Forum Moderators: open
If we divide up the web into two elements - static(ish) web pages and dynamic(ish) users, we can then say:
Context-based approaches to search (eg Google/Applied Semantics) focus on the first, setting up algorithms with the web page as their focal point: ‘What does this page probably mean?’
While user-profiling approaches take the user as the focal point, building up algorithms around A.N.Other: ‘What does this user probably mean?’
With msn/Longhorn seemingly taking the user-profiling route, and Google/Applied Semantics taking the other – it seems like theres a fundamental difference in the philosophy of search developing.
I’d love to know what thoughts people might have on the simlarities & difference between these approaches, and especially the effect it would have on the design/layout/content of websites (and therefore our dear old game of SEO... :)
For example: would a user-profiling web result in websites aiming more specifically at personality types? Would a context-based web have a relatively much higher emphasis on content links?
If the context-based approach gives rise to the Semantic Web, what would a user-profiling approach result in?
The Persona Web? :)
If the context-based approach gives rise to the Semantic Web, what would a user-profiling approach result in?
I believe both need to co-exist in order for all the ideas out there to work propery.
If someone searches for jaguars, while occasionally searching for lions, panthers, and tigers they must like big cats right? In order to help provide special personalized results on big cats, something is needed to make sure a web page is about jaguar cats and not cars or sports teams. Now it might be easy to determine sites that talk about cats but what if this person is also known to browse around on sites that sell posters?
How do you know to offer them personalized results on posters of florida panthers(cats) and not the football team?
This is probably a problem that all these search companies face. How far to go along with the personalized targeting?
If Google goes personal and with their context-targeted advertising, do they keep things slightly targeted or attempt to go full-force?
If someone searches for florida panthers and they are known to like big cats, do you still keep the ad-serving technology keyword targeted and serve ads selling tickets to watch the Florida Panthers play football? Or will Google's goal be to look at various sites' keyword inventories to determine if they are sports associated or wildlife and then serve ads for supporting the preservation of panthers in Florida.
Am I on topic with what you talking about? :)
In particular, I hadn't figured the necessary mix between the approaches, in that the person would help set the context.
(I had been thinking of a different sort of user-profiling, along the lines of a separate personality profiling test/quiz, but I think this sort of ongoing user-learning is where I need to start.)
So, the user has established a pattern of, say, 'big cats', and that sets the context for weighting the choice of 'between these two, you're probably after this' should it arise in the future.
Following on, I would say that a personal profiling system can only be build on top of a general profiling system i.e. 1) there exist these choices, and 2) you tend to chose along these lines…’
I guess, bigger picture, that might be what Google's doing now - especially with the new 'extra-broad-matching' on its AdWords - i.e. establishing the meta-concepts with their associated concrete examples – ‘you can have this and this or this because they’re all the same thing, but not this because it's not within the overall coherence’.
(a cunning/wonderfully efficient idea - using other peoples money to determine web structure! :)
Taking this thinking further would be the really interesting idea of a point, before which user-learning works poorly because the meta-concepts aren't defined well enough (both on the web and in the mind of the search engine – forgive my anthropomorphism :) to be picked up in a coherent enough way to link back into an everyday searchers life...
... so you’re going along making your searches like normal, but...
... after this point, when the search engine is sufficiently sophisticated to pick up that 'st. bernard' goes along with the 'alsation' and 'spanial' that you typed in before, so you must be interested in “dogs”, and along with that perhaps “cats” under an interest in ”’animals’”, and so ‘how about this?’, ‘oh that’s great!’ – and you get into this great feedback loop, particularly because you can say ‘yes/no that’s what I did/didn’t mean’.
The thing being, for this approach to be taken up and the winner taking all, it has to work for most people for most of the time, not just dogs and cats, but boats, cars, houses, college courses, office furniture, you name it – an incredible challenge!
Maybe it’s a question of a search engine having to first adapt to the web (and the web to the search engine), before it can start adapting to us?
Is that how you see it working?
Of course who ever comes up with this technology will capture a big chunk of the search universe.
I am excited about the possibilities in this realm.
Thanks for the insights folks!
"Personalization has become an ever more important part of search, as a means to deliver more accurate results to users", Eric Schmidt, Google's chief executive officer, said Tuesday......The company is increasingly focusing on personalization as a means to improve its search results, with the acquisition last week of Kaltix, a start-up that builds search tools based on personalization and context."The primary mission of Google is to get you what you want, rather than what someone thinks you want," Schmidt said.
