Google demands its adsense users to choose some trigger keywords/phrases. If they match a search of a Google user, the advertiser's ad might get displayed.
If the user clicks any such ad, they show interest in the topic of the ad, outlined by the trigger keywords. -- So, if one (or Google itself) were after to figure the interests of a user, Google could take these keywords and tag the user with them. -- The Google cookie one gets practically everywhere in the web, expiring in 2037, provides a unique handle for the user, for a long time. (And I really wouldn't wonder if they'd track 90% of all web users, individually.)
On the other hand, whenever a user enters a page featuring Google content, the user might become identified by their cookie. If Google is about to place ads on that site, by the cookie ID, Google immediately could figure the fields of interests of the visiting user. Also, immediately, it could update the user's record, since by visiting the page Google knew another bit about that user: He's likely interested in the topics dealt with on the page just entered. -- Which either the advertising real estate provider reveals by tagging its own page or which Google might do itself by traditional information science approaches.
Then, Google could select any ad that matches the user's profile and render it to the page currently being put out/the user is waiting for that it finishes rendering.
Is there a way to mimic that approach to gather tags of tags?
Updates: none so far
Showing posts with label MOM nodes. Show all posts
Showing posts with label MOM nodes. Show all posts
Monday, June 11, 2007
Tuesday, February 20, 2007
What is MOM?
MOM is the acronym for "Model of Meaning". It is founded on one basic assumption, that a notion always consists out of simpler notions or, alternatively, out of raw data -- like "light sensored"
Like ontologies, the MOM building of notions is represented as a graph, a heavily wired network, where the notions make the nodes and the edges indicate which notion consists out of which other notions. Other but common ontologies, MOM avoids edges injecting knowledge which is unreachable to a system that gets the network as its knowledge base. For example, in thesauri there are edges like "this is an antonym to that" and "this a broader term to that". And most of the thesaurus edges (called "relationships") are abstracting. So, for example a dog node may be immediately attached to a animal node. But that implies: The system cannot sense why the dog may be an animal. It fully depends on the humans who foster the network. The same in the case of antonym or even more sophisticated relationships. -- As far as I know, most ontologies tend to make use of such knowledge injecting edges.
Related to this, there is one another big difference between MOM and common classifications/thesauri (also known as "controlled vocabularies"): MOM builds upon a slightly different definition of the term "notion".
Usually, notions are thought of as a triangle of term, item, and thought of the item. MOM does not focus on the item, nor does it have any interest on the terms. (Different cultures have different terms for the same items, hence why care?) Thus, the MOM nodes in fact don't represent notions but just thoughts of items. That makes a huge difference: Relying on the traditional definition of notions, a classification or a thesaurus can validly define a motor vehicle to consist of e.g. a set of wheels, engine, body (amongst others), but validly it can not define a van to consist of a motor vehicle and other parts, since a motor vehicle, simply, is not a part of a van. MOM, on the other hand, ignores the chance that a physical item might be attached to a notion. Hence, yes, part of a van may be a motor vehicle.
This insight allowed to drop the dominating kind of relationship of traditional controlled vocabularies, the abstraction relationship. Which MOM dropped in fact. Which led to a single remaining kind of edges: The partial one. Questions?
Updates: none so far
Like ontologies, the MOM building of notions is represented as a graph, a heavily wired network, where the notions make the nodes and the edges indicate which notion consists out of which other notions. Other but common ontologies, MOM avoids edges injecting knowledge which is unreachable to a system that gets the network as its knowledge base. For example, in thesauri there are edges like "this is an antonym to that" and "this a broader term to that". And most of the thesaurus edges (called "relationships") are abstracting. So, for example a dog node may be immediately attached to a animal node. But that implies: The system cannot sense why the dog may be an animal. It fully depends on the humans who foster the network. The same in the case of antonym or even more sophisticated relationships. -- As far as I know, most ontologies tend to make use of such knowledge injecting edges.
Related to this, there is one another big difference between MOM and common classifications/thesauri (also known as "controlled vocabularies"): MOM builds upon a slightly different definition of the term "notion".
Usually, notions are thought of as a triangle of term, item, and thought of the item. MOM does not focus on the item, nor does it have any interest on the terms. (Different cultures have different terms for the same items, hence why care?) Thus, the MOM nodes in fact don't represent notions but just thoughts of items. That makes a huge difference: Relying on the traditional definition of notions, a classification or a thesaurus can validly define a motor vehicle to consist of e.g. a set of wheels, engine, body (amongst others), but validly it can not define a van to consist of a motor vehicle and other parts, since a motor vehicle, simply, is not a part of a van. MOM, on the other hand, ignores the chance that a physical item might be attached to a notion. Hence, yes, part of a van may be a motor vehicle.
This insight allowed to drop the dominating kind of relationship of traditional controlled vocabularies, the abstraction relationship. Which MOM dropped in fact. Which led to a single remaining kind of edges: The partial one. Questions?
Updates: none so far
Subscribe to:
Posts (Atom)
