Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Unravelling Dynamics of Migrant Solidarity:

A Comprehensive Analysis of Social Media
Discourse Amidst Crisis
Bontha. Mamatha1 Teja Chalikanti2
1 2
Assistant Professor, Department of computer science and engineering,
Department of computer science and engineering, Guru Nanak Institutions Technical Campus, Ibrahipatnam,
Guru Nanak Institutions Technical Campus, Ibrahipatnam, Telangana, India
Telangana, India

Abstract:- Expanding upon our prior research endeavors, In a significant contribution to the field, we also
our study delves into the dynamics of communal support address the potential pitfalls associated with scrutinizing
towards migrants and refugees in the wake of trends in social cohesion. For instance, we underscore
transformative events, focusing specifically on the period how the balance between solidarity-affirming and anti-
before and after the emergence of the COVID-19 solidarity sentiments can be influenced by various factors,
pandemic. We achieve this through the meticulous such as the choice of sampled tweets, the linguistic
collection and examination of an extensive and original characteristics of these tweets, the national identification
dataset composed of tweets associated with migration. of Twitter users (whether known or anonymous), and the
Our primary objective entails the evaluation of meticulous selection of pertinent tweets.
alterations in social cohesion and solidarity expressed
towards migrants during these distinct temporal phases. Keywords:- Solidarity, Crisis, Social Media, Long Short-
Term Memory(LSTM), Random Forest, Social Dynamics,
To initiate this investigation, we meticulously assess Trend Analysis, Societal Responses, Anti-Solidarity.
a corpus of more than 2000 tweets, deciphering the
presence of either supportive or adverse sentiments I. INTRODUCTION
towards immigrants. This is accomplished by employing
two distinct approaches for annotation: one reliant on the In times of crisis, the online realm becomes a crucial
expertise of individuals and the other soliciting space where shifts in pro- or anti-social behaviors are
contributions from the general public. Building upon displayed, carrying significant implications for both societal
these annotated tweets, we develop a Long Short-Term cohesion and political mobilization. This study centers its
Memory (LSTM) model, enriched by a multitude of data focus on the realm of Twitter, employing the context of
augmentation strategies. Impressively, the performance expressions concerning migrant solidarity and anti-solidarity
of this model approaches the upper echelons of human as a lens through which to unravel these intricate dynamics.
accuracy. This finely-tuned model serves as the Social solidarity, a concept denoting individuals' willingness
foundation for the subsequent automated labeling of over to extend aid and share resources beyond self-interest, serves
240,000 tweets spanning the period from September 2019 as a cornerstone of pro-social conduct that upholds societal
to June 2021. functioning in moments of turmoil. Conversely, the
manifestation of anti-solidarity sentiments directed towards
Through a meticulous analysis of these automated migrants and refugees represents a unique facet of antisocial
labels, we elucidate the evolving landscape of migrant- behavior, encompassing actions that lack empathy and
oriented sentiments over this critical period, consideration for the well-being of others. This interplay is
encompassing both the prelude and the aftermath of the pivotal, as the fabric of a well-operating society hinges upon
COVID-19 outbreak. Notably, our findings underscore the predominant absence of such antisocial tendencies.
the escalating prominence and contentiousness of Fundamental to these conceptualizations is the assessment of
expressions of solidarity towards migrants during the individuals' altruistic orientations or rejections thereof,
initial phases of the pandemic. However, as the timeline particularly concerning out-groups such as migrants and
progresses, this solidarity seems to recede in significance, refugees.
with a slight dip in tweet volume below levels observed in
more standard contexts by the summer of 2021. Amid the backdrop of the ongoing COVID-19 crisis,
Interestingly, a subset of tweets linked to the COVID-19 the significance of online media has surged, especially during
crisis displays an elevated proportion of sentiments prolonged lockdowns that confined individuals to their
opposing solidarity. homes. This digital landscape has evolved into a conduit for
engaging with the external world, consuming news, and
voicing political viewpoints. Correspondingly, recent
investigations suggest an upswing in both anti-social

