A VOICEMAIL system that labels messages according to the caller's tone of voice could soon be helping people identify which messages are the most urgent. The software, called Emotive Alert, is designed by Zeynep Inanoglu and Ron Caneel of the Media Lab at the Massachusetts Institute of Technology.

It might be installed at the phone exchange or in an intelligent answering machine, where it will listen to incoming messages and send the recipient a text message along with an emoticon indicating whether the message is urgent, happy, excited or formal.

It works by extracting the distribution of volume, pitch and speech rate - the ratio of words to pauses - in the first 10 seconds of each message, and then comparing them with eight stored "acoustical fingerprints" that roughly represent eight emotional states: urgent or not urgent; formal or informal; happy or sad; excited or calm.

The fingerprints were created by "learning" software, which was fed hundreds of snippets from old voicemail messages that had been assigned emotional labels by the researchers. In use, the software looks for the acoustical fingerprint that is closest to the characteristics of the voice message and sends the recipient the corresponding emoticon. It also sends a text message indicating the two best-matching emotional labels.

In tests on real-life messages, the software was able to tell the difference between excited and calm and between happy and sad, but found it harder to distinguish between formal and informal, and urgent and non-urgent. Inanoglu suspects this is because excitement and happiness are often conveyed through speech rate and volume, which the software measures - while formality and urgency are normally expressed through words, or personal nuances in volume and speech speed, which it does not measure.

“In the future, machines will know more about our emotions and respond in accordance with them”She has also combined the software with a speech-recognition system that links patterns of words, such as an increased use of negatives, to particular emotions. However, doing this deprives it of one of its talents: the ability to label messages in any language. She hopes soon to build a system in which the acoustical fingerprint can be personalised for somebody's most frequent callers.