Engineering and Technology Journal e-ISSN: 2456-3358

Volume 10 Issue 05 May-2025, Page No.- 5177-5185

DOI: 10.47191/etj/v10i05.46, I.F. – 8.482
© 2025, ETJ

Fake News Detection Using Natural Language Processing (NLP) and

Machine Learning
Hisham Ahmed Mahmoud1, Ibrahim M. Ibrahim2
1
Akre University for Applied Sciences Technical College of Informatics Directorate of Educational Training and
Development/Duhok
2
Akre University for Applied Sciences Computer Networks and Information Security Technical College of Informatics-Akre

ABSTRACT: Fake news has become a significant social problem that affects public opinion and social trust. The dissemination of
misinformation at lightning speed through online media channels presents a serious problem in distinguishing between credible and
deceptive content. Much research has been done on machine learning and natural language processing (NLP) approaches to create
automated detection systems that address this problem. These methods classify news stories as synthetic or legitimate based on
statistical, contextual, and language characteristics. Several machine learning techniques are presented in this book, including
deeper, more complex models as well as more conventional classifiers like Logistic Regression (LR), Naïve Bayes (NB), Support
Vector Machines (SVM), and Decision Trees. Additionally, the hybrid models that combine ensemble learning and natural language
processing (NLP) are showcased for their potential to improve classification accuracy and dependability. Feature extraction
techniques like word embeddings, N-grams, and Term Frequency-Inverse Document Frequency (TF-IDF) are credited with
organizing textual data into machine learning inputs. In addition, network-based techniques and sentiment analysis are examined as
substitute techniques for pattern recognition in the fight against false information. The findings provide strong evidence that
combining statistics and deep learning techniques enhances the effectiveness and precision of false news detection. This study
establishes the groundwork for future research to create robust and scalable systems for identifying fake news by shedding light on
the capabilities and limitations of various approaches.

KEYWORDS: Fake News Detection; Natural Language Processing (NLP); Machine Learning; Deep Learning; Text Classification;
Feature Extraction; Sentiment Analysis; Hybrid Models; Transformer Models; Word Embeddings; TF-IDF; Misinformation.

1. INTRODUCTION make accurate judgments regarding false or misleading

The dissemination of false information has become one of the information[6]. In such a scenario, Natural Language
most urgent concerns of the digital age. With the advent of Processing (NLP) and Machine Learning (ML) are cutting-
fast, user-driven models such as social media, messaging edge technologies that can enable automated systems to read,
apps, and networked blogging, the distribution of false analyze, and classify news stories based on linguistic,
information is no longer constrained by traditional editorial semantic, and statistical patterns[7]. Machine learning
screening mechanisms[1, 2]. This unfiltered information models can be trained to detect subtle text cues and anomalies
dissemination has wide-ranging effects ranging from that are typical of deceptive content[8]. These include
influencing popular perceptions and dictating electoral sensationalized titles, emotive language, semantic
outcomes to spreading health misinformation and inducing dissonance, and stylistic deviance from genuine reporting.
social unrest[3]. Particularly at such historic event such as the Coupled with NLP models such as tokenization,
2016 U.S. Presidential Election and global-wide COVID-19 lemmatization, sentiment analysis, and syntactic parsing,
pandemic, disinformation succeeded in outpacing official such models can form the basis for a robust fake news
sources of information as far as visibility and engagement are detection system. Slightly more sophisticated techniques
concerned[4]. The traditional mechanisms for verifying using word embeddings (e.g., Word2Vec, GloVe),
news—relying on human fact-checkers, journalists, or vectorization techniques (e.g., TF-IDF, n-grams), and deep
regulatory agencies—are inherently slow and insufficient to learning architectures (e.g., RNNs, LSTMs, CNNs) further
deal with the amount and pace at which content is produced enhance the system in learning from context and complex
online[5]. Thus, there exists a pressing necessity for systems language constructs[9]. Further recent developments in
to automatically process vast amounts of text in real-time and transformer models, i.e., BERT, RoBERTa, and GPT, have

