Using examples to script traffic newscasts

Artificial intelligence learns by example, so the more example stories you provide, the better RadioNewsAI will be at replicating their style.

For more information on training RadioNewsAI with examples, check out our guide. On this page, we'll specifically be covering how to use examples to help script natural-sounding traffic newscasts.

While the app includes a default traffic model, it works best when given examples scripted the way you want it to sound on your station. To do this:

1. Open RadioNewsAI and click on 'Sources' from the menu.
2. Select your traffic source(s) by clicking on 'View'.
3. Click 'Test Traffic Feed' to pull the latest delays from TomTom's API. You may need to adjust your source settings to ensure it presents the data you want.
4. Copy the test data, then go to 'Models' from the main menu.
5. Select your traffic model, then click on the 'Examples' tab.
6. Paste the data into the 'Source content' box. For example:

id: xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
from: Manchester
to: Boston
rawfrom: Manchester
rawto: Boston
road: A 1 2 3
roadNumbers: ["A123"]
direction:
length: 2
delay: 12
event: Stationary traffic

7. In the 'Newscast content' box, translate the raw data into a script that reads naturally. For example:

There's slow-moving traffic on the A 1 2 3 between Manchester and Boston.
Expect delays of around 12 minutes.

8. Save your example and repeat the process a few times to improve script variety.

The more examples you provide, the less repetitive the AI will be when scripting travel updates. Stations have had the best results with 10-15 different examples.