International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 01 | Jan 2026 www.irjet.net p-ISSN: 2395-0072
Fake News Detection
Dr.C.P.Divate1, Ms.N.R.Bhokre2, Saksham S. Shrivastav3, Vikram R. Sargar4, Adarsh K. Yevale 5 , Priyanka V. Chougule 6, Onkar M. Shinde 7
1Dean, Dept of Computer Engineering, Shri Ambabai Talim Sanstha’s Sanjay Bhokare Group of Institute Miraj(poly), Maharashtra, India
2Lecturer, Dept of Computer Engineering, Shri Ambabai Talim Sanstha’s Sanjay Bhokare Group of Institute Miraj(poly), Maharashtra, India
3,4,5,6,7 Student Dept. of Computer Engineering, Shri Ambabai Talim Sanstha’s Sanjay Bhokare Group of Institute Miraj(poly), Maharashtra, India ***
Abstract - The rapid growth of social media and online news platforms has led to an unprecedented spread of misinformation and fake news, which can influence public opinion, create social unrest, and undermine trust in digital media. Manual verification of online content is timeconsuming and impractical at a large scale. This paper presents a fake news detection system based on Natural Language Processing (NLP) and rule-based analysis techniques,withouttheuseofmachinelearningmodels.
Key Words: fake news detection; natural language processing; rule-based system; text analysis; misinformation.
1. INTRODUCTION
The consumption of news has shifted significantly from traditional printandbroadcastmediatodigital platforms, includingonlinenewsportalsandsocialnetworkingsites. Thistransformationhasgreatlyenhancedtheaccessibility, immediacy, and reach of information, allowing users to receivereal-timeupdatesfromacrosstheglobe.However, thissameeaseofdistributionhasalsofacilitatedtherapid spread of fake, misleading, or deliberately manipulated news.Unliketraditionalmedia,onlineplatformsoftenlack rigorouseditorialoversight,enablingunverifiedcontentto circulatewidelywithinminutes.
Numerousstudieshaveshownthatfalseinformationtends to spread faster and reach a wider audience than verified news, as it often appeals to emotions, sensationalism, or confirmationbias.
This issue becomes particularly critical during sensitive events such as elections, public health emergencies, pandemics, and large-scale social movements, where misinformationcaninfluencepublicopinion,createpanic, underminetrustininstitutions,andevenleadtoreal-world consequences.Asaresult,thedetectionandpreventionof fake news have become essential challenges in the digital age, highlighting the need for automated, reliable, and scalablesolutions..
2. PROBLEM STATEMENT
The rapid expansion of digital news platforms and social mediahassignificantlytransformedthewayinformationis produced,shared,andconsumed.Whilethisdigitalshifthas made news easily accessible and enabled real-time information dissemination, it has also resulted in the widespreadcirculationoffake,misleading,anddeliberately manipulated news content. Such misinformation can influencepublicopinion,distortfacts,createsocialunrest, and generate panic among the public, especially during criticaleventssuchaselections,naturaldisasters,andpublic healthemergencies.
Due to the massive volume of content generated every minute on online platforms, manual verification of news authenticityhasbecomeextremelytime-consuming,laborintensive, and impractical. Human fact-checkers cannot efficientlykeepupwiththespeedandscaleatwhichfalse information spreads across digital platforms. As a result, fake news often reaches a large audience before it can be identifiedandcorrected.
Manyexistingfakenewsdetectionsystemsrelyprimarilyon keyword-basedfilteringorshallowtextanalysistechniques. Thesetraditionalapproachesfocusonsurface-levelpatterns rather than understanding the contextual and semantic meaning of news articles. Consequently, they often fail to detect sophisticated fake news that uses subtle language, emotional manipulation, or partially true information, leadingtoinaccurateorunreliablepredictions.
Furthermore,mostexistingsystemssimplyclassifynewsas eitherrealorfakewithoutprovidingaconfidencescoreor explanationforthedecision.Thislackoftransparencymakes itdifficultforuserstoassessthecredibilityoftheoutputor understand how reliable the classification is. Therefore, thereisacriticalneedforanintelligent,accurate,anduserfriendlyfakenewsdetectionsystemthatleveragesadvanced natural language processing and machine learning techniquestoevaluatenewscredibilityeffectivelywhilealso providingmeaningfulconfidencelevelstousers.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 01 | Jan 2026 www.irjet.net p-ISSN: 2395-0072
3. PROPOSED SOLUTION
The proposed solution involves the development of an automatedfakenewsdetectionsystemthatutilizesmachine learning techniques to accurately classify news articles as eitherrealorfake.Theprimaryobjectiveofthesystemisto reducehumaninvolvementwhileachievinghighaccuracy, scalability, and adaptability to the continuously evolving patterns of misinformation present in digital media. By analysingtextualcontentalongwithlinguisticandstatistical features of news articles, the system is capable of distinguishing genuine information from misleading or fabricatedcontent.
Thesystembeginswiththecollectionofdatafromreliable and publicly available datasets that contain labelled news articlescategorizedasreal orfake.Toensuredata quality and consistency, the collected data undergoes a preprocessingphase.Thisphaseincludestextnormalization, removal of stop words, elimination of punctuation and special characters, tokenization, and stemming or lemmatization.Thesepreprocessingstepshelpreducenoise in the data and enhance the effectiveness of the machine learningmodels.
Afterpreprocessing,relevantfeaturesareextractedfromthe textualdatausingtechniquessuchasBagofWords(BoW), Term Frequency–Inverse Document Frequency (TF-IDF), and word embeddings. Feature extraction transforms unstructuredtextintonumericalrepresentationsthatcanbe efficientlyprocessedbymachinelearningalgorithms.
