ProductCategorization

From RCC 2007

Jump to: navigation, search

How can a large dataset, with tags that might be adhoc, be built, administered and delivered?

Don't let the perfect be the enemy of the good

Tech

how will the data appear on the page, how will it look? What is the delivery mechanism?
barrier to adoption for web clients of the data has to be low
clear distinctions between "unassailable" , generally quantitative data, and user-reviews and commentary. Both present different problems

[edit]

Tools

ontologies ex. papers
ML / provable systems (are these really one category Sandra?) Conf Link
Relational Theory Wikipedia Article
Unsupervised learning techniques Intro
IBM UIMA Unstructured Information Management Architecture
RDF models Simile Home Page

Mashups are happening right now. Companies like [StrikeIron] are providing data source services, and some tools like their Excel plugin that reads StrikIron sources.

RDF is notoriously abstracted and "not Done Yet"

Note that really large datasets are in practice not relational tables, but carefully chosen flat tables, because of the enormous compute resource cost associated with managing them.

Some Relevant Mashup Links

http://wiki.mashupcamp.com/index.php/LicensingCommericalMashup

http://blog.programmableweb.com/?p=521

http://redmonk.com/sogrady/2007/01/17/john-herren-mashup-camp/

[edit]