Impact Journalism Day (June 25th, 2016) focuses on solutions-based journalism. Fifty-five leading media organisations from 50 countries, including The Irish Times, are sharing stories of innovative solutions to social issues around the world, in an initiative developed by Sparknews. Read our articles here: irishtimes.com/news/impact-journalism-day and follow the conversation on Twitter through #ImpactJournalism and #StoryOfChange

A common fear for first-time parents is not being able to understand their baby’s cry. But now a Taiwanese team have developed an app to tell parents if their baby is hungry, tired, in pain or needs a clean nappy.

Over three years, researchers collected about 300,000 sounds from 100 babies at the National Taiwan University Hospital. From this they developed the Infant Crying Translator, an app that uses a cloud-based database to decode babies' cries from newborn to six months old.

The app takes only 15 seconds to come up with a translation; tap “record” and a clip of the infant’s captured cry is uploaded to the database. The file is then compared to its samples and a verdict appears onscreen.

After winning an innovation award from Taiwan’s government in 2014, Infant Crying Translator went on sale for Android and iPhone in 2015 and now has about 10,000 users worldwide.

According to user feedback, accuracy reaches 92 per cent for babies under two weeks old. For babies under two months, accuracy is lower but showing an improvement, to reach as high as 85 per cent. For a four-month-old baby, accuracy is 77 per cent and rising.

Better judgments

“When we first launched, the app wasn’t accurate for babies older than two weeks. Back then the database only had audio of very young babies that we gathered from the hospital,” says the head researcher, Chuan-Yu Chang. “Now the library has many, many more sound files uploaded by users, and it can make increasingly better judgments for a wider range of ages.”

In addition, a machine-learning algorithm lets parents personalise the app for their baby. “Based on the actual situation for each baby, the cloud database will make unique data models to improve the recognition rate,” says Chang.

Developers at Yunlin University of Science and Technology are now working with the Taiwanese government to incorporate the baby translator in a mobile suite for expectant mothers, aimed at making parenthood less intimidating.

Chang’s team is convinced that a reliable baby reader has an important place in the global market. He also admits that many parents, even on a bad day, could outperform the app. But for first-time parents of any nationality the app could offer some clarity and relief when dealing with a crying newborn.

“From my own experience as a father, I know that sometimes when the baby cries the parents feel a bit like crying, too,” Chang says. “Humans have emotions and they make mistakes. The app doesn’t get flustered. It simply reads the data.”