IJISRT23AUG1644 www.ijisrt.com 2100

Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
behaviors and their digital counterparts during times of crisis. influence the balance between expressions of solidarity and
These trends are observable across both online platforms and anti-solidarity.
real-world contexts. Yet, the durability of these dynamics and
their intricate relationships remain shrouded in uncertainty. II. LITERATURE SURVEY
Within this research sphere, the integration of Natural
Language Processing (NLP) techniques to dissect online pro- “Mediating solidarity, Global Media and
and anti-social behaviors within textual content from social Communication” by N. Fenton [1], the rise of alternative
media platforms emerges as a potent tool. Such an approach political media and identity politics has led to political
holds promise in providing insights that can illuminate social fragmentation, challenging traditional allegiances.
scientists and policymakers, supplementing conventional Postmodern theorists view fragmentation positively for
survey-based evidence that probes potential threats to social recognizing diverse political desires and debunking
cohesion and the vulnerability of marginalized social homogeneity. Yet, political efficacy requires more than
clusters. diversity; solidarity is essential for a viable political
community. While new communication technologies mediate
Presently, endeavors to comprehend online pro- and global solidarity, linking online politics to real-world
anti-social behavior confront several limitations that temper movements remains a challenge.
their conclusions with bias. These deficiencies underscore the
significance of our current study, which seeks to rectify such “The emotional antecedents of solidarity in social media
shortcomings. By engaging a comprehensive exploration of crowds, New Media & Society” by D. Margolin and W. Liao
sentiments present in a textual dataset extracted from Twitter, [2], this study investigates how emotional expression within
our research endeavors to offer a nuanced and accurate social media crowds affects their solidarity dynamics.
portrayal of the dynamics surrounding migrant solidarity and Focusing on crowds discussing ongoing National Football
anti-solidarity expressions. Such an inquiry holds the League games, the research utilizes game outcomes as quasi-
potential to refine our understanding of the underlying random influences to mitigate confounding variables.
mechanisms that steer individuals' responses to crises within Findings reveal that crowd participation is self-sustaining
digital spaces. over a week but can be triggered or dampened within an hour
based on expressed emotions. Anger fosters participation,
To address these critical gaps, this study embarks on a sadness hinders it, while positive emotions and anxiety have
multifaceted investigation that involves the meticulous complex connections to engagement. This sheds light on
annotation of over 2000 tweets, a process that categorizes emotional drivers of crowd behavior and their impact on
sentiments as either aligned with solidarity or in opposition to sustained participation and solidarity dynamics.
it. Leveraging a two-pronged approach for annotation
involving both domain experts and crowdsourced “The role of grassroots food banks in building political
contributors, our methodology aims to yield robust and solidarity with vulnerable people” by S. Koos and V. Seibel
comprehensive assessments. These annotated tweets [3], Amid Spain's economic crisis, food banks emerged as
subsequently form the basis for training a Long Short-Term emergency solutions, criticized for perpetuating dependency
Memory (LSTM) model. Augmented with diverse strategies, and inequality. However, in Madrid post the 15M movement,
this model's performance nearly rivals human-attained grassroots food banks transformed assistance into solidarity,
accuracy. By leveraging the precision of this model, we particularly in Tetuán. Political and interpersonal solidarity
conduct automated labeling for a staggering 240,000 tweets thrived within these banks, fostering inter-recognition and
spanning the timeframe from September 2019 to June 2021. identifying the roots of recipients' hardships. This cultivates a
sense of cohesion, challenging inequality and stigma inherent
In analyzing the automated labels, our study strives to in traditional charity models. Grassroots food banks not only
shed light on the temporal trajectory of sentiments pertaining provide aid but also forge new solidarity avenues, countering
to migrant solidarity and its counterpoint during the prelude hierarchical relationships found in formal food banks, thus
and progression of the COVID-19 crisis. Noteworthy shifts in promoting social inclusion.
the prominence and contestation of these sentiments are
brought to the forefront of our analysis. Early stages of the “Stereotypical gender associations in language have
pandemic see an upsurge in the saliency of migrant solidarity decreased over time, Sociological Science” by J. J. Jones, M.
expressions, which, however, wane in importance since the R. Amin, J. Kim, and S. Skiena [4], Utilizing a vast digital
latter part of 2020, with tweet volumes dipping slightly book collection, we track the evolution of gender stereotypes
below levels witnessed prior to the crisis by the summer of in written English from 1800 to 2000. Employing word
2021. Interestingly, a subset of tweets associated with the embeddings, we quantify male gender bias across four
COVID-19 crisis manifests an increased ratio of anti- domains: career, family, science, and arts. While gender
solidarity sentiments. associations have diminished over time, biases persist. Career
and science terms exhibit positive male gender bias, while
Moreover, our study contributes a critical layer of family and arts terms show negative bias. We asecond shift,
reflection upon the challenges inherent in dissecting trends finding partial evidence. Traditional gender ideologies are
within social cohesion dynamics. We underscore the intricate embedded in English texts, but their strength appears to wane
interplay of factors such as language, national identification across time.lso examine shifting associations reflecting the
of users, and the precise selection of tweets, all of which

