Category: Uncategorized
Large-Token LLMs: Using GCP's New 32K PaLM Model To Summarize Two CSPAN Broadcasts
Earlier this month we explored using Google's new 32K PaLM LLM model to summarize an entire evening news broadcast. Let's…
Hallucinating Detail In Simple Summaries: Why LLM "Grounding" Doesn't Work To Combat Hallucination
One of the most commonly recommended methods of reducing hallucination in LLMs is called "grounding" in which the LLM is…
Deep Linking Democracy: An Index Of Half A Million Legislative Mentions On CSPAN Spanning 2009-2023
Today in collaboration with the Internet Archive's TV News Archive, we are immensely excited to announce a massive new index…
Visual Explorer: Live Transcription Of All 24 Active Uncaptioned Channels Through Google's Chirp
Last week we announced a massive historical backfile of 9.5 million minutes of machine-generated multilingual transcripts of global television news…
Experiments In Summarizing Global Media Tenor: Views Towards China – Part 2
Continuing our experiments with using LLMs to summarize global media coverage, let's use the final workflow that yielded the best…
Experiments In Summarizing Global Media Tenor: Views Towards China – Part 1
How might we use LLMs to summarize at-scale media tone and portrayals of countries? Let's look at a few different…
Tracing Global Media Tone & Anxiety Towards China & Ukraine 2020-Present
Using the GKG, what might we be able to learn about global media tone towards China and Ukraine? Let's first…
Visual Explorer: Belarusian, Russian, Ukrainian & Iranian Channels Now Upgraded to Google's Chirp Transcription
We are incredibly excited to announce that over the coming 24 hours, all of our Belarusian, Russian, Ukrainian & Iranian…
Visual Explorer: Vastly Improved Russia Today Transcripts Through GCP's Chirp Model
To help researchers and journalists examining Russian propaganda and narratives around the invasion of Ukraine, we are launching vastly improved…
Visual Explorer: Transcripts Now Available For Iran's PressTV
We are excited to announce today the availability of machine-generated transcripts for Iran's English-language PressTV channel, generated through Google's Chirp…
Watch: Russia 24's Tucker Carlson Show Release Teaser
Russian television news channel Russia 24 has aired a trailer for a weekend show featuring Tucker Carlson. While details of…
Stay Tuned: Major New Visual Explorer ASR Announcements Coming Next Week
Stay tuned for some major new announcements next week regarding speech transcription in the Visual Explorer using Google's new Chirp…
Google Developer Experts' DevFest: Experts Bootcamp
Kalev is back in Silicon Valley today for the Google Developer Experts' DevFest: Experts Bootcamp event at Google.
WashPost: Kevin McCarthy And His Caucus Contest Valuable Terrain: Attention
The Washington Post's Philip Bump examines media attention to the GOP caucus. Read The Full Article.
Iran's PressTV Covers Israeli Societal Unrest
Iran has been devoting airtime to covering societal unrest in Israel over proposed and enacted judicial reforms, portraying itself as…
Nature: Understanding Urban Flood Resilience In Chinese Sponge Cities Via GDELT GKG Analyses
Media data analytics has become prominent in enhancing urban flood resilience. Most of the media articles were posted during the…
Transcribing 9.5M Minutes Of Global Television News Through Google's Chirp
The TV News Visual Explorer encompasses selections of television news from 108 channels spanning 50 countries and territories in 35 languages…
The Perils Of LLMs For Translation Tasks On Lower-Resource Languages: Estonian Noun Declension
One of the great promises of large language models (LLMs) is their ability to revolutionize translation and linguistic tasks. A…
Google's Chirp & Truly Multilingual Global Speech Transcription: An Example Of Three Languages In 60 Seconds
Google's new Universal Speech Model called Chirp is a large speech model speech transcription system that offers state-of-the-art speech transcription…
Coming Soon: 6.6M Minutes Transcribed Through Google's Chirp
Our forthcoming Chirp ASR demo has reached 6.6 million minutes (110,000 hours) of global television news!
Temporal Relationship Between Daily Reports Of COVID-19 Infections And Related GDELT And Tweet Mentions
Social media platforms are valuable data sources in the study of public reactions to events such as natural disasters and…
The Rich Multilingual World Of Global Television News: New Possibilities For Understanding Codeswitching In News
For those accustomed to the monolingual English-only broadcasts of the United States, it can be nearly impossible to imagine the…
More Multilingual Transcription With GCP's Chirp: An Arabic-English South Sudan Broadcast
While GCP's new Chirp ASR model does not officially support multilingual audio content, it has proven highly adept at correctly…
Coming Soon: 2M Minutes Transcribed Through Google's Chirp
Our forthcoming Chirp ASR demo has reached 2 million minutes of global television news!