − −
=
=⋅
∏
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
236
(1) If does not correlate with the target language e and the source
language length m.
(2) If is related to the target language e length l, then we
have:
(22)
(3) If is related to and , then there exists ,
. is the probability given and .
After incorporating the Lagrangian factors , the process of obtaining the highest
value of machine translation probability is transformed into the process of obtaining
the highest value of the auxiliary function at the random state, then the English
machine translation model based on the robot cognitive-emotional interaction model is
as follows:
(23)
The above English machine translation model is transformed into a reverse
machine translation model to accomplish accurate machine translation of the English
language using the statistical machine method of maximum entropy. The
transformation within the model ensures that the machine translation efficiency of the
machine translation model is improved by obtaining improved parameter values
through the great likelihood prediction method as follows:
(24)
(25)
and then obtain the formula:
(26)
After incorporating the new property, replaces and the
framework of the obtained extended statistical machine translation is:
(27)
Equation (27) enables the implementation of more efficient retrieval and the
acquisition of high-quality English machine translation results.
p(m|e)
p(αj|αj−1
1,fj−1
1,m,e)
1 1
1 1
1
( | , , , ) 1
j j
j
p a f m e l
α− − =+
p(αj|αj−1
1,fj−1
1,m,e)
fj
fal
ε=P(m|e)
t(fj|eal)=p(fj|aj
1,fj−1
1,m,e)
t(fj|eal)
eal
fj
λ1
0 0 1
( , ) ( | ) ( ( | ) 1)
( 1)
m
l l
m
al al l
j
s
h p L t f t f e
lθ θ θ
θ
λ α λ
= = =
=− −
+∑ ∑ ∑ ∑
∏
θ
=arg ma x
θ
s
∏
s=1
pθ(fs|es
)
γ
=arg ma x
γ
s
∏
s=1
pγ(es
)
el
1
=arg ma x
e
l
1
{p
γ
(el
1
)⋅P
θ
(fj
1
|el
1
)}
P
θ(ej
1|fl
1)
P
θ(fj
1|el
1)
el
1
=arg ma x
e
l
1
{p
γ
(el
1
)⋅P
θ
(ej
1
|fl
1
)}
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
237
5. EXPERIMENTAL DESIGN AND ANALYSIS OF
RESULTS
5.1. EXPERIMENTAL DESIGN
To facilitate the performance analysis and comparison experiments of the robot-
based cognitive-emotional interaction model proposed for the text, an English text
translation robot based on the robot-based cognitive-emotional interaction model of
this paper is built using the open-source chatbot ChatterBot. First, the English
translation model is used to match English translation answers of environmental
protection articles with the translation robot logic adapter, and the top answers with
higher confidence are returned as the candidate answer set; then, the sentiment
strategy is evaluated using the model of this paper, and the optimal sentiment strategy
is selected. Finally, the candidate answers are optimally ranked based on the
response sentiment of this paper's model, and the answer with the highest ranking
level is selected as the robot response output. In addition, since the number of
emotional states to be explored increases exponentially with the number of interaction
rounds, the maximum number of interaction rounds T=8 (rounds) for two bits of
intelligence and the number of candidate emotional states selected in each round n=8
(kinds) are set in this model for emotional state evaluation.
The experimental data uses the sample dataset from the NLPCC2017 shared task
Emotional Conversation Generation, which contains a total of 11207 Chinese-English
parallel question-and-answer corpus of articles about environmental protection. 6000
question and answer pairs are randomly divided as the validation set, 5000 question
and answer pairs as the test set, and the remaining question and answer pairs are
used as the training corpus for the chatbot to translate from English to Chinese.
The experiment focuses on the translation and affective accuracy of the English
translation of environmental protection articles, as well as the actual effect of human-
computer interaction sessions, so the following cognitive models are selected for
comparison experiments:
(1) A single robot cognitive model, Chatterbot, outputs responses based on the high
confidence level of each answer in the candidate answer set. Since it does not have
cognitive-emotional computing capability, it is only used for model validation
comparison experiments;
(2) Emotional Chat Machine ECM, which can produce appropriate responses in
terms of content-related grammar and emotional coherence;
(3) Adversarial network SentiGAN model, capable of generating generic, diverse
and high-quality sentiment texts;
(4) Two-way asynchronous sentiment session generation method E-SCBA, which
can generate text with a logical and emotional degree;
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
238
(5) An emotional interaction model based on the guided cognitive reassessment
strategy GCRs can reduce the robot's dependence on external emotional stimuli and,
to some extent, prompt positive emotional expressions.
