Semantic Overflow

Question

3

2

The current web is heavily polluted with various forms of spam. Can semantic technologies help? If the spammers start using fake metadata then it could reduce the usefulness of what we currently have.

Could we make use of personal networks to assign reputation to sites and people? I've heard of people using their FOAF network as an email filter, i.e. only accept incoming email from people in their extended network. Google's social search can give you search results using network data in your profile that can be based on FOAF and XFM data.

This question was partly inspired by the classic Metacrap article.

I know I keep mentioning FOAF and XFN, but those are the semantic technologies which I have played with most.

edited Nov 2 at 12:33

asked Nov 2 at 8:24

steevc
185●6

Jiří Procházka · Accepted Answer · 2009-11-03 22:16:29Z UTC

Semantic Web has to track provenance and trust of data, judged by various people/agents/groups.

You should be able to state that you completely trust your friend's data, or that you trust somebody's judgment of trust - like some spam detection software/service.

You should be able to have multiple of such trust profiles, each for different purpose.

Unfortunately ontologies and tools for this are not yet well developed.

Personally I foresee a common Semantic Web gateway application for remote calls by applications, which would use the user's identity and trust information to present user a subset of the Semantic Web - only the information he trusts.

Ian Davis · Answer 2 · 2009-11-02 11:00:21Z UTC

I wrote recently about some of the vectors by which Spam could attack the semantic web. I don't offer any solutions but I think it's important to understand the vulnerable areas first. See http://iandavis.com/blog/2009/09/linked-data-spam-vectors

andymurd · Answer 3 · 2009-11-02 12:26:11Z UTC

My take is that spam is just a form of incorrect data, so it will be dealt with in a similar manner to markup that says (for example) that Sweden is a country in Africa. Semantic apps will need to calculate how much they trust their datasources, and look for consensus amongst disparate sources.

Search engines like Google already assign a measure of trust (PDF here) to domains and pages and assume that good pages don't link to bad ones. Something similar will soon have to be applied to semantic data.

Andrew Matthews · Answer 4 · 2009-11-03 21:05:49Z UTC

Perhaps the semantic web will be better able to accomodate advertising if the advertising has to be semantically relevant to gain linkage. The irritation about spam (discounting malware for a moment) is that the spam is frequently irrelevant. If it weren't irrelevant, I doubt people would be so irritated by it.

Obviously, there's nothing to stop spammers from, say, scanning the ontology referenced in some content and then randomly sprinkling the ontology's concepts all over their own advertisement. To detect that, you'd need something a bit more intelligent than a baseline semantic web application. Another evolutionary pressure towards general purpose AI.

I guess it depends what form of spam you're refering to:

unsolicited emails
comment spam
irrelevant wiki edits
semi-relevant answers that plug specific products...

Each of these has a different context that determines the contextual relevance requirements upon them.

Michel van Tol · Answer 5 · 2009-12-04 19:00:35Z UTC

0

I believe one measure that is worth mentioning is the sha-1 encoding of the mailbox in FOAF, but obviously it can only be used to check whether a known value corresponds to it. Whether it really does counter spam, I don't know...

answered Dec 4 at 19:00

Michel van Tol
51●4

How will the semantic web counter spam?

5 Answers