RSSWhich way to the Semantic Web?
Put together on April 22, 2008 10:55 pm by Dimitris
What do you think?
I’ve been watching closely Guy Kawasaki’s latest venture, Alltop. It’s a very simple project: split popular and diverse fields of interest into categories (e.g. from health to books and from programming to wine etc) and find the top links regarding each subject. I see Guy on Twitter every now and then request links of authoritative reference and breaking news sources from his followers. That way gradually an edited, user-generated corpus of links emerges. All such links are collected in one place for anyone who would like more information on a subject. In other words, Alltop aims to collate all must-know links about a particular subject. And if you were to ask a sommelier e.g. ‘Hey, where should I look for information about wine’ it’s to that list that they would point you.
The idea is neither new or terribly innovative – see e.g. popurls – and is based on a simple website (which cost only a few thousands of dollars to build) with no RSS or any other frills. And yes, quite a few people have argued that for Alltop’s target audience even RSS is something unnecessary. Alltop is aimed to the vast majority of individuals who just use the web as a quick tool to find the information they need and then move on with their work – instead of worshipping it like some of us.
This would mean that Alltop is simply about generating huge amounts of traffic – and the corresponding ad revenue – by bringing to the hordes of ‘average’ internet users what they need most: the top links out there. All nice and straightforward and similar in concept to what Mahalo and Squidoo are doing (which not so coincidentally are also run by two of the most well-known startup ‘luminaries/celebrities’).
However, I’m not sure if that’s all there is in Kawasaki’s mind (or Calacanis and Godin’s for that matter). These collections of links per subject can and should be used for another purpose too. At the not so distant future they can make up the base of vertical search engines to:
a. return more relevant results and
b. bring them closer to the much-hyped Semantic Web.
Now, about the first aim, if someone were to build a search engine optimised for a particular field, say healthcare, they could, in principle, achieve slightly (or even considerably) better results than Google, just by ‘handling with more focus’ what the query is and where the possible answers come from. And indeed another startup, Hakia, is doing exactly this. Starting from the narrow field of healthcare they plan to snatch some of Google’s traffic by developing their own search that will make users go ‘Google is a fine generic search engine but when it comes to healthcare Hakia rules’. And when they achieve that of course Hakia plans to move to the next field – simply by applying the same process there.
What is more interesting however, is the potential to develop a natural language processing platform for this particular subject. Narrowing down the field (e.g. healthcare) simplifies matters considerably compared to trying to tackle the whole semantic web problem at its full complexity. For example, instead of processing a vast number of data (that of the whole web – or even of just Wikipedia, as Hakia rival Powerset does), narrowing down the scope means analysing only a few hundred or at least thousands links. Furthermore, having a reduced corpus to process can mean the potential to go further into the deep web – something even Google is only now starting to do. Finally, focusing on a single field drastically reduces the ‘vocabulary’ the processing platform needs to be able to understand (esp. in fields like medicine, IT, marketing etc where the vocabulary is limited and heavily recycled even between different languages).
And once this practice is spread somewhat to various fields and queries to specialised search engines can be processed in natural language in each of them, a meta search engine can be developed. It will only need to understand the field of the user’s query initially and followingly which of the available natural language search engine (or engines) to tap on for the answer. I think such an approach of unifying a number of semantic sub-webs has more chances to work than inching towards a doubtful all-encompassing solution in one go.
For a by far more authoritative analysis of the Semantic Web prospects, check Tim-Berners Lee’s interview.
tags: analysis, idea |
Digg | |  |
Del.icio.us | |  |
Ma.gnolia | |  |
Newsvine | |  |
Technorati |
Comments
Leave a Reply - in English Please