5.2. EMOTIONAL ACCURACY ANALYSIS
To avoid the ambiguity of the robot's emotional expression that makes it difficult for
the participants to recognize the response emotion state, the response emotion state
should have a certain degree of accuracy in the expression of the expected emotion
category. To visually evaluate the accuracy of the robot's emotion generation state
under the action of each model, the accuracy of the target emotion category of the
response emotion is calculated:
(28)
The results are shown in Table 2. As can be seen from Table 2, the models in this
paper are better than other models in terms of sentiment accuracy, which is mainly
because the confidence of the input response sentiment state to each basic sentiment
state transfer probability is used as the update factor when the robot sentiment state
transfer probability is updated. This is mainly because the confidence of the input
response emotional state to each basic emotional state transfer probability is used as
the update factor in this paper, which effectively increases the influence of the input
response expected emotional category on the robot's emotional state transfer
probability.
Table 2. Statistical table of sentiment accuracy for different models
5.3. RETRIEVING TRANSLATION VALIDITY
To facilitate the verification of the effectiveness of model answer retrieval
translation, two information retrieval evaluation indexes, MRR and MAP, were used to
calculate the sorting accuracy of each model candidate answer, 60 sentences were
randomly selected from the test set for the experiment, and the average of the sorting
accuracy was taken as the final result of the experiment, and the results are shown in
Figure 5.
Acc(Ek+1
RH )=Pk+1(i)−
1
5∑
j≠i
Pk+1(j)i,j= 1,2,⋯
,6
Cognitive models Accuracy
ECM 0.785
GCRs 0.821
E-SCBA 0.802
SentiGAN 0.856
This article 0.895
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
239
Figure 5. Different models retrieve translation accuracy statistics
Figure 5 shows the statistical results of the average accuracy of retrieving
translation ranking for different cognitive model answers (m=6), and it can be seen
from the table that this paper's model achieves relatively satisfactory results compared
with other models. This is because the model in this paper ensures more effective
retrieval by combining quantitative evaluation of contextual affective states and
quantitative analysis of factors influencing human-like affective states when ranking
candidate answers, and transforming the translation model into a reverse machine
translation model by incorporating Lagrangian factors in the English translation model,
to obtain high-quality English translation results. Reinforcement learning is used to
establish the correlations between contextual long-term affective states to achieve a
comprehensive and optimal assessment of the following state response with better
cognitive affective ability.
5.4. VALIDATION OF INTERACTION TRANSLATION
To effectively evaluate the effectiveness of interactive sessions, 20 volunteers were
invited to participate in multiple human-computer interaction translation tests under
different models in this paper. At the same time, to increase the objective
comparability among the models, each model was subjected to 30 rounds of multi-
round human-computer interaction conversation experiments. Thirty English
sentences were randomly selected from the test set and used as the initial input for
each model to conduct interactive sessions. The average number of conversations
and the average interaction time for each model are shown in Table 3.
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
240
Table 3. Conversation translation rounds and interaction time statistics table
As shown in Table 3, this model outperforms other models in terms of the average
number of conversation rounds and the average interaction time, which indicates that
the translation robot under the effect of this model can better express the meanings
expressed in English, can better communicate with people in English, and can
effectively extend the human-robot interaction session time. This is because the
response emotions obtained from the model in this paper are more diverse, positive
and accurate by considering human-like emotion generation in the continuous space
of multiple emotion states and combining with the robot's emotion state update, which
effectively guides the participants to participate in human-robot interaction.