5177 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025

 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
changed the game in the domain by introducing pre-trained terms were reduced to their basic forms using stemming and
models capable of understanding fine-grained semantic lemmatization. In order to reduce noise and improve the
dependencies between words and phrases[10]. These models quality of the input data, text normalization—which includes
far outperform traditional ML methods with the assistance of lowercasing and punctuation removal—was also carried out.
deep contextual information and transfer learning so that it is
2.3 Feature Engineering
now achievable to identify fake news more precisely and
Numerical features were taken from the text to be used as
effectively[11]. This paper provides a comprehensive
inputs for the machine learning algorithms once the data had
comparison of various machine learning and NLP-based
been preprocessed. This was achieved by employing
methods for detecting fake news. It puts traditional classifiers
techniques that aid in capturing word significance and
such as Logistic Regression, Naïve Bayes, and Support
sequence patterns, such as Count Vectorization, N-gram
Vector Machines against existing deep learning models and
modeling, and Term Frequency-Inverse Document
transformer-based models. It also talks about hybrid methods
Frequency (TF-IDF). Feature selection methods like the Chi-
that combine statistical models with neural networks to
Square Test and Mutual Information were applied to increase
leverage the strength of both[12]. The study evaluates the
efficiency and concentrate the model on the most pertinent
performance of these models using public datasets and
data. By highlighting important characteristics and reducing
compares their accuracy, scalability, and generalizability to
dimensionality, these techniques aid in precise
practical applications. Moreover, the paper stresses the
categorization.
challenge of translating disinformation tactics, such as AI-
created content and multimodal fakes, and underlines the 2.4 Model Development and Training
necessity of explainable AI in crafting trustworthy detection Numerical features were taken from the text to be used as
systems[13]. In general, fake news detection is a inputs for the machine learning algorithms once the data had
multidisciplinary problem that must be solved by integrating been preprocessed. This was achieved by employing
linguistic analysis, algorithmic creativity, and ethical techniques that aid in capturing word significance and
concerns. By leveraging the capabilities of NLP and machine sequence patterns, such as Count Vectorization, N-gram
learning, this research aims to help develop intelligent modeling, and Term Frequency-Inverse Document
systems that can aid in digital literacy, safeguard information Frequency (TF-IDF). Feature selection methods like the Chi-
integrity, and ultimately build a more enlightened global Square Test and Mutual Information were applied to increase
society. efficiency and concentrate the model on the most pertinent
data. By highlighting important characteristics and reducing
2. RESEARCH METHODOLOGY dimensionality, these techniques aid in precise
The methodology used to develop a false news detection categorization.
system based on machine learning (ML) and natural language 2.5 Evaluation and Future Scope
processing (NLP) techniques is described in this section. The Several widely used performance measures—accuracy,
method uses a systematic pipeline that starts with data precision, recall, and F1-score—were used to holistically
gathering and ends with assessment and recommendations for assess the performance of the trained machine learning
further developments. models. These measures provide a multi-dimensional
2.1 Dataset Collection perspective on how accurately a model can detect real and
A wide range of authentic and fraudulent news stories are fake news. Accuracy indicates the overall accuracy of the
included in the dataset utilized in this study, which was predictions, whereas precision indicates the proportion of true
obtained via Kaggle. To aid with supervised learning, each fake news identified out of all items identified as fake. Recall,
item has a label. The dataset, which includes a variety of news in contrast, measures the capability of the model to accurately
items to ensure fair representation and generalizability, is detect all the existing fake news articles. The F1-score, a
crucial for training and testing classification algorithms. The harmonic mean of recall and precision, is especially useful
work promotes equitable model evaluation and adheres to when working with imbalanced datasets, where one class may
reproducible research principles by utilizing a carefully significantly dominate the other—a real-world scenario in
selected and publicly accessible dataset. fake news detection [Link] the current models
demonstrating strong classification power and producing
2.2 Text Preprocessing
positive outcomes, there remains ample room for
The raw text data was preprocessed using a number of
improvement in adjusting to the constantly shifting landscape
methods to guarantee clean and consistent input for machine
of misinformation. Future research has to aim at creating real-
learning models. Tokenization, which divides the text into
time fact-checking solutions that can perform well within
discrete tokens or words, was a step in this process.
dynamic environments where disinformation spreads rapidly
Stopwords and unnecessary letters were then eliminated.
across different platforms. Secondly, the use of Explainable
Additionally, to improve uniformity throughout the dataset,
5178 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025
 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
Artificial Intelligence (XAI) needs to be looked into for beneficial representations from unlabelled, raw text.
greater transparency and trust, particularly in high-stakes Together, these innovations hold the promise of creating
applications where model outputs need to be explainable and more adaptive, scalable, and consistent fake news detection
accountable. Another avenue with promise is researching systems capable of addressing the advanced and adaptive
Self-Supervised Learning, which could decrease dependence techniques of modern disinformation campaigns.
on large amounts of labeled data by enabling models to obtain