IJISRT23AUG1644 www.ijisrt.com 2101

Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
“Norms of online expressions of emotion: Comparing  Random Forest:
Facebook, Twitter, Instagram, and WhatsApp, New Media & The widely embraced machine learning ensemble
Society” by S. F. Waterloo, S. E. Baumgartner, J. Peter, and technique, Random Forest, is adeptly applicable in both
P. M. Valkenburg [5], This study aimed to explore emotional classification and regression undertakings. Functioning as a
expression norms on social media, focusing on perceived variant of decision tree-based models, it orchestrates multiple
appropriateness of six emotions (sadness, anger, decision trees harmoniously for predictive tasks.
disappointment, worry, joy, and pride) across four platforms.
Data from 1201 young Dutch users (15-25 years) in March  Core Aspects of Random Forest:
2016 revealed positive expressions were generally seen as  Decision Tree Fusion: Random Forest orchestrates
more fitting than negative ones across platforms. Differences diverse decision trees to culminate in predictions. Each
emerged among platforms: negative emotions were most tree is trained on a distinct random data subset.
acceptable on WhatsApp, followed by Facebook, Twitter,  Prediction Aggregation: The ultimate prediction outcome
and Instagram. Positive emotion expression's appropriateness stems from aggregating the predictions of all individual
ranked highest on WhatsApp, then Instagram, Facebook, and trees.
Twitter. Gender differences existed, while age variations  Random Subset Training: The model assembles decision
were limited. trees, each cultivated on a random data subset (employing
bootstrapping), while every split integrates a randomly
III. METHODOLOGY selected subset of features.
 Embedding of Randomness: This deliberate incorporation
The exploration of social solidarity, a vital pro-social of randomness mitigates overfitting, thereby elevating the
behavior, has been a consistent theme in social sciences, model's capacity to generalize beyond training data.
tracing back to thinkers like Rousseau and Durkheim. While
past research predominantly employed surveys and  Outcome Derivation:
qualitative methods, the emergence of computational social For concluding predictions, the prevailing approach
science has introduced novel approaches, like integrating involves either majority voting or the averaging of
solidarity into Natural Language Processing (NLP). Notably, predictions engendered by the entire tree assembly.
computational studies have investigated crises such as the
European Migration Crisis and the Financial Crisis,
spotlighting media discourse disparities and solidarity
discussions across various platforms. However, these
investigations primarily focused on mainstream media, while
social media offers unique avenues for contestation,
reevaluation, and diversification of narratives surrounding
solidarity claims.