5.5. MODEL SATISFACTION ASSESSMENT
To evaluate the model satisfaction effectively, this paper conducts questionnaire
experiments in two aspects: single-round sentence translation and dialogue subjective
satisfaction, and multi-round sentence translation and conversation subjective
satisfaction. The evaluation indexes of subjective satisfaction with single-round
sentence translation and conversation are rationality, diversity, and empathy. The
experiment process is as follows: 100 question-and-answer phrases are randomly
selected from the test set for the test, 500 question-and-answer pairs are used in total,
and 200 volunteers are invited to conduct the questionnaire survey online and offline
through multiple channels. The evaluation indexes of subjective satisfaction of multi-
round sessions were fluency, positivity, interestingness and participation. The
experimental process is as follows: based on the evaluation indexes, a multi-round
session satisfaction survey is conducted for 20 HCI volunteers in the validation of
interactive sessions. At the same time, all indicators were evaluated using a three-
point scale (0,1,2): 0 indicates a low degree, 1 indicates an average degree, and 2
indicates a high degree. The final statistical results were averaged, and the higher the
score the higher the model satisfaction. The results of the model's single-round
sentence translation and dialogue subjective satisfaction survey are shown in Figure
6, and the results of the multi-round sentence translation and dialogue subjective
satisfaction survey are shown in Figure 7.
As seen in Figure 6, the model in this paper is significantly better than other models
in terms of dialogue rationality, diversity and empathy, especially in terms of diversity
Cognitive models N(rounds) T(s)
Chatterbot 7 67.51
ECM 10 98.54
GCRs 15 137.51
E-SCBA 12 119.47
SentiGAN 10 107.63
This article 18 152.92
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
241
of emotional expressions, which is because this paper makes full use of multiple
emotional states in the emotional space when making emotional decisions, and the
results show that the model in this paper can effectively improve the satisfaction of the
robot's single-round dialogue response in many ways.
Figure 6. Statistical chart of single-round subjective evaluation data
Figure 7. Statistical chart of subjective evaluation data for multiple rounds
As can be seen from Figure 7, the model in this paper has effectively improved the
overall fluency and positivity of the robot's emotional utterance translation expression,
the interestingness of human-robot interaction, and the participant's involvement
compared with other models, indicating that the contextual long-term dependency
relationship and the factors influencing emotion generation established in this paper
are reasonable and effective in the construction of the emotional interaction model,
which can further increase the participant's willingness to interact with human-robot
and build a natural and harmonious human-robot interaction relationship.
6. CONCLUSION
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
242
Robot cognitive-based affective interaction computing is to give computers the
ability to observe, understand and generate various emotional states similar to human
beings, so that they can interact naturally and intimately, vividly and interestingly like
human beings. In this paper, we propose a reinforcement learning-based emotional
interaction model for robot cognition. First, we use reinforcement learning to model the
emotion generation process, and use the one-dimensional emotion model theory as
the emotion state space of the robot, with small granularity of emotion division and
delicate expression, which motivates the robot to improve efficiency in the process of
emotional interaction; second, we consider quantifying the three emotional influencing
factors of similarity, positivity and empathy as the reward function for conducting
reward function for emotional state assessment, and derive the optimal emotional
strategy selection based on the utility function to realize the interaction motive of
emotional support, emotional guidance and emotional empathy for the participants;
thirdly, in the process of environmental protection English articles translation by the
robot, the Lagrange factor is introduced, which makes the process of machine
translation probability maximum transformed into the process of obtaining the highest
value of the auxiliary function at the random state. The retrieval speed of machine
translation is improved, the efficiency of machine translation is enhanced, and high-
precision translation results can be obtained more effectively. Finally, the Chinese-
English parallel question and answer corpus commonly used in environmental
protection articles is used as the experimental data set, and the optimal emotional
state is combined with the optimal emotional state to update the robot's emotional
state transfer probability, to realize the robot's state transfer in the translation process
and ensure the continuity of the translation process. The experiments verified the
validity of the model in terms of accuracy, MAP and MRR, and also proved that the
robot cognitive-emotional interaction model can competently translate environmental
protection English articles as a whole with faster translation efficiency and more
accurate retrieval translation quality. Due to the complexity of the human emotion
generation process and the diversity of factors influencing the probability of emotion
state transfer, the model in this paper only considers some of the influencing factors in
the process of emotion generation and translation of English articles. Therefore, future
work needs to consider all the factors influencing the process of human emotion
generation and translation of English texts to further optimize human-like emotion
state generation, so that the robot cognitive-emotional interaction model can help
people in all aspects of daily life in a real sense.
