Author: Kalev Leetaru
How Television News Story Segmentation Could Be Used To Help Citizens Connect With Their Local Governments
Local municipalities all across the United States hold a range of meetings, debates, hearings, press conferences, public comment opportunities and…
Prosperity and Posturing: New research uncovers domestic drivers of international aggression
"New research by Dr Ashani Amarasinghe, from the University of Sydney School of Economics, sheds light on how governments use…
Story Segmentation Of TV News: Testing On Evening News Broadcasts From A Decade & A Half Ago
Our new television news story segmentation workflow has demonstrated remarkably accurate results in our tests thus far, matching and even…
Using Gemini 2.5's Structured Outputs To Enforce Consistent Stable JSON Output For Story Segmentation
One key challenge with our new Gemini 2.5-based television news story segmentation initiative has lain in coaxing Gemini to produce…
Story Segmentation Of TV News: Segmenting A Week Of Evening TV News Into Stories & Adding Links To Each Clip
Earlier this week we showcased how accurately Gemini 2.5 Flash Thinking was able to segment American television evening news broadcasts…
Story Segmentation Of TV News: A First Glimpse At Cataloging The Major Congressional Debates From CSPAN
Two years ago we explored how LLMs could help America's citizenry better understand the functioning of their government by distilling…
Story Segmentation Of TV News: Experimenting With A Broader Array Of Show Types
Given Gemini's success at segmenting evening news broadcasts into their underlying stories, how well does it perform on a broader…
Snopes: Fact Check: Yes, video shows Trump stumbling up steps to Air Force One
Snopes uses the TV News Archive to verify a video of President Trump stumbling on the Air Force One stairs….
Story Segmentation Of TV News: Experimenting With Persian & Russian Broadcasts & English Translations
Continuing our series segmenting television news broadcasts into their underlying stories, yesterday we demonstrated that Gemini yields human-level segmentation accuracy…
Story Segmentation Of TV News: Comparing A Week Of Evening News Human Segmentation With Gemini Pro, Flash Thinking And Flash Non-Thinking
Last week we demonstrated a major leap in our ability to automatically segment television evening news broadcasts into their underlying…
Using Gemini 2.5 Pro To Assess Evidence Sourcing & Credibility In News Reporting On Conflict
One of the most critical pieces of information that must be asked about any conflicting reporting is the sourcing of…
Mass Translating Global Television News: Comparing Gemini Pro/Flash/2.5/2.0, Gemma & Google Translate
The Television News Archive spans more than 16.7 billion spoken characters of non-English news broadcasts from across the world spanning…
Video: Planetary Scale AI: Watching, Visualizing And Forecasting The Entire World In Realtime
The video of my talk at the Americas Google Developer Experts (GDE) Summit following Google I/O 2025 is now available!…
Story Segmentation Of TV News: Comparing Gemini 2.5 Pro/Flash Thinking/Flash & Gemini 2.0 Flash/Flash-Light
Yesterday we demonstrated a major leap forward in our efforts to apply at-scale story segmentation to television news: changing from…
Story Segmentation Of TV News: Having Gemini Chunk A Transcript Instead Of Following Timecodes
Unlike written news articles, which often focus on a single story, television news transcripts represent continuous stream-of-consciousness streams of words…
A First Glimpse Of Using Language Detection To Identify Language Examples From The TV News Archive
A year ago we explored applying language detection to selections from the Chirp-generated ASR transcripts of the TV News Archive,…
Story Segmentation Of TV News: Comparing Gemini 2.5 Pro/Flash, Gemini 2.0 Flash/Lite, Gemma 3 1/4/12/27B & ChatGPT 4o/o3
As part of our continuing series in using LLMs to segment television news broadcasts into their component stories, let's return…
Interpretable Machine Learning for Macro Alpha: A News Sentiment Case Study
This study introduces an interpretable machine learning (ML) framework to extract macroeconomic alpha from global news sentiment. We process the…
Planetary Scale AI: Watching, Visualizing And Forecasting The Entire World In Realtime
It was a tremendous honor to speak at the Americas Google Developer Experts (GDE) Summit last week to my fellow…
Using Gemini 2.5 As A Persona-Based News Recommender Service: Summarize Trump's Middle East Visit
Continuing our series in persona-based news recommenders, today we'll take our end-to-end recommender service workflow and ask Gemini 2.5 Pro to…
Using Gemini 2.5 As A Persona-Based News Recommender Service: Analyze A Day Of Global AI Trends
Continuing our series in persona-based news recommenders, today we'll take our end-to-end recommender service workflow and ask Gemini 2.5 Pro…
Using Gemini 2.5 As A Persona-Based News Recommender Service: An Investment Advisor On How To Make Money
Continuing our series in persona-based news recommenders, we're going to follow the same workflow of our end-to-end recommender service, but this time…
Behind The Scenes: XML Tags In TV News Closed Captioning Streams
Traditional broadcast closed captioning is both a powerful human-generated realtime transcript of television news and a source of uncountable oddities…
Automatic large-scale political bias detection of news outlets
Political bias is an inescapable characteristic in news and media reporting, and understanding what political biases people are exposed to…