 Existing System:
The Random Forest algorithm is a widely used machine
learning technique that enhances prediction accuracy by
combining multiple decision trees. The mechanics of a
Random Forest model are as follows:
 Ensemble of Decision Trees: A Random Forest comprises Fig 1: A Random Forest Overview
multiple decision trees, each trained on a distinct subset
of training data (a bootstrap sample), using a subset of Within our methodology, the potent Random Forest
features at each split. This controlled randomness technique holds sway. Employed for both classification and
prevents overfitting and bolsters the model's regression, it leverages its ensemble of decision trees to
generalization capacity. extrapolate predictive insights. This model is especially
 Training Process: Training involves these steps: a. effective for our investigation, wherein we explore the
Randomly select training data subsets (with replacement) evolving dynamics of sentiments and expressions of
to form bootstrap samples. b. Randomly choose feature solidarity within the context of crisis scenarios. By training
subsets for decision tree splits. c. Construct decision trees individual trees on diverse subsets of data and combining
using the bootstrap samples and chosen features, their outcomes, we foster a holistic understanding of these
recursively partitioning data based on feature values. d. intricate patterns. Furthermore, this approach integrates
Repeat (a) to (c) to create multiple trees. randomness purposefully, countering the pitfalls of
 Prediction: When predicting with a trained Random overfitting that could potentially taint our results. The
Forest model, each decision tree autonomously predicts collective wisdom of these decision trees, facilitated by the
the target variable. For regression, predicted values are Random Forest paradigm, fortifies our endeavor to discern
averaged for the final outcome. For classification, the the underlying sentiments and solidarity expressions in
ensemble applies voting (majority class) or averaging digital discourse during periods of crisis.
(probabilities) to determine the final class label.

IJISRT23AUG1644 www.ijisrt.com 2102

Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 Disadvantages: Within our methodology framework, the LSTM
 Single trees may be visualized as a sequence of decisions algorithm emerges as a pivotal instrument. Tailored to our
 A forest is less interpretable than a single decision tree research objectives, it scrutinizes the unfolding sentiment and
 Lack of interpretability solidarity discourse dynamics during crisis periods.
 Computational complexity Employed to assess large-scale data, the LSTM model
 Overfitting navigates through the intricacies of sequence-based content,
enabling the identification of patterns, trends, and shifts. Via
 Proposed System: continuous training, it learns to capture the nuanced interplay
An LSTM (Long Short-Term Memory) model between evolving sentiments and the expression of
represents a type of recurrent neural network (RNN) community support. The technique's adaptability for diverse
architecture meticulously engineered to grapple with types of sequential data fortifies our exploration across the
prolonged dependencies, effectively mitigating the vanishing temporal dimensions, enhancing our comprehension of the
gradient problem inherent in conventional RNNs. Its emotional landscape and solidarity dynamics amidst crises
widespread applicability encompasses sequence data
manipulation tasks, spanning domains like natural language  Salient aspects of LSTM models include:
processing, speech recognition, and time series analysis.  Memory Cells: The fundamental cornerstone of an LSTM
model resides in its memory cell. This dynamic entity
 LSTM Algorithm for Sequential Data Analysis: facilitates information retention and retrieval across
The Long Short-Term Memory (LSTM) algorithm, a extensive sequences. Featuring a self-loop connection, the
variant of recurrent neural network (RNN) architecture, memory cell ensures unimpeded temporal information
specializes in processing and comprehending sequential data. flow, thereby adeptly capturing long-range dependencies
Its design overcomes the vanishing gradient problem and  Gate Mechanisms: LSTM models incorporate gate
facilitates the assimilation of long-range dependencies in the mechanisms to regulate internal information flow within
data. memory cells. Three pivotal gate types comprise this
architecture:
 LSTM Workflow Highlights:  Forget Gate: Assesses the need to discard specific
 Data Setup: Input data for LSTM demands sequencing, memory cell contents. Informed by prior hidden states
encompassing text, time series, or similar formats. Data is and current inputs, it yields per-element values in the
divided into input and corresponding output sequences. memory cell, signifying the degree of information to
 LSTM Structure: Comprising multiple memory cells, each relinquish
with an internal state, LSTM integrates three core  Input Gate: Determines incoming data for memory cell
elements: input gate, forget gate, and output gate. These storage. It discerns pertinent segments of ongoing inputs
components wield control over information flow, and previous hidden states for inclusion, culminating in a
dictating storage, oblivion, and emission. provisional value designated for memory cell
 Forward Progression: The LSTM algorithm processes augmentation
input sequences incrementally. With each time step, input  Output Gate: Dictates the information to be emitted from
enters the LSTM network, and the model computes the memory cell. Synthesizing contemporary inputs and
outputs based on input, prior hidden state, and preceding preceding hidden states, it adjudicates the constitution of
memory cell state. Gates manage information movement information to be conveyed as the present hidden state
within memory cells  Backpropagation Through Time (BPTT): A vital training
 Backpropagation Through Time (BPTT): Post forward method for LSTM models, BPTT harmonizes with the
movement, the algorithm computes the variance between sequential data context. This variant of backpropagation
predicted and actual outputs, translating to loss propagates error gradients over temporal horizons,
computation. BPTT technique propels the error back in equipping the model to grasp the complete sequence
time to update LSTM network weights, thereby context
minimizing loss  Stacked LSTM: In certain scenarios, the strategic layering
 Training Phase: LSTM undergoes training through of multiple LSTM tiers enriches model depth. Each
iterative weight adjustments, facilitated by gradient stratum absorbs the output of its precursor, thus enabling
descent optimization techniques (e.g., stochastic gradient intricate representation learning
descent, Adam, RMSprop). This iterative process seeks to  Training and Optimization: Conventional training of
minimize the loss function by refining model weights and LSTM models leverages gradient-based optimization
biases schemes, encompassing stochastic gradient descent
 Prediction Capability: Post training, the LSTM algorithm (SGD) and advanced variants like Adam and RMSprop.
transitions into predictive mode. When provided with new The overarching objective entails loss function
input sequences, it generates corresponding output minimization, quantifying disparities between model
predictions. By propelling input through the network, it predictions and actual values
derives outputs at each time step  Applications: LSTM models find successful deployment
across manifold tasks. Natural language processing arenas
include language modeling, sentiment analysis, and
machine translation. Furthermore, they empower speech