DATA AVAILABILITY
The data used to support the findings of this study are available from the
corresponding author upon request.
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
243
CONFLICTS OF INTEREST
The author declares that there is no conflict of interest regarding the publication of
this paper.
REFERENCES
(1) Sovacool, B. K., & Rio, D. (2021). Corrigendum to "Smart home technologies
in Europe: a critical review of concepts, benefits, risks and
policies" [Renew Sustain Energy Rev 120 (2020) 109663]. Renewable and
Sustainable Energy Reviews, 111277.
(2) Hoskote, M., Le, G., Cherian, R., et al. (2021). Cancer patient perspectives on
survivorship goals from the Smart Patients online community. Supportive
Care in Cancer, 29(5), 2375-2384.
(3) Calp, M. H., Butuner, R., Kose, U., et al. (2022). IoHT-based deep learning
controlled robot vehicle for paralyzed patients of smart cities[J]. The
Journal of Supercomputing, 78(9), 11373-11408.
(4) Souza, A., Filho, M. R., & Soares, C. (2020). Production and evaluation of an
educational process for human-computer interaction (HCI) courses[J].
IEEE Transactions on Education, PP(99), 1-8.
(5) Mmab, C., Kr, B., Aga, B., et al. (2020). Cognitive ability and education: How
behavioral genetic research has advanced our knowledge and
understanding of their association - ScienceDirect[J]. Neuroscience &
Biobehavioral Reviews, 111, 229-245.
(6) Eder, M., Reip, M., & Steinbauer, G. (2021). Creating a robot localization
monitor using particle filter and machine learning approaches[J]. Applied
Intelligence, 1.
(7) He, Y., Yang, S., Chan, C. Y., et al. (2020). Visualization analysis of intelligent
vehicles research field based on mapping knowledge domain[J]. IEEE
Transactions on Intelligent Transportation Systems, PP(99), 1-16.
(8) Chetan, M. R., Dowson, N., Price, N. W., et al. (2022). Developing an
understanding of artificial intelligence lung nodule risk prediction using
insights from the Brock model[J]. European Radiology, 32(8), 5330-5338.
(9) Alghadir, A. H., Gabr, S. A., Al-Momani, M., et al. (2021). Moderate aerobic
training modulates cytokines and cortisol profiles in older adults with
cognitive abilities[J]. Cytokine, 138, 155373.
(10) Panahiazar, M., Chen, N., Lituiev, D., et al. (2021). Empowering study of
breast cancer data with application of artificial intelligence technology:
promises, challenges, and use cases[J]. Clinical & Experimental Metastasis.
(11) Martinez, E. O., Jorns, J. M., Kong, A. L., et al. (2022). ASO Visual Abstract:
Primary Breast Neuroendocrine Tumors—An Analysis of the National
Cancer Database[J]. Annals of Surgical Oncology, 29(10), 6349-6349.
(12) Griebling, T. L. (2020). Re: Do Canadian Radiation Oncologists Consider
Geriatric Assessment in the Decision-Making Process for Treatment of
Patients 80 Years and Older with Non-Metastatic Prostate Cancer? National
Survey[J]. The Journal of Urology, 204(2), 361-361.
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
244
(13) Kuo, J. Y., Chen, C. H., Roberts, J. R., Chen, L. C., & Yang, C. C. (2020).
Evaluation of the user emotional experience on bicycle saddle designs via
a multi-sensory approach. International Journal of Industrial Ergonomics, 80,
103039.
(14) Arana, S., Marquand, A., Hultén, A., Richiardi, J., Loth, E., & Murphy, D. G.
(2020). Sensory modality-independent activation of the brain network for
language. The Journal of Neuroscience, 40(14), 2271-2219.
(15) Lca, B., Wsa, B., Yu, F., Cai, L., & Li, H. (2020). Two-layer fuzzy multiple
random forests for speech emotion recognition in human-robot
interaction. Information Sciences, 509, 150-163.
