Re: "Two-way Search" demo anyone?

2009/7/28 François Dongier <francois.dongier@gmail.com>:
> Right, I was surprised to see Dave Winer suggesting just a pointer to the
> searcher's blog URI. It certainly would seem more efficient to have the
> query include a pointer to a *rich*, structured, user-editable user-profile.

True, a person's profile URI or even their own URI would bypass a
step. But if the person has a profile online it can be pointed to from
their blog/homepage for autodiscovery:

  <link rel="meta" type="application/rdf+xml" title="FOAF"
               href="http://example.com/~you/foaf.rdf"/>

(from http://xmlns.com/foaf/spec/#sec-autodesc )

When a person *doesn't* have a profile online, it may still be
possible to narrow the search focus through data extracted from their
blog (and/or other publications). If you're lucky there'll be some
meta tags or maybe links to well-known entities (e.g. Wikipedia URIs).
Links to less well-known material might also be useful if that remote
material contains some explicit data.

Natural language-derived entities drawn from a person's blog may well
be useful in the context of search, for example running Winer's blog -

http://scripting.com/
through
http://www.alchemyapi.com/api/entity/

- produces output including the entity "Berlusconi".

While Winer != Berlusconi, the fact that Winer has blogged about him
indicates some level of recent interest, just the kind of thing you
want for more focussed search.

So although the personal profile approach is probably the most
desirable, more general filtering using RDF based entity matching
(more or less 'similar' or 'related' pages) would be nice to have.

Cheers,
Danny.

-- 
http://danny.ayers.name

Received on Tuesday, 28 July 2009 11:19:54 UTC