 |
|
|
|
|
|
|
|
 |
 |
 |
 |
 |
 |
 |
|
 |
|
||||||
|
||||||||||||||||||||||
 :: About :: Newsletter :: Membership :: Reports :: Meetings :: Conferences :: Listserv
::
Links :: SAOUG's Blog |
How do you make a computer think like a human? A case study regarding a clippings service in the banking industry Dianne Kruger, Rand Merchant Bank Over the years the RMB InfoZone has provided a press clipping service to the staff of RMB – much like other libraries. Over the years it grew and grew until we were supplying news tracks on over 1400 companies and over 400 topics – which were being provided by means of individualised emails to over 500 clients. Then the crunch came – the person responsible for this monster retired! What to do?? We didn’t want just news feeds we wanted the personalisation, the emails, the archive and also allow growth of this service. We realised that we had been rapidly outgrowing the capacity of one person (and her unpaid husband who shared the load) and that her early retirement had just brought forward the inevitable – the need for automation. However, it is not quite as simple as just typing in a couple of keywords and waiting for the news to flow in – it needs to be accurate news without too much noise – thus requiring filtering. We also realised how much we rely upon the innate sorting and subject knowledge of the human brain. For example – trying to retrieve news on the South African economy requires x amount of structured keywords – the task was daunting!! As we progressed we also realised that we were having to use the librarianship skills that Google and other such search engines have made us neglect – detailed search construction such as those used in the days of old time dial-up Dialog! Really brain wracking stuff but so rewarding – trying to teach a computer to think like a human! Due to the size and scope of the project it was very exciting, after selecting a service provider (3 were considered and trials were run) we only had a 6 week planning and implementation period as we wished to launch it during the first weeks of January 2008 - as everyone else returned back to work from their holidays.
|
|
|
Updated:
|
||