(16) Zucco, C., Calabrese, B., & Cannataro, M. (2017). Sentiment analysis and
affective computing for depression monitoring. In 2017 IEEE International
Conference on Bioinformatics and Biomedicine (BIBM) (pp. 1988-1995). IEEE.
(17) Liu, X., Xie, L., & Wang, Z. (2017). Empathizing with emotional robot based
on cognition reappraisal. China Communications, 14(9), 100-113.
(18) Wu, D., Han, X., Yang, Z., Li, B., & Zhang, Y. (2021). Exploiting transfer
learning for emotion recognition under cloud-edge-client collaborations.
IEEE Journal on Selected Areas in Communications, 39(2), 479-490.
(19) Kraljević, N., Schaare, H. L., Eickhoff, S. B., & Nickl-Jockschat, T. (2021).
Behavioral, anatomical and heritable convergence of affect and cognition
in superior frontal cortex. NeuroImage, 243, 118431.
(20) Loev, S. (2022). Epistemic feelings are affective experiences. Emotion
Review, 14(1), 1-11.
(21) Zhong, Q. B., Fang, B. F., Guo, X. P., Li, Z. J., & Zhang, J. (2021). Task
allocation for affective robots based on willingness. IEEE Access, 9,
80028-80042.
(22) Culot, C., & Gevers, W. (2021). Happy is easy: The influence of affective
states on cognitive control and metacognitive reports. Cognition and
Emotion, 35(12), 1195-1202.
(23) Zhang, R., Wang, Z., & Mai, D. (2017). Building emotional conversation
systems using multi-task Seq2Seq learning. In The Sixth CCF International
Conference on Natural Language Processing and Chinese Computing (pp.
612-621). Springer.
(24) Rodríguez L. F., Gutiérrez-García, J. O., & Ramos, F. (2016). Modeling the
interaction of emotion and cognition in autonomous agents. Biologically
Inspired Cognitive Architectures, 17, 57-70. https://doi.org/10.1016/
j.bica.2016.07.008
(25) Fatahi, S., & Moradi, H. (2016). A fuzzy cognitive map model to calculate a
user's desirability based on personality in e-learning environments.
Computers in Human Behavior, 63, 272-281.
(26) Mohsin, M. A., & Beltiukov, A. (2019). Summarizing emotions from text using
Plutchik's wheel of emotions. In CY/7th Scientific Conference on Information
Technologies for Intelligent Decision Making Support (ITIDS2019) (pp. 291-294).
Atlantis Press.
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
245
(27) Ali, M., Al Machot, F., Haj Mosa, A., et al. (2018). A globally generalized
emotion recognition system involving different physiological signals.
Sensors, 18(6), 1905.
(28) Pan, H., Xie, L., Lv, Z., et al. (2020). Hierarchical support vector machine for
facial micro-expression recognition. Multimedia Tools and Applications,
79(41), 31451-31465.
(29) Qi, X., Wang, W., Guo, L., et al. (2019). Building a Plutchik's wheel inspired
affective model for social robots. Journal of Bionic Engineering, 16(2),
209-221.
(30) Belkaid, M., Cuperlier, N., & Gaussier, P. (2018). Autonomous cognitive
robots need emotional modulations: Introducing the eMODUL model. IEEE
Transactions on Systems, Man, and Cybernetics: Systems, 49(1), 206-215.
(31) Maqache, N., & Swart, A. J. (2021). Remotely measuring and controlling
specific parameters of a PV module via an RF link. 3C Tecnología. Glosas de
innovación aplicadas a la pyme, 10(4), 103-129. https://doi.org/
10.17993/3ctecno/2021.v10n4e40.103-129
(32) Che Xiangbei, Li Man, Zhang Xu, Alassafi Madini O., & Zhang Hongbin. (2021).
Communication architecture of power monitoring system based on
incidence matrix model. Applied Mathematics and Nonlinear Sciences, 7(1),
83-92. https://doi.org/10.2478/AMNS.2021.1.00098
https://doi.org/10.17993/3ctic.2023.121.222-246
3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529
Ed.42 | Iss.12 | N.1 January - March 2023
246