Figer1: Structured Methodology for Automated Fake News Classification

4. LITERATURE REVIEW accuracy even more by combining statistical learning and

Identification of false news has been an emerging area of deep learning models[21]. Real-time detection of fake news
research, and various studies have utilized machine learning has driven innovation in user behavior analytics, network-
and NLP techniques to increase precision[14, 15]. Language based detection, and explainable AI models to enhance
and semantic examination, such as sentence structure, use of transparency[22]. All these despite, fake news detection
vocabulary, and coherence, have been researched by scholars, remains challenging due to the dynamic nature of
and researchers have discovered that false news articles misinformation, innovation of multimodal fake news (text,
usually utilize hyperbolic word choice, lower lexical images, and videos), and algorithmic bias concerns[23].
diversity, and extreme readability measures[16]. Sentiment Future research promises to develop adaptive, scalable, and
and emotional analysis show that fake news employs intense ethically responsible models in order to counter
emotions, typically playing on the reader's emotions through misinformation more effectively in a continuously changing
emotive language[17]. Stylometric analysis, examining digital space.
writing style and syntactic structures, has proven successful
in identifying fake content[18]. Traditional machine learning 4.1 Linguistic and Semantic Analysis
models such as Naïve Bayes, SVM, and Decision Trees have One of the pioneer approaches in fake news detection is
been widely employed for classification, but with the recent linguistic analysis, which inspects textual characteristics like
advancements in deep learning such as RNNs, LSTMs, sentence structure, use of vocabulary, and semantic
CNNs, and hybrid architectures, detection accuracy has coherence[24-26]. It has been observed by researchers that
significantly improved by learning contextual relations in text fake news articles tend to have:
data[19]. The transformer-based architectures such as BERT, • Exaggerated or sensational language: Fake news
GPT, and T5 employ self-attention to enhance stories frequently use emotionally charged words,
performance[20]. Ensemble and hybrid approaches extend
5179 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025
 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
hyperbolic expressions, and misleading headlines to Semantic analysis techniques such as Latent Semantic
attract attention. Analysis (LSA) and Word Embeddings (Word2Vec, GloVe,
• Lower lexical diversity: Studies suggest that deceptive FastText) have been applied to detect hidden patterns among
articles tend to reuse specific words and phrases, words and detect patterns characteristic of misinformation.
indicating a lack of linguistic variation.
• Sentence complexity and readability: Fake news
articles may exhibit overly simplistic or overly complex
sentence structures, deviating from journalistic norms.

4.2 Advanced Techniques for Fake News Detection scores, also assists in detecting fake news[28]. Machine
Fake news detection has evolved through techniques such as learning algorithms such as Naïve Bayes, SVMs, Decision
sentiment and emotion analysis, which have shown that fake Trees, and Random Forests categorize content using
news tends to incorporate emotionally charged or subjective linguistic features but demand heavy feature engineering. In
language, as opposed to the neutral tone of authentic order to combat increasingly complex disinformation,
news[27]. Emotion classifiers that identify fear, anger, or scientists are integrating deep learning and NLP for higher
surprise assist in identifying deceptive content. Stylometric accuracy.
analysis, such as n-gram patterns, grammar, and readability

Figure 2: Distribution of Core NLP and ML Techniques for Fake News Detection

5180 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025

 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
4.3 Advanced Deep Learning Techniques for Fake News Bayes with deep learning to provide improved accuracy and
Detection explainability[30]. Knowledge-based systems and external
With the rise of misinformation, identifying fake news is fact-checkers assist in making it reliable. User behavior
becoming more crucial with deep learning. RNNs, LSTMs, analysis and network-based tracking are components of real-
and CNNs learn textual patterns and contextual relationships, time detection now. Explainable AI (XAI) is also used to
improving classification accuracy[29]. Hybrid models that build trust. Some of the challenges are the novel
employ a combination of CNNs and LSTMs, and misinformation tactics, the rise of multimodal content, bias,
transformer-based models like BERT, GPT, T5, and XLNet and scalability[31]. Future work seeks to create adaptive,
offer state-of-the-art results with deep language and ethical, and robust AI systems using techniques like self-
contextual understanding. Researchers also use ensemble supervised, reinforcement, and federated learning.
methods and combine statistical models like SVM and Naïve