IJISRT23AUG1644 www.ijisrt.com 2103

Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
recognition systems, furnish time series forecasting dynamics can be assessed. Ensuring comparability, we
capabilities, and facilitate anomaly detection. maintain a consistent baseline of solidarity discussion prior to
and post the pandemic's outset.

 LSTM Algorithm:
LSTM, denoting long short-term memory networks,
constitutes a subset of Deep Learning techniques. This
variant of recurrent neural networks (RNNs) specializes in
grasping long-range dependencies, especially applicable to
sequence prediction problems. Among RNN types, LSTM
networks are the most prevalent. The crux of LSTM involves
memory cells and gates (including the forget and input
gates), controlling memory cell contents through input and
forget gate modulation.

 Training:
Researchers, particularly those engaged in NLP or data
science, often tailor their machine learning models to suit the
task, like analyzing English-only tweets. However, this
Fig 2: Basic structure of the LSTM model decision can potentially introduce bias to findings. Various
factors can sway results, influencing the conclusions drawn
In our methodology, we harness the potential of LSTM from online discourses.
models, intricately designed for capturing intricate sequential
patterns, by applying them to our large-scale dataset. By  Analyzing Accuracy on Test Set:
exploiting their inherent capacity to handle long-term Our model achieved a commendable accuracy rate of
dependencies and nuanced context, we aim to assess the 93.75% on the test dataset.
evolving dynamics of sentiments and solidarity expressions
within the scope of our research domain. Through diligent Within our methodology, we draw on meticulously
training using advanced optimization techniques, we intend curated datasets comprised of tweets. Our focus lies in
to unravel the intricate relationships that underlie the understanding the transformation of sentiments and solidarity
observed trends and changes in social media discourse over expressions, pinpointing how they evolved preceding and
time. This cutting-edge approach bolsters our endeavor to during the COVID-19 crisis. By adopting LSTM-based Deep
provide a comprehensive understanding of the interplay Learning techniques, we harness the capability of these
between emotions, sentiments, and communal support during networks to grasp intricate patterns in sequence-based data.
crisis periods. This approach aligns with the paradigm shift introduced by
the pandemic and its influence on online discussions.
 Advantages: Notably, the pre-COVID-19 period serves as a benchmark,
 Network is a type of RNN ensuring robust comparisons across these distinctive
 Learning sequential data prediction problems temporal phases.
 LSTM also has some layers which help it to learn and
recognize the pattern for better performance  System Architecture:
 Capturing Long-Term Dependencies
 Handling Variable Length Sequences
 Learning Complex Temporal Patterns
 Handling Noisy and Missing Data

