Data Scraping & AggregationML Infra, Classifiers & RL
A Risk-Stratification Framework Linking Clinical Milestones to PAH Survival Outcomes
Aug 21, 20266 min readRead
Back
Data Scraping & Aggregation
A Verified, Multi-Source News Platform With AI-Anchor Narration and Hourly Refresh
How Pfactorial Technologies built Tru-Insight, a news aggregation app that verifies story authenticity with a machine-learning classifier and narrates headlines through an AI anchor.
August 21, 2026
Share
ENGAGEMENT SNAPSHOT
Figure 1 - Key figures from this engagement, at a glance.
EXECUTIVE SUMMARY
Our client wanted to launch a news product that solved the fragmentation of having to visit many different sites and apps to get trustworthy, current news across local, national, business, and world categories.
Simply aggregating headlines from many sources without a way to check authenticity would have amplified the exact trust problem the product set out to solve, and a feed that updates in batches or once a day breaks the "most recent news" promise the product is built on.
Pfactorial built Tru-Insight: an app that scrapes and aggregates from many trusted sources, classifies each story's authenticity with a machine-learning model, refreshes every category hourly, and narrates verified headlines through an AI-generated anchor, alongside a date-range API exposing the same verified content.
Why this engagement is representative This engagement demonstrates Pfactorial's ability to combine aggregation, an ML-based verification layer, and generative-media presentation into one coherent consumer product, not three disconnected components.
THE CHALLENGE
Building a news product that could be trusted and stayed current meant solving problems a simple aggregator doesn't have to.
1. Aggregating news doesn't solve trust, it can amplify the problem
Pulling from many sources without a way to separate credible reporting from misinformation just puts more unverified content in front of the reader, faster.
2. News has to actually stay current, not just look aggregated
A feed that updates in batches or once a day breaks the promise of showing users the most recent news, so refresh had to be a scheduled guarantee across every category.
3. Coverage spans national, business, and hyperlocal news at once
The same platform needed to serve broad national coverage across many countries and granular, state-level local news for India, which are different data problems solved separately.
4. Presentation was part of the differentiation, not an afterthought
The product needed an AI-narrated anchor experience, not just a scrollable list of headlines, which meant a generative video and speech pipeline alongside the news pipeline.
The real brief Not "another headline aggregator" but "one app where every story has been checked for authenticity, is never more than an hour old, and can be delivered as narrated video, not just text."
THE SOLUTION
Pfactorial built Tru-Insight around four cooperating stages - collection, authenticity classification, serving, and AI-anchor narration - running on a scheduled hourly cycle.
Figure 1 - Every story is verified before it's served, and narration runs against the same verified content, not a separate feed.
Architectural principles
- Verify before you publish, not after - Every collected story runs through the authenticity classifier before reaching a user, so verification is a gate, not a downstream flag.
- Freshness as a scheduled guarantee, not a best effort - An hourly refresh cycle across every source category keeps the "most recent news" promise structural rather than aspirational.
- National breadth and local depth as separate coverage tracks - Broad country-level aggregation and granular Indian state-level local news are treated as distinct coverage goals, so neither dilutes the other.
- Presentation is a first-class output, not a rendering afterthought - The AI-anchor narration pipeline is built as its own stage, on par with collection and classification, not a UI skin applied at the end.
CAPABILITIES DELIVERED
Each capability targets a distinct part of turning scattered news sources into one trusted, current product.
CAPABILITY | WHAT IT DOES |
|---|---|
Multi-source news aggregation | Scrapes and consolidates news from many trusted sources into one feed, ending the need to visit multiple sites. |
ML-based authenticity classification | A Random Forest classifier flags likely misinformation before a story reaches a reader. |
Hourly refresh | Every category re-pulls on an hourly cycle, so users see current news rather than stale batches. |
Broad geographic coverage | National headlines for 50+ countries plus hyperlocal coverage for 28 Indian states, with fallback top headlines for any other location a user requests. |
AI-anchor narration | A generative video and text-to-speech pipeline produces a synthetic news anchor that reads out headlines and updates. |
Date-range news API | External users can request news by date or date range, up to one year back, by category. |
Figure 2 - The same verification and refresh guarantees apply whether a user reads the feed or watches the AI anchor.
