[Extension 'scheme-3' (56364) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable strScheme

[Extension 'outputScheme' (56370) on page 'science-cubed-pop-hit.htm'] Line 50: ReferenceError: strScheme is not defined

[Extension 'Show Nav' (8929) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable showNav

[Extension 'Hide Nav 2' (71380) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable showNav2

[Extension 'Show Slot 1' (8800) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable showSlot1

[Extension 'Show 3 Links' (8868) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable noOfLinks

[Extension 'Show 8 Slots' (8818) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable noOfSlots

[Extension 'Hide Slot 3' (9108) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable showSlot3

[Extension 'Hide Slot 4' (42266) on page 'science-cubed-pop-hit.htm'] Line 1: TypeError: assignment to undeclared variable showSlot4

	Pop hit detector

Chart hit or miss
The UK music industry is worth over £4bn and with a diversifying digital market, UK artists and entrepreneurs are looking for new ways to engage with fans. Hoping to reach a wider audience, Dr Tijl de Bie and his team, at the University of Bristol’s Intelligent Systems Laboratory, have found a way for science to predict whether a song is a chart hit or miss.

Machine learning by example
Sparked by academic curiosity, they wanted to show a music audience the potential of the machine learning method. Machine learning is a set of artificial intelligence tools to allow the computer to learn from examples. De Bie explains how, ‘The computer is able to look at the chart to see the songs that are at the top and at the bottom and learns how to distinguish them, based on the examples, so we never have to explain to the computer how to distinguish them. That’s the power of machine learning.’

Extracting a range of features from the audio input, such as the duration of the song, the tempo, how simple the song is from a musical perspective, they then can be turned into a numerical value. Then the machine learning method looks at all those sets of feature values for all the songs in the top 40 charts. As each feature value and chart position is calculated, the equation becomes more refined.

Digital Portable Radio © Murat Giray Kaya - iStockphoto

Score your own tunes
This is not a recipe for how to write a hit. It is a method that for a given song can give a score whether it is more likely to go the top of the charts than to the bottom. The team are happy that their scoring is 60% correct in predicting the chart hit, based solely on the audio information. As De Bie points out, there are other aspects of the song, such as the marketing budget and the popularity of the artist that affect the hit potential.

The ScoreAHit app has been developed with Panagiotis Tigkas, a former University of Bristol MSc student, which fans can register to use online. With the web app you can score existing songs by title and band name, or musicans can score any of their own songs, and potentially add value to the music.

LearnEnglish Science activities
Why not do a language activity based on this cubed story, Pop hit detector? You can double-click on any word on this page for a dictionary definition.