Figure 3: Proportion of Deep Learning Strategies for Fake News

Classification

5. DISCUSSION AND COMPARISON functions. SVMs perform best with sparse representation
The performance of disinformation detection systems largely such as those produced by TF-IDF or N-gram
relies on the choice of machine learning algorithms, vectorization[36]. They are, however, computationally
representativeness and quality of the dataset, feature intensive and require tuning, especially for larger datasets.
extraction techniques employed, and hyperparameter tuning Decision Trees and Random Forests, as rule-based learners,
approach[32]. Here, a variety of models were explored, enjoy the advantage of interpretability and can learn non-
ranging from simple classifiers to state-of-the-art deep linear relationships. They work well in finding hierarchical
learning and transformer-based models[33]. Of the traditional decision paths but are prone to overfitting, particularly when
machine learning methods, Logistic Regression (LR) and feature engineering is weak or the dataset is imbalanced.
Naïve Bayes (NB) are the most basic, quickest, and efficient Ensemble techniques such as Random Forest and XGBoost
ones, especially for high-dimensional text classification improve model stability by taking the average of numerous
tasks. They require similarly fast computation resources and learners, which improves generalization and reduces the
are interpretable, making them ideal for baseline studies and likelihood of bias[37]. Apart from traditional models, deep
quick deployment in low-resource environments[34]. learning models have made significant advancements in
Logistic Regression is able to produce good performance in identifying fake news since they are capable of learning
cases when text features are linearly separable, but Naïve complex semantic and syntactic patterns from texts.
Bayes, despite relying on the strong assumption of feature Convolutional Neural Networks (CNNs) are adept at
independence, is surprisingly capable of producing good extracting local features and spatial hierarchies from text,
results for natural language processing problems due to word whereas Recurrent Neural Networks (RNNs) and Long Short-
distribution topological properties within text[35]. Support Term Memory (LSTM) networks are adept at capturing
Vector Machines with their ability to handle high- sequential dependencies, and therefore are well placed to
dimensional spaces of features tend to perform better than capture context and narrative flow[38]. Deep learning
simpler models, especially after optimization using kernel models, however, typically require large labeled data and a
5181 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025
 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
tremendous amount of computing power for training. In this, relationships in addition to sequential context modeling.
transformer models such as BERT (Bidirectional Encoder Ensemble methods such as bagging and boosting (e.g.,
Representations from Transformers), RoBERTa, and GPT AdaBoost, XGBoost) have also been known to work well,
are the state-of-the-art models[39]. The models utilize the usually outperforming single classifiers since they can reduce
self-attention mechanism to capture the contextual relations variance and bias. Additionally, the incorporation of metadata
in the text more effectively than RNNs and CNNs. Fine- attributes (such as source of article, author credibility, and
tuning pre-trained transformer models for fake news data has social engagement measures) and user behavior analysis will
shown high accuracy and F1 values over traditional enhance the accuracy of models, particularly in real-world
classifiers as well as older deep learning models[40]. Though application where context becomes vital[42]. Work has
they work excellently, transformer models come with trade- demonstrated that combining content-based features with
offs. They are memory-intensive and can require GPU propagation-based signals is a better enhancement for fake
capability and enormous memory space. Their complexity news detection systems to improve their credibility. In brief,
also can reduce transparency, leading to explainability and while basic machine learning models offer speed and ease,
bias issues in high-stakes uses like news classification. transformer models and deep learning offer improved
Hybrid approaches and ensemble models combining several performance in high-level textual applications[43]. Based on
approaches have been especially helpful[41]. For example, the requirements of the application, whether scalability,
stacking a logistic regression on the outputs of an SVM and interpretability, available computing resources, or real-time
an LSTM can take advantage of both linear and non-linear execution, a model must be selected accordingl

Table1: Comprehensive Comparative Analysis of Machine Learning Models and Techniques for Fake News Detection

