Comments on: Maybe text mining SHOULD be playing a bigger role in data warehousing

By: Michael

Michael — Sat, 14 Mar 2009 00:52:24 +0000

The marriage of Structured databases to Text or other syntactic/contextually related data is becoming a requirement for companies in the webspace. Much of the data by such companies is formless, or is being built at a rate which obviates the use of the traditional lifecycle management systems which include structured data analysis and data modeling. Time To Market is driving those processes out of the delivery cycle, making the job of doing analytics after the fact, very very difficult indeed. This problem is also manifested by the resistance from web companies to invest in software systems using legacy licensing schemes.

By: Jos

Jos — Fri, 14 Nov 2008 10:03:26 +0000

Wow, what a bunch of buzzwords. By the way, you should learn a little bit about how text mining and information extraction are two different thigs.

By: Curt Monash

Curt Monash — Sat, 25 Oct 2008 19:50:24 +0000

Remember, we’re talking about exhaustive extraction here.

http://www.texttechnologies.com/2007/10/05/when-to-use-exhaustive-extraction/

By: Daniel Tunkelang

Daniel Tunkelang — Sat, 25 Oct 2008 04:13:12 +0000

I’m all for promoting scalable software for information access (since its sales help pay my bills), but I’m a bit skeptical on the 10-20x multiplier. A big motivation for performing information extraction / text mining is to *reduce* the text to a form which a higher signal-to-noise ratio. There will be overhead, but such a large blow-up suggests replication rather than mining.

By: Jay Levitt

Jay Levitt — Fri, 24 Oct 2008 21:25:08 +0000

I have a much shorter theory:

Numbers make colorful graphs.

And graphs get you budgets.