For this project, we have designed following modules

 Dataset:
For German tweets, tweets without specified Twitter
account locations were also included. Tweets were gathered
exclusively based on specific hashtags relevant to our topics:
refugee and financial solidarity.

 Pre-processing:
We employ the unforeseen initiation of the COVID-19
crisis, commencing with the initial European lockdown in
late February to early March 2020. This period enables the
analysis of how social solidarity discourse evolved before
and during the pandemic. The pre-COVID-19 phase
establishes a baseline against which pandemic-related Fig 3System Architecture

IJISRT23AUG1644 www.ijisrt.com 2104

Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Moreover, the choice of dataset and linguistic
limitations, especially with German tweets, plays a role in
shaping our findings. This selection highlights the nuances
that can potentially impact the broader conclusions drawn
from digital dialogues. Through our comprehensive
approach, we intend to unravel the complexities inherent in
online discourse during crisis events, shedding light on how
emotions, sentiments, and solidarity dynamics manifest and
transform across different stages of the crisis

 Output Snapshots: Fig 8: Initially, as it is a classification model, we need to

train the model using labelled data

Fig 4: it is the front-end execution, where user needs to login

Fig 9: After a successful training finally, we can predict the

tweet which belongs to solidarity/ant solidarity

IV. CONCLUSION

This endeavor marks a significant stride by introducing

expansive human and algorithmically annotated datasets,
Fig 5: user login credentials specifically designated for delineating solidarity and its
counterpoint, anti-solidarity. Leveraging the textual fabric of
social media posts, our annotations discern expressions of
(anti-)solidarity directed towards pertinent target groups.
Notably, we attain commendable consensus among both
expert and crowd annotators, despite the intricacies inherent
in this pioneering NLP task. Our methodological ingenuity
extends to augmented LSTM models, whose proficiency
closely parallels expert agreement. Employing these models,
we orchestrate an extensive trend analysis, dissecting over
Fig 6: uploading the dataset on which task is to be performed 270,000 social media posts before and after the COVID-19
pandemic's onset.

Within the domain of migrant solidarity discourses

explicitly interwoven with COVID-19, our scrutiny unveils
an abrupt surge in COVID-19 relevance as Europe
implements initial measures in mid to late March 2020. This
prominence swiftly ebbs, persisting at subdued levels. While
scrutinizing migrant solidarity discourse with COVID-19-
related keywords, (anti-)solidarity trends prove contingent on
subpopulations under study. However, an intriguing pattern
Fig 7: The above preview shows the random tweets which emerges, with all subpopulations showcasing heightened
are recorded from the twitter solidarity during the pandemic's early phase. An isolated
focus on either phenomenon, be it migrant solidarity or anti-
solidarity, can yield a deceptive portrayal of societal
movement toward political orientations.

Our investigation underscores the volatility of outcomes

due to multiple influencing factors, encompassing language
preferences, disclosure of user location, and distinct sampling
strategies, including hashtag inclusion or exclusion.

