Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
TruthScanML is presented as a two-stage fake-news detector: an offline text classifier first assesses an article, then online evidence gathering and scoring adds context. Its author, Harsh Tiwari, describes a stack built with TF-IDF and Logistic Regression, plus credibility and freshness scoring, natural-language-inference-assisted verification, and an INCONCLUSIVE result for uncertain cases. The published project description does not report accuracy or enough implementation detail to reproduce or independently assess the system.
How does TruthScanML work?
The project description presents TruthScanML as a hybrid workflow, rather than a single model that settles a claim from its wording alone. Its first stage classifies text offline; a second stage gathers online evidence and scores it. The author describes this latter stage as including source credibility, freshness, and NLI-assisted verification. These are the author’s stated components, not independently validated capabilities or reported outcomes. Read Harsh Tiwari’s project article on DEV Community.
Offline text classification
The author says the offline classifier uses TF-IDF with Logistic Regression. TF-IDF represents text through the importance of its terms, and Logistic Regression uses those features to produce a classification. The project description does not specify the training data, label definitions, preprocessing, model settings, or how the output is calibrated. Those details are needed to reproduce the classifier or judge how its output should be interpreted.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Online evidence and verification
The described online layer is intended to supplement the text-based classification with evidence from multiple sources. The project article names credibility and freshness scoring and NLI-assisted verification, but does not identify the sources, explain the scoring formula, name the NLI model, or describe what happens when sources disagree. It therefore establishes the broad design, not how reliably that design verifies a particular claim.
#1 Best Overall
- Use scikit-learn to track an example ML project end to end
- Explore several models, including support vector machines, decision trees, random forests, and ensemble methods
- Exploit unsupervised learning techniques such as dimensionality reduction, clustering, and anomaly detection
- Dive into neural net architectures, including convolutional nets, recurrent nets, generative adversarial networks, autoencoders, diffusion models, and transformers
- Use TensorFlow and Keras to build and train neural nets for computer vision, natural language processing, generative models, and deep reinforcement learning
When the result is inconclusive
The author says TruthScanML can return INCONCLUSIVE when uncertain. That is a useful distinction from forcing every item into a true-or-false result, but the article does not explain the confidence threshold or decision procedure that triggers it.
What the project description does—and does not—establish
The project article names Python, FastAPI, Streamlit, and scikit-learn as its implementation stack. It does not provide the information needed to evaluate the detector’s accuracy or reproduce its reported behavior.
Rank #2
- Training and labels: No dataset, label definitions, or data-splitting method is stated.
- Evidence sources: The number and origins of sources are not stated.
- Scoring and model choices: The credibility and freshness calculations and the NLI model are not identified.
- Evaluation: No performance measurements or evaluation design are reported.
Without these details, it would be misleading to attach a performance figure to TruthScanML or infer how it handles unfamiliar topics, misleading framing, or conflicting evidence.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Why detection results can be hard to generalize
A 2026 comparison by Pietro Dell’Oglio, Alessandro Bondielli, Francesco Marcelloni, and Lucia C. Passaro evaluated 12 representative approaches across 10 datasets. Its protocol covered English text-only binary classification and compared in-domain, multi-domain, and cross-domain settings. The authors report that fine-tuned models can perform well in-domain yet struggle to generalize; cross-domain approaches can narrow that gap, but require more data. These findings describe the study’s evaluated approaches, not TruthScanML. The paper also notes that mapping different datasets onto “Real” and “Fake” labels can erase semantic nuance. Read the 2026 comparison in Information Sciences.
A review of fake-news detection studies from 2018–2023 identifies additional methodological concerns: class imbalance can bias results toward a majority class, models can overfit or underfit, and TF-IDF or n-gram features may miss semantic relationships. These are general risks to check for—not demonstrated flaws in TruthScanML. Read the review on PubMed Central.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What to look for in a meaningful evaluation
To judge a detector like TruthScanML, readers would need evidence beyond a successful demonstration. Useful evaluation details include:
Rank #4
- Which datasets and claim types were used, how “real” and “fake” were defined, and whether class imbalance was addressed.
- Whether training and test examples were separated by source, topic, and time, not only randomly split, and whether results include cross-domain testing.
- How the online evidence sources were selected, how freshness and credibility are measured, and how contradictory or missing evidence affects the verdict.
- Results for each class and for inconclusive cases, along with the decision threshold and examples of failure.
These checks connect the project’s stated design to the broader evaluation issues identified in the cited studies. They are evaluation criteria, not claims that the project has completed or failed those tests.
Recommended Free Tools
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
