ReadWrite Web has a recent article on how semantic search works, Semantic Search: Myth and Reality. The article points out that semantic search will not solve all our search problems, it’s simply impossible. For those working with datasets, there will always be a great deal of cross-tabulation, manipulation (to reformat and present data) and manual work to be done to bring together answers to queries. Yes, semantic search might help save time in the initial stages of a query by giving more meaning to the terms we use to query with, but search is still, at its heart, a human construct and will always be open to interpretation and error. The article makes the important point that -
These are computationally challenging problems that really have nothing to do with understanding semantics. The misconception has been perpetuated since early days of the Semantic Web that somehow, because we will annotate the web, we will be able to solve these super complex problems. This is simply not true. There are fundamental limits to what we can compute, and a class of problems that have an exponential number of possible solutions is not going to be magically solved because we represent data as RDF.
Right now we search based on word occurance (ignoring for the moment Google’s fancy rankings, inbound link rankings, and other fancy search criteria and patterns in high-end databases). In the future we will search using semantics for words+concepts+inferred trust. We will probably never be able to search purely by opinion or emotion. And that’s fine - because we can make that judgement for ourselves. We’re human after all.
Entries (RSS)
June 5th, 2008 at 11:24 pm
words+concepts+inferred trust …. there is another element … the notion of ourselves as a context. I’ve been doing some work on trying to understand how to model personal ontologies in order to inject them into an application to provide it with context about who we are. Think of it as a glorified personal profile. One of the difficulties though is the highly temporal nature of our existence … and the lack of any real tools with which to readily express that.
For example …
If my personal ontology could describe to Google that I’m currently on holiday in mexico, then when I search for Chihuahua, it would give me results about the town rather than the breed of dog.
It’s a crude example … but it’s something I’m interested in.
Nice write up though Fiona
June 5th, 2008 at 11:30 pm
Yes! Absolutely! I would love to be able to do that. I would be very interested to see what you come up with.
August 5th, 2008 at 7:03 pm
@Nadeem Personal ontologies are hugely important but there’s already a lot of information in Web 2.0 sites that can help. For example, I might tag some pages with “toread” in del.icio.us, some music with “party” in last.fm and like some friendfeed entries about kittens.
When we take these highly personal tags and cross-match them with tags shared across many users it should be possible to estimate a set of interests (an APML file?) that says I like to read articles on system admin, party to 80’s hip-hop and look at pictures of kittens. There’s a further challenge, though - my interests change over the years so there needs to be a weighting favouring recent items.