Design note Choosing a classical Random Forest classifier for authenticity scoring over a heavier deep-learning model was a deliberate trade - fast enough to score every incoming story within the hourly refresh window at full coverage volume, at the cost of some of the nuance a larger model might catch, a defensible choice for a product whose core differentiator is coverage and freshness alongside verification.
ENGINEERING FOR SCALE AND RELIABILITY
Four decisions kept collection, verification, and narration operating together on an hourly cycle without one stage bottlenecking the others.
Scraping via requests and Playwright instead of relying solely on publisher APIs
Direct scraping gave broader source coverage than would be available through a patchwork of individual publisher APIs and license terms.
A Random Forest classifier for authenticity scoring instead of a heavier model
Fast enough to score every incoming story within the hourly refresh window across all covered sources and regions.
Wav2Lip and text-to-speech for narration instead of licensing a presenter service
Building the AI-anchor pipeline directly kept narration in-house and tied to the same content pipeline that had just been verified.
A dedicated date-range API instead of exposing only the live feed
External integrators can query up to a year of historical, verified news by category, not just what's currently on the home screen.
DELIVERY APPROACH
The build proceeded from source aggregation through to verification, coverage expansion, and finally the narration layer.
1. Source aggregation pipeline - building the scraping layer to pull national, business, and local stories from multiple trusted sources.
2. Authenticity classifier - training and integrating the Random Forest model that scores every collected story before publication.
3. Hourly refresh scheduling - wiring the pipeline to re-run on an hourly cycle across every covered category and region.
4. Geographic coverage expansion - extending coverage to 50+ countries at the national level and 28 Indian states at the local level, with a fallback path for other locations.
5. AI-anchor narration - integrating a video and text-to-speech pipeline to generate narrated presentation of verified headlines.
6. News API - exposing a date and date-range endpoint returning up to a year of verified, categorized news.
RESULTS AND IMPACT
- Key outcomes from this engagement.
Tru-Insight aggregates and verifies news across 50+ countries and 28 Indian states, refreshing every category hourly and running each collected story through a Random Forest authenticity classifier before it reaches a user.
Because narration runs as its own pipeline stage against the same verified content, users can consume the same hourly-refreshed, authenticity-checked news either as a normal feed or as AI-anchor-narrated video, and external integrators can pull up to a year of the same verified history through the date-range API.
What it enabled commercially
The client can market authenticity-checked, always-current news with an AI-narrated presentation as a clear differentiator against generic aggregator apps, and offer a licensable date-range news API as a second, integration-facing revenue path.
WHY PFACTORIAL
This engagement reflects Pfactorial's ability to combine content aggregation, an ML verification layer, and a generative-media presentation pipeline into one coherent consumer product, not three disconnected components.
- Service lines this engagement draws on.
Engagement enquiries Pfactorial Technologies works with organisations that need trustworthy, always-current content products with a real verification layer, not just aggregation. If you're evaluating a media or news-technology platform, we're happy to give you an honest read on scope and risk before anyone commits to anything. · pfactorial.ai
APPENDIX A - TECHNOLOGY STACK
The technology stack underpinning the system, grouped by the layer it serves.
Result and Analysis
ENGAGEMENT SNAPSHOT
How Pfactorial Technologies built Tru-Insight, a news aggregation app that verifies story authenticity with a machine-learning classifier and narrates headlines through an AI anchor.
CASE STUDIES
You might also like...
ML Infra, Classifiers & RL
A BERT-Driven Restaurant Chatbot That Classifies Guest Intent, Not Keywords
Aug 21, 20266 min readRead
Automotive & Vehicle
A Browser-Native Neural Network Simulation That Learns to Drive From Experience
Aug 21, 20267 min readRead
Computer VisionML Infra, Classifiers & RL
A Deep Learning Classifier That Separates True-Negative Mammograms from Genuine BIRADS 0 Cases
Aug 21, 20266 min readRead
Conversational AI & ChatbotsML Infra, Classifiers & RL
A Layered Safety Pipeline for a Healthcare Patient Companion
Aug 21, 20268 min readRead
ML Infra, Classifiers & RL
A Machine-Learning Screening Model for Pulmonary Hypertension from Routine Medical Records
Aug 21, 20266 min readRead