• Bias / propaganda detection: transformer-based classifiers fine-tuned on public political news datasets, plus some hand-engineered features (e.g., source-level priors, readability, sentiment). In offline tests I get 93% accuracy on bias detection(happy to share more detail if people care).
• Claim extraction: sentence segmentation + a lightweight classifier to label check-worthy clauses (counts, quotes, time-bound events, entity claims).
• Fact-checking: MNLI model (currently DeBERTa-based) over (claim, evidence-passage) pairs with some heuristics to merge multiple snippets.
• Frontend: Angular + server-rendered news pages for speed and SEO.
The methodology is documented here with more detail:
• How far I can push MNLI-style models before needing a more explicit retrieval-augmented system or custom architectures.
• Whether my current claim extraction approach is good enough for high-stakes use, or if I should move to a more formal information extraction pipeline.
• How to expose uncertainty and failure modes in a way that’s actually useful for non-technical readers.
Why I’m posting
I’d like feedback from this community on:
• ML / NLP choices you strongly disagree with.
• Evaluation: what would be a more convincing test suite or benchmark?
• UI/UX for showing “supported/refuted/inconclusive” without overselling model confidence.
I’m very open to critique. If you think this is conceptually wrong or socially dangerous, I’d also like to hear that argument.
Thanks for reading,
Marcell
Show HN: Neutral News AI – Multi-source, MNLI-checked news summaries · HackerTrans
I’m the co-founder of Neutral News AI: a site that tries to answer a simple question:
“What actually happened here, across multiple biased sources, and can we check the claims against the original articles?”
Link: https://neutralnewsai.com Analyzer: https://neutralnewsai.com/analyzer No signup needed to read the news or run a basic analysis.
What it does
• Crawls multiple outlets (left / center / right + wires / gov sites) for the same story.
• Generates a short, neutral summary constrained to those sources (no extra web search).
• Extracts atomic claims (events, numbers, quotes) from the draft.
• Uses an MNLI model to test each claim against the underlying articles:
• entailment → “Supported”
• contradiction → “Refuted”
• neutral → “Inconclusive”
• Surfaces a “receipt ledger” per article: claim text, verdict, quote, source, timestamp.
• Exposes the underlying models on an Analyzer page where you can paste any URL and get:
• political bias score,
• sentiment / subjectivity,
• readability metrics,
• a rough credibility signal.
Stack and models
• Backend: Python, PostgreSQL.
• Crawling / aggregation: scheduled scrapers + RSS + manual curated source lists.
• Bias / propaganda detection: transformer-based classifiers fine-tuned on public political news datasets, plus some hand-engineered features (e.g., source-level priors, readability, sentiment). In offline tests I get 93% accuracy on bias detection(happy to share more detail if people care).
• Claim extraction: sentence segmentation + a lightweight classifier to label check-worthy clauses (counts, quotes, time-bound events, entity claims).
• Fact-checking: MNLI model (currently DeBERTa-based) over (claim, evidence-passage) pairs with some heuristics to merge multiple snippets.
• Frontend: Angular + server-rendered news pages for speed and SEO.
The methodology is documented here with more detail:
https://neutralnewsai.com/methodology
What I’m unsure about
• How far I can push MNLI-style models before needing a more explicit retrieval-augmented system or custom architectures.
• Whether my current claim extraction approach is good enough for high-stakes use, or if I should move to a more formal information extraction pipeline.
• How to expose uncertainty and failure modes in a way that’s actually useful for non-technical readers.
Why I’m posting
I’d like feedback from this community on:
• ML / NLP choices you strongly disagree with.
• Evaluation: what would be a more convincing test suite or benchmark?
• UI/UX for showing “supported/refuted/inconclusive” without overselling model confidence.
I’m very open to critique. If you think this is conceptually wrong or socially dangerous, I’d also like to hear that argument.
Thanks for reading, Marcell