Studying Time: 2 minutes
Synthetic intelligence is a fraught subject for journalists — simply ask the man who
However for all its dangers, AI additionally presents alternatives we’re simply now beginning to perceive. For instance, Wisconsin Watch has been an early person and accomplice with Gigafact on an AI-powered software they’ve constructed that may assist analyze the hundreds of hours of podcasts, social media movies and discuss radio applications that may very well be spreading misinformation day-after-day.
The software, often known as Parser, can course of an hourlong audio file in a matter of minutes and never solely present a transcript, but in addition establish particular claims made in the course of the audio section and even the particular person making the declare.
Wisconsin Watch truth briefs reporter Tom Kertscher has been utilizing Parser to make it simpler to search out stunning and doubtful claims. Earlier than Parser he would take heed to these hourlong podcasts and radio exhibits himself, making an attempt to choose up on what Wisconsin politicians have been saying. In monitoring how a lot time it took to supply a truth transient, we present in some instances nearly half the time was spent simply looking for a declare.
Parser has sped up that course of, making it attainable to scan by means of much more audio recordings of interviews.
“We are able to cowl a lot extra floor with Parser, checking many extra politicians and interviews than we may manually,” Kertscher mentioned.
Gigafact started growing Parser after Wisconsin Watch offered that suggestions on how a lot time it might probably take to remain on high of each declare that each politician makes. However the issue of misinformation is way greater than simply preserving tabs on politicians.
Final 12 months the investigative journalism class at UW-Madison labored on a undertaking about discuss radio in Wisconsin. One of many key findings was the notable quantity of misinformation being unfold on the airwaves, particularly amongst conservative pundits.
To try this undertaking, college students spent a big period of time listening to 6 radio hosts whose viewpoints spanned the political spectrum. They took 4 hours for every host from the week after the Tremendous Bowl — 24 hours of audio whole — and manually processed the audio right into a database of claims. Even with a transcription software, the method took simply over 100 hours to supply an inventory of claims to fact-check.
Earlier this 12 months, I labored with Gigafact utilizing Parser to course of 24 hours from the identical hosts the week after this 12 months’s Tremendous Bowl. We got here up with an inventory of claims in two hours.
Wisconsin Watch and Gigafact offered that case examine in utilizing AI at a current Journalism Educators Institute convention hosted by the College of Wisconsin-Madison’s Faculty of Journalism and Mass Communication. We’ll current it once more this week on the Investigative Reporters and Editors convention in New Orleans.
And if you happen to haven’t learn it but, add our investigative journalism undertaking Change is on the Air to your summer time studying record. Sadly, for the scholars who devoted so many hours to listening and re-listening to these discuss radio hosts, it was not produced utilizing AI. However possibly subsequent time.
Wisconsin Watch is a nonprofit, nonpartisan newsroom. Subscribe to our newsletters for unique tales and our Friday information roundup.
Republish our articles totally free, on-line or in print, beneath a Artistic Commons license.
