Friday, April 6, 2012

Data mining gone wrong... well tampered with anyway

Source:http://datamining.typepad.com/data_mining/2011/04/when-data-mining-goes-horribly-wrong.html

This article shows one of the fatal flaws with data mining if the processes for the computer to follow have been too widely defined or just told to collect and combine what data it relevant but not checking for relevancy. In this article Google made a mistake in the programming they used to mine for data in a social site called West Seattle Blog. The program started to combine information about different places and comments about these places because for the similarity of the names. This lead to conflicting comments like,"the Chinese food is great," and ," the PI moron strikes again," which are comment about two different places but posted on a bowling alleys comment page. Also the Seattle PI which is a local newspaper which has a link on the Google supported blog has been rerouted to display a page about a west Seattle killer responsible for slayings in Seattle having mental problems instead of the homepage of the newspaper. Google has to get on this problem and rectify the issues in their programming or it could get ugly. 

1 comment:

  1. I think this is just one example that demonstrates how huge and involved the internet is becoming. It is even hard for programs like Google to keep some things straight because there is so much information out there. Im sure there are many examples of issues like this. I know as a user of these types of search engines, data like this can be very frustrating. You search for one thing, and a million other irrelevant things show up. It can also make it hard to find reliable sources of information when you are not even sure if what you are searching is what you are actually receiving.
    Not only will mistakes and poor data mining like this frustrate users like me but it will make some search engines prevail over others for their accuracy as the internet grows. Key thought: type as many words as possible when searching to get results that you want!

    ReplyDelete