AI systems may blow weathermen away
* 28 September 2005
* NewScientist.com news serviceWeather forecasters could find themselves pushed out of a job by an artificial intelligence system designed to write clearer, less ambiguous reports.
Computer scientists at the University of Aberdeen, UK, were asked to generate an “artificial weatherperson” by operators of offshore oil rigs, who wanted more clarity in their forecasts. The vocabulary used by different forecasters can be vague and highly variable, says Ehud Reiter, who led the Aberdeen team.
While this is simply an irritation to most of us, it can be a big headache for the offshore oil industry, where unexpected bad weather can damage equipment and threaten safety.
When Reiter and his team compared past weather bulletins with the raw forecast data on which they were based, they found a striking variability in the choice of words used by different forecasters. For instance, when they talked of “evening” weather, some meant conditions around 6 pm, while others meant much nearer midnight. “Late morning” could mean anywhere between 9 am and noon. The UK’s Met Office is also reviewing how effectively its forecasters communicate.
To remove such uncertainties, the team programmed a natural language generation (NLG) software package to transform data on the forecast weather into an unambiguous written bulletin (Artificial Intelligence, vol 167, p 137).
Similar systems could be used to clarify the medical notes of hospital critical care beds to prevent the phrases used to describe a patient’s condition being misinterpreted.
From issue 2518 of New Scientist magazine, 28 September 2005, page 27
[via]