Comments on: Another Difference Between Enterprise Search and Web Search

By: Life, the Universe, and SEO | The Noisy Channel

Life, the Universe, and SEO | The Noisy Channel — Mon, 24 Nov 2008 04:56:16 +0000

[...] So, is SEO good or bad? Today, it’s actually necessary. Google and other web search engines rely on SEO efforts to compensate for the limitations of their indexing. This is an example of where sites share in the responsibilitiy for contextualizing a user’s experience. [...]

By: Daniel Tunkelang

Daniel Tunkelang — Sun, 09 Nov 2008 03:01:42 +0000

Jeremy, I’m with you. I’ve invited Nitin Mangtani to come over here and join the discussion. We’ll see if he bites.

By: jeremy

jeremy — Sun, 09 Nov 2008 00:08:15 +0000

No, no fight. I was just seeking to clarify what each of us meant about clustering.

Am I incorrect? Does Google Web not offer (user-facing, exploratory-search directed) clustering, and Google Enterprise offer it? If so, how can it be true, as the google enterprise guy said in that interview that you posted a few days ago, that they are trying to provide the same experience in the enterprise as they are on the web? That seems not to be the case, which was my main point.

By: Daniel Tunkelang

Daniel Tunkelang — Fri, 07 Nov 2008 18:50:53 +0000

Fight, fight, fight! But seriously, let me me try to clarify.

Many queries lead users to portals even if they don’t have navigational intent. A search for a notable book title likely yields Amazon and Wikipedia pages as top results. A search for a person yields LinkedIn and Facebook pages. These aren’t navigational queries in the sense of Broder et al., but they do lead to portals.

And of course I agree that, to the extent that web search companies are doing interesting work, it’s almost entirely opaque to the user.

By: jeremy

jeremy — Fri, 07 Nov 2008 17:46:43 +0000

Clustering that is exposed to the user, I mean. Clustering w/ interactive feedback. Google Enterprise has it. Google Web does not.

By: FD

FD — Fri, 07 Nov 2008 16:52:14 +0000

i “suspect” there is more clustering in web search than you think.

wrt the tail. head queries are pretty easy to detect and handle. tail queries are harder to detect, sometimes easy to handle, sometimes hard. in general, that’s what differentiates systems.

By: jeremy

jeremy — Fri, 07 Nov 2008 16:35:57 +0000

However, there’s a very long, rich tail and the majority of the interesting work goes on at that end.

Could you give specific examples? Because frankly, I don’t see evidence of this work that you’re talking about.

If anything, I see a shocking lack of this work on the web. Take, for example, a company like Google. In their enterprise search, they do clustering of results. In their web search they do not. Why?

By: Moteurs de recherche d’entreprise versus Google

Moteurs de recherche d’entreprise versus Google — Fri, 07 Nov 2008 15:42:08 +0000

[...] Tunkelang souligne que chercher des documents au sein d’une entreprise est parfois plus difficile que de [...]

By: FD

FD — Fri, 07 Nov 2008 14:58:45 +0000

I think one should be very careful about suggesting that the majority of web search is about getting people to portal sites. Certainly, the most popular queries have navigational intent. However, there’s a very long, rich tail and the majority of the interesting work goes on at that end. Representing web search by navigational queries is like representing english by stopwords.

That said, I’m not arguing that web IR = enterprise IR = legal IR = cross-lingual IR. Each corpus and query base has its particularities that need to be understood well before applying IR techniques.

By: Daniel Tunkelang

Daniel Tunkelang — Fri, 07 Nov 2008 02:29:48 +0000

Jeremy: you’re right. I do think that the popular evaluation web search reflects a selection bias towards the queries that it handles well. Moreover, the ecosystem adapts: there is a entire industry of search engine optimization, which goes beyond pursuing traffic to creating landing pages best designed to exploit it. For queries that don’t land in this sweet spot, web search isn’t quite as compelling an experience.

Daniel L.: history is written by the victors.

By: Daniel Lemire

Daniel Lemire — Fri, 07 Nov 2008 02:19:54 +0000

(Yet, it is now the reference.)

By: jeremy

jeremy — Fri, 07 Nov 2008 02:19:44 +0000

Daniel T: I agree with what you are saying. However, an even larger-overlooked problem is the fact that the web itself also contains regions or subsets that are quite enterprise-like in nature. They have exactly the properties that you describe. And even though these web pages are just that.. web pages, most internet search engines do a horrible job finding them.

By: Daniel Tunkelang

Daniel Tunkelang — Fri, 07 Nov 2008 01:55:14 +0000

It is funny. When I went through the history of search in my MIT talk, I tried to make the point that search existed for a long time before the web, but that the mass adoption of the web meant that most people’s first experience with search was web search. Before the web, searching electronic collections was something that most of us associated with libraries–and often with professional librarians.

By: Daniel Lemire

Daniel Lemire — Fri, 07 Nov 2008 01:29:14 +0000

Excellent point.

Isn’t it funny that people doing enterprise search have to “differentiate” themselves from web search?

If we go back in time, “enterprise search” came first… by something like 30 years. Ok. We did not have full text search in the sixties, but many of the first “enterprise” computers were used to help retrieve documents. Web search is a relatively recent addition. Yet, it is not the reference.