# Model/Technique Type Key Techniques Strengths Limitations

TF-IDF, Weak on non-linear

1. Logistic Regression (LR) Traditional ML Fast, interpretable
CountVectorizer patterns
Lightweight, good
Assumes feature
2. Naïve Bayes (NB) Traditional ML N-gram, TF-IDF for text
independence
classification
Handles high-
High computational
3. Support Vector Machine Traditional ML TF-IDF + Kernel dimensional data
cost
well
Easy to visualize
4. Decision Tree Traditional ML Gini/Entropy splits Prone to overfitting
and understand
Bagging, multiple Robust, reduces Slower, less
5. Random Forest Ensemble
trees overfitting interpretable
High accuracy,
Complex parameter
6. XGBoost Ensemble Gradient boosting regularization
tuning
built-in
Weak learners + Improves weak
7. AdaBoost Ensemble Sensitive to noisy data
boosting classifiers
K-Nearest Neighbors Distance metrics Simple, no training Inefficient with large
8. Traditional ML
(KNN) (Euclidean) phase datasets
Reduces Not great with non-
9. Ridge Classifier Regularized ML L2 Regularization
overfitting linear data
Fast updates, good
10. Passive Aggressive Class. Online ML Hinge loss-based Prone to instability
for large streams
Recurrent Neural Sequential data Good for sequence Vanishing gradient
11. Deep Learning
Network processing learning issue
Solves long
Slow training, many
12. LSTM Deep Learning Gated RNN units dependency issues
parameters
in text
Faster than LSTM,
Less expressive than
13. GRU Deep Learning Simplified LSTM performs
LSTM
comparably
Great for
Lacks sequential
14. CNN Deep Learning Convolutional layers extracting local
understanding
text features

5182 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025

 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
Deep contextual
Pre-trained Large memory &
15. BERT Transformer (DL) awareness, top-tier
transformer compute requirements
accuracy
Better fine-tuning,
16. RoBERTa Transformer (DL) Optimized BERT Very resource-heavy
robust
Contextual
Generative pre-
17. GPT Transformer (DL) generation + Lower interpretability
training
classification
Handles longer
Permutation
18. XLNet Transformer (DL) dependencies than Training complexity
Language Modeling
BERT
Flexible for
High training resource
19. T5 Transformer (DL) Text-to-text format classification and
demand
generation
Combines
Hard to interpret,
20. Hybrid (SVM + LSTM) Hybrid Stacking techniques strengths of both
complex structure
models
Local + sequential Enhanced feature Needs large training
21. Hybrid (CNN + LSTM) Hybrid
features extraction data
Ensemble Voting Majority voting Improved stability
22. Hybrid Model selection is key
Classifier among classifiers and generalization
Still under
Pretext tasks for Reduces need for
23. Self-Supervised Learning Emerging development for fake
feature learning labeled data
news
Reward-based Good for dynamic, Not widely adopted yet
24. Reinforcement Learning Emerging
feedback adaptive systems in fake news
Privacy-friendly, Communication
25. Federated Learning Privacy-Preserving Distributed training
scalable overhead
Adds emotional
Polarity scores, May miss sarcasm or
26. Sentiment Analysis NLP Technique dimension to
emotion classifiers subtleties
analysis
Authorship or
Writing style, POS
27. Stylometric Analysis NLP Technique style-based Style can be mimicked
patterns
detection
SHAP, LIME, Improves
Model Interpretation May slow down
28. Explainable AI (XAI) attention transparency &
Tool prediction process
visualization trust

6. CONCLUSION emotional and relational content of fake news. Despite the

Fake news detection remains a significant issue in today's massive progress, some limitations remain for developing
digital era, with the spread of misinformation affecting fully reliable fake news detection systems. Challenges such
societal beliefs, political discourse, and public decision- as adversarial manipulation, emerging misinformation
making. A number of machine learning and NLP-based tactics, and real-time detection require continued research.
approaches to the classification of news articles as real or fake Research in the future needs to prioritize the development of
have been presented in the paper, demonstrating that hybrid model adaptability, the inclusion of multi-modal data sources,
models that integrate deep learning and statistical methods and the advancement of explainable AI techniques to make
can significantly improve detection precision. Traditional automated detection systems more transparent and
machine learning models such as Logistic Regression, Naïve trustworthy. Overall, the combination of NLP and machine
Bayes, and Support Vector Machines have worked well to learning provides a solid foundation for tackling fake news,
classify deceptive content based on linguistic and statistical with ongoing innovation required to ensure scalability and
features. Deep learning techniques, in particular effectiveness in the real world. The findings of this research
Transformer-based architectures, have however enhanced contribute to the growing body of research committed to
classification accuracy by capturing complex contextual mitigating the impact of misinformation and shaping a more
dependencies within text. Feature engineering methods such informed digital society.
as TF-IDF, N-grams, and word embeddings have played a
significant role in preprocessing textual data for the machine REFERENCE
learning models. Sentiment analysis and network-based 1. C.K.H. and P.G.C. Deshpande, Fake News
approaches have also provided further insights on the Detection Using Deep Learning Techniques.