From this article [news.com.com]
Bolstered by this :) and having laid the ground a bit with the user-learning discussion (thanks msgraph! I noted the ‘personalization and context’ above :) I’ll continue with where I was going...
While I was at uni, I was particularly interested in how to arrange things to help you think – the whole area of representation and reasoning. So we looked at Donald Norman etc, and some really interesting stuff. When it came to dissertation time, one possibility was investigating nuances in the game of Scrabble...
...I’ll explain why I found this so interesting :) The tutor involved with this was very familiar with the process of arranging things to help thought processes – so familiar, in fact, that he wanted to take it the next stage further, and rather than investigating how to arrange the tiles on a scrabble holder in different ways to help someone come up with new words – he wanted to look at how this process varied with personality type.
Say for example with your standard Myers-Briggs categorization. So the effect would be: this personality type needs things presented like this, this personality type another way.
Couple of years later (and now really starting to wonder what I would have found had I pursued that particular research topic :) I’m involved in SEO and hear that search engines are going the way of user-profiling. So naturally I assume they're tying search to something like Myers-Briggs.
Its from this stand point that I’m wondering whether there could be two types, to search developed:
1) one that focuses on presenting you with the information according to whether you’re a more ‘sensing’ or ‘intuition’ type person, or ‘thinking’ or ‘feeling’.
2) the other on getting the context/cicrumstances as accurate as possible – getting all the appropriate information, and not worrying a great deal about its presentation.
So a key question would be: will people take a loss in context (circumstantial support) to be able to handle things in a way that suits them better? For me, this is a question that is bound to come up.
All this presumes the two approaches have sufficiently mutually exclusive goals & methods for them to evolve separately.
And I think they do. The key issue is, I believe, whether doing SEO for a website aiming at a personality sensitive search engine (‘truly’ personal), versus SEO for a hyper-context/circumstance sensitive search engine (‘truly’ semantic) would be different – which, by now, I think that it's clear that I do ;)
I just don’t think computers will ever be able to join the personal with circumstance, and for me that’s the heart of the matter.
If it does come to a Personal v. Semantic Web situation, or if a good enough compromise is found...
...what would what develop with the personality-profiling approach?
A number of different types of web experience (and therefore different types of internet as it evolves) according to the 3/4/16 major different ways of seeing things?
‘Yeah, I surf the intuitive web. You?’, ‘Nah, I’m more a sensing web guy’
or...
‘You won’t find him on the thinking web’
:)
Who's bitten recently by a tiger may want to buy a Jaguar today.
If you know him as involved in big living cat entertainment and hear him say: "Oh, I would like to have a jaguar!" - you likely don't understand that he's speaking about a car, actually.
"The primary mission of Google is to get you what you want, rather than what someone thinks you want,"
I doubt if that's really possible. As illustrated in my example above, it's always about 'giving me what someone thinks I want.' Except someone is able to read my mind. Or he asks for specification.
(also in the above should have been 'will ever be able to join the personal with the circumstantial' - just been in the copywriting forum, and felt prompted to review my english :)
I guess how I see it would need to work would be a search engine with a number of visual clues as to what the influences are on the choices being made, ranging from the large influences to the small and temporary.
With the decision making process clear, the onus is then on you to provide complete enough information, so you realise why a mistake may be made.
(Possibly available as a downloadable program - an end-user 'funnel' for search, so your more personal information stays on your computer?)
So for example, I've been bitten by a tiger, and I know that if I don't log on this information the search engine (more a social agent by now?) will make choices as before.
Which is completely reasonable, as I wouldn't expect a human to be able to do any better - if I don't tell a friend a relevant piece of information, they will make a 'mistake' but the fault will be mine.
So you'd go to 'Google-up-close-and-personal.com', you type in your search...
'Oh, I would like to have a jaguar!'
...and then the program (whatever) would provide a range of clues around the screen, indicating what influences it thinks are relevant and hence the decisions it's made using those.
And you see the image of a tigers head. (Biting the hand that fed it? ;) somewhere near to the offer of the cars available.
You think, 'No, that was ages ago, I'm over that' and remove the icon, or say, reduce it to minor influence.
The program then re-computes;
'Here are all the jaguars available for purchase in your area, and theres an ocelot on special offer down the road...'
