Radio companies have been using artificial intelligence in their scheduling software, recommendation engines, and speech recognition for years, but the pace of change is now accelerating, and the development of generative AI has opened up the potential of artificial intelligence in radio. has made even greater strides.
Various speakers from last month radio days europe At the conference, I pointed out that tying all the new technologies together is already changing the way radio companies operate and strategize for the future.
Every new development is revolutionary in its own right, but the greatest potential for change lies in combining them all.
The most common scenario is to combine all tools to create new content for news or radio presentations. It works like this:
• Speech to Text engine converts speech to text
• Text in content that feeds into generative AI engines such as ChatGPT
• Generative AI engine analyzes content to learn station syntax, style and sentence structure
• When asked to write a news story or topic of interest, it scans the Internet for information on that topic and uses what it learns about the style of the station to compose. Use machine learning within the service to generate the best content.not just the first one found
• Generative AI output is sent to a text-to-speech engine for text-to-speech conversion.
• A speech synthesizer reads the text to the listener and applies compression and other AI-driven audio processing tools to make it sound right for the listener.
• When the listener is interacting with the content on the smart speaker, the listener can ask questions and the AI interaction engine will return detailed information in real time or whether the listener would like to hear more or relevant information. You can ask a few questions to find out. story.
Some experts believe that combining these new AI tools with smart speakers could mean the end of Google and the emergence of new types of search powered by voice and smart speakers. increase.
Satu Keto From YLE’s Radio Technology and Development Division Anna VosarykovaCzech Radio, told conference attendees how Finnish and Czech national broadcasters are using this technology to produce AI-generated crime-fiction podcasts, horror stories, interactive audio dramas, news reports, and generate website content.
“For media, the future is no longer digital transformation…we are already doing it. Transformation is now digital transformation.” Satu Ket said.
While learning about the possibilities from experimentation, speakers are encouraged to demonstrate their responsibilities, such as ensuring transparency by telling viewers that they are using AI, or warning viewers that they are using AI. He also pointed out that one broadcaster has learned that there are many issues to consider. Potential fraud with deepfake videos and false voices. “While our 2021 experiment showed that it is easy to steal a person’s voice that could be exploited for hoaxes, we also see how this technology can be used to improve listener interactions using digital twins. I was.
“We need to play around with these tools to understand how they help us focus on our work … to maintain a strong rapport with our audience.” , there are transparency and legal issues to discuss…we need to be able to authenticate ourselves. The audience.”
Over the last few years, we have seen malicious scammers forge radio social media accounts. In Europe, the government is already considering these issues and is expected to lead regulations on how this technology can be used, similar to his GDPR regulations for the internet. Responsible broadcasters will play a key role as society considers the impact of the latest AI tools.
content target
For advertisers and listeners, some new tools have many benefits and solve some of the privacy concerns associated with tracking. One of these is contextual targeting to improve recommendations .
By using speech-to-text recognition engines on radio programs and podcasts, intelligent podcast and streaming platforms are helping listeners to more accurately present content without knowing anything other than what they’ve heard before. will be Currently, recommendation engines aren’t very smart and rely on people tagging and categorizing shows and podcasts, and these categorization tools are matched against the tags and categories of previously-heard shows. and recommends, “If you like it, you might like it.” .’ However, these recommendations are often very inaccurate.
Contextual targeting uses a speech-to-text tool to analyze every sentence spoken on a podcast or catch-up radio show to learn more about what was actually being discussed and then select the show you just heard it from. match the actual content of the to provide closer information. Content matches are more likely to interest you.
white paper sounderdescribes how advertising can benefit from this technology without using any of the listener’s personal data. Advertisers can buy by listener interest rather than by general category or listener demographic. The Sounder whitepaper explains:
“Contextual targeting serves ads based on content relevance. Audience targeting targets specific types of users regardless of what content they are consuming at the time the ad is received. Like Apple’s genre. By analyzing show episodes and content to identify opportunities for contextual targeting, rather than relying solely on podcast metadata, media buyers are more confident in finding opportunities that are more relevant to their brand’s needs. You can have it and find it.Inventory contextual assessments can include more podcasts than ever before, significantly increasing the amount of content available to advertisers.”
audience research
radio analyzer has been around for some time, but its latest offering includes an audience research tool that maps audience meter rating data to program clocks and broadcast audio to determine which segments of a show are viewed by viewers. It provides a visual guide to what worked and what didn’t work for you.
Measurement tools use minute-by-minute analysis to help you determine exactly what engages listeners and what programmers, producers, and presenters can do to keep them entertained.
Here are some takeaways from analyzing the personality radio programs presented at the conference:
• Irrelevant weather forecasts lose listeners. Make your report convenient for your listeners by telling them to bring an umbrella or dress warmly. Make information available for weather forecasts, as well as other program segments.
• Talk at the end of the song to transition the listener from the music they love to the talk.
• Emotions such as crying and laughing attract the listener.
• Sports results are boring to listeners. Talk more about unique sports news items, not just results.
• Don’t put too many topics on the morning show. Use a few topics and come back to them in a new way every hour.
• Talking about the same topic throughout the day isn’t very effective after the first listen. Listeners feel that they have already heard and do not pay attention or interest. So if you go back to a topic, please explain how it was updated when you introduced it.
• Catch your audience quickly by going straight to the point of each topic.
journalism
Converting written news articles to speech, and audio articles to web reports, are becoming more and more common in newsrooms. Because we use AI to preserve the angles and basic facts of an article while formatting it into a different format for TV, radio, or the web. The journalist reformats the story differently for each medium so he doesn’t have to create three different versions of the story for each medium. Of course, journalists from responsible media organizations still check each platform’s work before publishing it.
Marianne Bouchardthe founder of a nonprofit that fosters news innovation, said:
“Use robots to collect data instead of trusting them to make decisions. Make your own editorial decisions. Be ethical and tell your audience what you are doing. Please… we need transparency.”
AI monitoring tools can monitor social media sites for keywords and images. For example, an AI analysis of his Russian TikTok found that TikToks mentioning the word “war” or showing pictures of soldiers with guns were not shown in Russia early in the invasion of Ukraine. I did.
Still confused about AI?
Mona Macaulaya leader in data transformation and strategy, artificial intelligence is data + algorithm = outputIt’s good for tracking and analytics, but doesn’t create any significant original content on its own.
Machine learning is the next evolution of AI and is what drives services like ChatGPT.
Machine learning creates algorithms, not just outputs, data + output = algorithmA feedback loop is created so the AI can continue to learn. Improve what you output.
New tools discussed by various presenters at the conference include:
RadioGPT – www.futurimedia.com/radigpt
Royalty-free music generated by AI – https://soundraw.io/create_music
Speech processing AI tools –
https://auphonic.com/features/leveler
www.cleanvoice.ai
Audience Data Matching and Sentiment Analysis – www.podderapp.com
AI-generated, copyright-free production music (main image) – https://soundraw.io/create_music
Investigative fact-checking tool for journalists used by Radio France – https://mbnuijten.com/statcheck
Secure Speech Transcription Service – https://www.mygoodtape.com