IJISRT23AUG1644 www.ijisrt.com 2105

Volume 8, Issue 8, August – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Nonetheless, our empirical insights firmly establish that REFERENCES
migrant solidarity experienced an amplified presence and
contention during the pandemic's onset. Subsequently, its [1]. N. Fenton, Mediating solidarity, Global Media and
prominence dwindled, modestly falling below pre-pandemic Communication, vol. 4, no. 1, pp. 37–57, 2008.[1] D.
levels by summer 2021. Noteworthy is the concurrent rise in Margolin and W. Liao, The emotional antecedents of
anti-solidarity tweets within a subset of COVID-19-related solidarity in social media crowds, New Media &
posts, albeit outweighed by the prevalence of solidarity Society, vol. 20, no. 10, pp. 3700–3719, 2018.
expressions. [2]. S. Santhanam, V. Srinivasan, S. Glass, and S. Shaikh, I
stand with you: Using emojis to study solidarity in crisis
This study serves as a clarion call for mindful research events, arXiv preprint arXiv: 1907.08326, 2019.
design, advocating long-term observation, judicious pre- and [3]. Z. Tufekci, Social movements and governments in the
post-crisis comparisons, and meticulous sampling strategies. digital age: Evaluating a complex landscape, Journal of
The evanescent nature of crisis-related effects necessitates International Affairs, vol. 68, no. 1, pp. 1–18, 2014.
such considerations. Through our comprehensive [4]. H. Silver, Social exclusion and social solidarity: Three
methodology and insightful findings, we contribute to the paradigms, International Labour Review, vol. 133, nos.
wider discourse on societal dynamics during periods of 5&6, pp. 531–578, 1994.
upheaval, shedding light on the complex interplay between [5]. D. Clarke, Pro-Social and Anti-Social Behaviour. New
sentiments, expressions of solidarity, and the broader socio- York, NY, USA: Routledge, Taylor & Francis Group,
political landscape 2003.
[6]. M. Mohler-Kuo, S. Dzemaili, S. Foster, L. Werlen, and
V. FUTURE ENHANCEMENT S. Walitza, Stress and mental health among children/
adolescents, their parents, and young adults during the
Our investigation has effectively illuminated the first COVID-19 lockdown in Switzerland, International
susceptibility of outcomes to various determinants. These Journal of Environmental Research and Public Health,
encompass the language employed by users in their tweets, vol. 18, no. 9, p. 4668, 2021.
the disclosure or concealment of their geographic origin, and [7]. sCAN project, Hate speech trends during the COVID-
the selection of sampling strategies, whether enriched by 19 pandemic in a digital and globalized age,
hashtags or not. An essential lesson lies in the understanding https://scanproject.eu/resources-and-
that surges of (anti-)solidarity might not be intrinsically publications/#Covid-19, 2021.
linked to the pandemic itself; they could potentially emerge [8]. M. L. Williams, P. Burnap, A. Javed, H. Liu, and S.
from concurrent, interwoven political occurrences. Thus, our Ozalp, Hate in the machine: Anti-black and anti-muslim
analysis calls for a profound delve into the data to prevent the social media posts as predictors of offline racially and
propagation of deceptive conclusions. religiously aggravated crime, The British Journal of
Criminology, vol. 60, no. 1, pp. 93–117, 2020.
While our study provides robust testament, it also [9]. C. R. Seiter and N. S. Brophy, Social support and
beckons further exploration. We acknowledge that aggressive communication on social network sites
pronounced (anti-)solidarity instances might be intertwined during the COVID-19 pandemic, Health
with broader political narratives, necessitating a meticulous communication, doi: 10.1080/10410236.2021.1886399
evaluation of the underlying dynamics. Our findings remain a
steadfast testament that migrant solidarity underwent a
discernible escalation and contestation during the pandemic's
advent. Nevertheless, the prominence subsequently
experienced a wane from late 2020, with tweet volume
inching slightly below pre-pandemic thresholds by the
summer of 2021.

Within the same timeframe, an intriguing revelation

unfolded – a rise in anti-solidarity tweets within a subset of
COVID-19-related posts, albeit eclipsed by the
preponderance of solidarity messages. As we chart a path
towards future research, these insights underscore the pivotal
significance of methodological decisions. The integral role of
long-term observation, meticulous pre- and post-crisis
comparisons, and astute sampling strategies in crisis-focused
inquiries takes center stage. As we chart a course for future
enhancements, these considerations are poised to shape a
comprehensive and nuanced understanding of societal
dynamics during times of upheaval, steering us toward a
more informed comprehension of the intricate fabric of
sentiment and solidarity expressions.

IJISRT23AUG1644 www.ijisrt.com 2106