5183 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025

 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
International Conference on Advances in Processing, in Advances in Intelligent Networking
Information Technology, 2019. and Collaborative Systems. 2020. p. 223-234.
2. A Santhosh Kumar1, P.K., K Sharan Athithya1, V 15. Jain, A., et al., A smart System for Fake News
Sri Ajay Sundar1, Retraction: Fake News Detection Detection Using Machine Learning, in 2019
on Social Media Using Machine Learning (J. Phys.: International Conference on Issues and Challenges
Conf. Ser. 1916 012235). Journal of Physics: in Intelligent Computing Techniques (ICICT). 2019.
Conference Series, 2021. 1916(1). p. 1-4.
3. Abdulrahman, A. and M. Baykara, Fake News 16. Kaliyar, R.K., Fake News Detection Using A Deep
Detection Using Machine Learning and Deep Neural Network. International Conference on
Learning Algorithms, in 2020 International Computing Communication and Automation
Conference on Advanced Science and Engineering (ICCCA), 2018.
(ICOASE). 2020. p. 18-23. 17. Khaleel, Y.L., Fake News Detection Using Deep
4. Al Asaad, B. and M. Erascu, A Tool for Fake News Learning. International Conference on Advances in
Detection, in 2018 20th International Symposium on Information Technology, 2024.
Symbolic and Numeric Algorithms for Scientific 18. Khanam, Z., et al., Fake News Detection Using
Computing (SYNASC). 2018. p. 379-386. Machine Learning Approaches. IOP Conference
5. Al-Asadi, M.A. and S. Tasdemir, Using Artificial Series: Materials Science and Engineering, 2021.
Intelligence Against the Phenomenon of Fake News: 1099(1).
A Systematic Literature Review, in Combating Fake 19. Kong, S.H., et al., Fake News Detection using Deep
News with Computational Intelligence Techniques. Learning. International Conference on Advances in
2022. p. 39-54. Information Technology, 2020.
6. Chauhan, T. and H. Palivela, Optimization and 20. Koosha Sharifani 1, M.A., Yaser Akbari 3, Javad
improvement of fake news detection using deep Aghajanzadeh Godarzi 4, Operating Machine
learning approaches for societal benefit. Learning across Natural Language Processing
International Journal of Information Management Techniques for Improvement of Fabricated News
Data Insights, 2021. 1(2). Model. International Conference on Advances in
7. Costin BUSIOC, S.R., Mihai DASCALU, A Information Technology, 2022. 12(9).
Literature Review of NLP Approaches to Fake News 21. Madani, M., H. Motameni, and R. Roshani, Fake
Detection and Their Applicability to News Detection Using Feature Extraction, Natural
RomanianLanguage News Analysis. Language Processing, Curriculum Learning, and
TRANSILVANIA, 2020. Deep Learning. International Journal of Information
8. Doss, A.K.S.A.M.R., Smart Systems: Innovations in Technology & Decision Making, 2023. 23(03): p.
Computing. 2021. 1063-1098.
9. Fatima, B.N., Fake news detection using machine 22. Mahmud, T., et al., Integration of NLP and Deep
learning. IEEE Access, 2020. Learning for Automated Fake News Detection.
10. Faustini, P.H.A. and T.F. Covões, Fake news International Conference on Advances in
detection in multiple platforms and languages. Information Technology, 2024.
Expert Systems with Applications, 2020. 158. 23. Manzoor, S.I., J. Singla, and Nikita, Fake News
11. Federico Monti1, F.F., 2 Davide Eynard1,2 Michael Detection Using Machine Learning approaches: A
M. Bronstein1,2,3, Fake News Detection on Social systematic Review, in 2019 3rd International
Media using Geometric Deep Learning. IEEE Conference on Trends in Electronics and
Access, 2019. Informatics (ICOEI). 2019. p. 230-234.
12. Gerardo Ernesto Rolong Agudelo1, O.J.S.P., 2 and 24. Meesad, P., Thai Fake News Detection Based on
Julio Barón Velandia1, Raising a Model for Fake Information Retrieval, Natural Language
News Detection Using Machine Learning in Python. Processing and Machine Learning. SN Comput Sci,
