Lists Home |
Date Index |
>Please explain. I'm nervous about Golems
>that reason to recommendations based on humans
>trusting each other. If I ask the opinion of
>the International Brotherhood of Bowlers,
>I get a different answer than if I ask the guys
>International Wrestling Federation (maybe not much but
>some difference). So I have to ask, who trusts whom
>and how can I know that in advance who to ask?
Agreed, it's a big tricky thing, especially the bootstrapping part.
Though there are already a few precedents like the stuff behind online
shopping and even spam filters. Tricks like google & citeseer's use of
references might come in handy too. But sorting out the good assertions from
the bad (or just plain dull) is only possible when the things in the
assertion can be unambiguously identified - are we talking about the World
Wrestling Federation or the World Wildlife Fund? (again, maybe not much but
some difference). Doesn't really matter what kind of identifiers are used,
just as long as they're shareable.
>The semantic web cannot escape the problem of
>identifying the "preferred reading". Humans
Nope, but the web can already help a bit, and a bit more metadata and agents
that can use it should help enhance that.
Old TimBL note, mentions trust :
>From: Danny Ayers [mailto:email@example.com]
>>but requires a different means to enforce uniqueness.
>not necessarily - the 'web of trust' kind of stuff should cover
>I can't remember seeing it discussed on the level of identifiers