Thesystememploysmultiplesupervisedmachinelearning classifiers, including Naïve Bayes, Logistic Regression, SupportVectorMachines(SVM),andRandomForest.These algorithms are trained using the extracted features to identify patterns commonly associated with fake and real news.Acomparativeanalysisoftheclassifiersisperformed todeterminethemostaccurateandreliablemodel.
Oncetrained,theselectedmodelisevaluatedusingstandard performancemetricssuchasaccuracy,precision,recall,and F1-scoretoensurereliabilityandeffectiveness.Thefinalized modelisthenintegratedintoauser-friendlyapplicationthat allowsuserstoinputanewsarticleorURLandreceiverealtimeauthenticitypredictions.
To maintain robustness and long-term effectiveness, the system is designed to be scalable and continuously updatable with new datasets. This enables the model to adapt to emerging misinformation trends and evolving languagepatterns.Overall,theproposedsolutionoffersan efficient, cost-effective, and automated approach to addressing the growing challenge of fake news in digital media.
4. LITERATURE REVIEW
1. Wang (2017) highlightedthattextualcuessuchasword choice,sentencecomplexity,andemotionaltonecanserveas indicators of misinformation, even without advanced learningmodels.Thisworkdemonstratedtheimportanceof linguisticfeaturesinidentifyingdeceptivecontent.
2. Shu et al. (2020) emphasized the role of source credibility and contextual information in fake news detection.Theirstudiesshowedthatunreliablesourcestend to publish content with similar linguistic and stylistic patterns,makingrule-basedsourceverificationaneffective approach.
Otherstudies explored sentimentanalysistechniques and foundthatfakenewsarticlesoftenexhibitextremepositive ornegativesentimentcomparedtofactualreporting.Rulebasedsystemscombiningsentimentscores,keywordlists, and writing-style heuristics were shown to provide transparentandexplainabledetectionmechanisms.
Surveypapersconcludethatwhilerule-basedsystemsmay lack adaptability, they offer advantages such as interpretability, low computational cost, and ease of implementation, making them suitable for educational, prototype,andlow-resourceenvironments.
Research Objectives:
1. To implement Logistic Regression for efficient andaccurateclassificationofnewscontent.
2. To design and develop a machine learning–based modelfordetectingfakeandrealnewsautomatically.
3. Toevaluatetheperformanceoftheproposedmodel usingstandardaccuracyandclassificationmetrics.
4. To use Sentence-BERT (SBERT) for extracting meaningfulsemanticfeaturesfromnewstext.
5. METHODOLOGY
The proposed Fake News Detection System follows a structured NLP-based pipeline without using machine learningmodels.
Initially,newsarticlesarecollectedfromonlinesourcesor user input. The text data undergoes preprocessing steps includingconversiontolowercase,removalofpunctuation, stop-wordelimination,andtokenizationtoreducenoiseand standardizeinput.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 01 | Jan 2026 www.irjet.net p-ISSN: 2395-0072
After preprocessing, multiple rule-based analyses are applied:
Keyword Analysis: Detection of commonly used fake-newstermsandsensationalphrases
Sentiment Analysis: Identification of extreme emotionalpolarityusinglexicon-basedmethods
Writing Style Analysis: Examination of headline length, punctuation usage, capitalization, and repetition
Source Credibility Check: Verification against predefinedlistsoftrustedanduntrusteddomains Eachanalysiscontributestoacumulativecredibility score.Basedonpredefinedthresholds,thesystem classifiesthenewsarticleas real or potentiallyfake A simple user interface can be developed using Python-based tools to allow users to input news textandreceiveinstantresults.
6. Benefits of an NLP-Based Fake News Detection System
[1] A. User-Centric Benefits
AnNLP-basedfakenewsdetectionsystemofferssignificant advantages to end users by empowering them to make informeddecisionswhileconsumingdigitalcontent.Oneof thekeybenefitsisaccurateinformationverification,which allows users to quickly assess whether a news article is potentiallyfakeorreliable.Thisimprovesuserawareness and encourages critical thinking when interacting with onlineinformation.
Thesystemalsohelpsreduceexposuretomisinformationby identifying suspicious content at an early stage. By discouragingusersfromsharingmisleadingorfalsenews,it plays an important role in limiting the spread of misinformationacrosssocialmediaplatforms. Additionally,thesystemisdesignedwitheaseofuseinmind. A simple, text-based interface ensures that even nontechnical users can easily input news content and receive clear, understandable results, making the system widely accessible.
[2] B. Societal and Media Benefits
From a societal perspective, the system contributes to improving media credibility by supporting journalists, editors,andfact-checkersinpreliminarycontentscreening. Automated analysis helps identify potentially misleading articlesbeforetheyarepublishedorwidelycirculated.
Theabilitytoperformreal-timeornearreal-timeanalysis enablesafasterresponsetomisinformation,whichiscritical inpreventingfakenewsfromgoingviral.Thisisparticularly valuable during sensitive events such as elections, public healthcrises,andsocialmovements.
Furthermore, the system supports digital literacy by encouraging responsible content consumption and ethical sharing practices. By making users more aware of misinformationpatterns,itpromotesahealthierandmore trustworthydigitalinformationecosystem.
[3] C. Technical Benefits
From a technical standpoint, the NLP-based rule-driven approach offers several advantages. The system operates withlowcomputationalcost,asrule-basedNLPtechniques requireminimalprocessingpowerandcanrunefficientlyon standardhardwarewithouttheneedforexpensiveGPUsor largetrainingdatasets.
Thescalabledesignallowsthesystemtobeeasilyintegrated into websites, browser extensions, news portals, or social