International Conference on Advances in 2021. 2(6): p. 425.
Information Technology, 2019. 25. Mishra, S., et al., Analyzing Machine Learning
13. Haqi Al-Tai, M., B.M. Nema, and A. Al-Sherbaz, Enabled Fake News Detection Techniques for
Deep Learning for Fake News Detection: Literature Diversified Datasets. Wireless Communications and
Review. Al-Mustansiriyah Journal of Science, 2023. Mobile Computing, 2022. 2022: p. 1-18.
34(2): p. 70-81. 26. Mridha, M.F., et al., A Comprehensive Review on
14. Ibrishimova, M.D. and K.F. Li, A Machine Learning Fake News Detection With Deep Learning. IEEE
Approach to Fake News Detection Using Access, 2021. 9: p. 156151-156170.
Knowledge Verification and Natural Language
5184 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025
 “Fake News Detection Using Natural Language Processing (NLP) and Machine Learning”
27. Mugdha, S.B.S., S.M. Ferdous, and A. Fahmin, 36. Shaik, M.A., et al., Fake News Detection using NLP,
Evaluating Machine Learning Algorithms For in 2023 International Conference on Innovative
Bengali Fake News Detection, in 2020 23rd Data Communication Technologies and Application
International Conference on Computer and (ICIDCA). 2023. p. 399-405.
Information Technology (ICCIT). 2020. p. 1-6. 37. Shaikh, J. and R. Patil, Fake News Detection using
28. Nasir, J.A., O.S. Khan, and I. Varlamis, Fake news Machine Learning, in 2020 IEEE International
detection: A hybrid CNN-RNN based deep learning Symposium on Sustainable Energy, Signal
approach. International Journal of Information Processing and Cyber Security (iSSSC). 2020. p. 1-
Management Data Insights, 2021. 1(1). 5.
29. Pandey, S., et al., Fake News Detection from Online 38. Smitha, N. and R. Bharath, Performance
media using Machine learning Classifiers. Journal Comparison of Machine Learning Classifiers for
of Physics: Conference Series, 2022. 2161(1). Fake News Detection, in 2020 Second International
30. Prachi, N.N., et al., Detection of Fake News Using Conference on Inventive Research in Computing
Machine Learning and Natural Language Applications (ICIRCA). 2020. p. 696-700.
Processing Algorithms. Journal of Advances in 39. Srivastava1, A., Real Time Fake News Detection
Information Technology, 2022. 13(6). Using Machine Learning and NLP. International
31. Ray Oshikawa1, J.Q., WilliamYang Wang2, A Research Journal of Engineering and Technology
Survey on Natural Language Processing for Fake (IRJET), 2020. 7(6).
News Detection. IEEE Access, 2020. 40. Subhadra Gurav1, S.S., Supriya Shinde3, Prachi
32. Rodríguez, Á.I. and L.L. Iglesias, FAKE NEWS Wabale4, Sumit Hirve5, Survey on Automated
DETECTION USING DEEP LEARNING. IEEE System for Fake News Detection using NLP &
Access, 2019. Machine Learning Approach. International
33. S, D. and B. Chitturi, Deep neural approach to Research Journal of Engineering and Technology
Fake-News identification. Procedia Computer (IRJET), 2019. 6(1).
Science, 2020. 167: p. 2236-2243. 41. Thota, A., et al., Fake News Detection: A Deep
34. Saeed Amer Alameri, M.M., COMPARISON OF Learning Approach. SMU Data Science Review,
FAKE NEWS DETECTION USING MACHINE 2018. 3(1).
LEARNING AND DEEP LEARNING 42. Uma Sharma, S.S., Shankar M. Patil, Fake News
TECHNIQUES. International Conference on Detection using Machine Learning Algorithms.
Advances in Information Technology, 2020. International Journal of Engineering Research &
35. Sajjad Ahmed, K.H., Flavio Corradini, Development Technology (IJERT). 9(3).
of Fake News Model Using Machine Learning 43. Zhixuan Zhou1, Huankang Guan1, Meghana
through Natural Language Processing. Moorthy Bhat2 and Justin Hsu2, Fake News
International Conference on Advances in Detection via NLP is Vulnerable to Adversarial
Information Technology, 2020. Attacks. International Conference on Advances in
Information Technology, 2020.

5185 Hisham Ahmed Mahmoud 1, ETJ Volume 10 Issue 05 May 2025