Talking Point

Steven Collier's WebLog

Alternatives to SharePoint Search

I must admit that I've now become quite sceptical about all search vendors. They all claim to have the latest, greatest algorithms but really are all based on the same fundamental technology.

 

All searches are about the same when it comes to identifying documents which match the search parameters, SharePoint does this just fine, see how your search brings back 200+ documents, the file you wanted is in there, just on page 44 rather than where you want it.

 

The place where searches try to differentiate is in the ranking of these results, putting what you need on the first page. All the products I have seen just seen seem to use variants of the same type of formula relating the results to the frequency/significance of the words you are looking for. I can’t see why one would be especially better than any other. It’s an easy thing to try, most of these folk will provide evaluations or come with a demo version you can try onsite.

 

I don’t really believe that the view improvements such as relating documents together or graphical visualisation is really that great from a users perspective. They don’t use this functionality from Google, they just want to type a word and get a reply.

 

You will find some definitions around on how Google’s Page Rank technology works (http://www.iprcom.com/papers/pagerank/), it’s worth understanding so you can understand how you will never be able to recreate this on an intranet of only thousands of employees.

 

I’m waiting for search to start considering information beyond the simple lexical analysis of the documents. Why can’t the search know about me, and tailor my search results to put the IT Department expense claim form before the Finance Department Expense Claim form ? Why can’t the search learn that last time someone searched for X the document they wanted was on page 4, and the next time move the document up the ranking ? etc.

 

Until then, I know I’m going to get the most bang-for-buck by tailoring SharePoint search through thesaurus, best bets, noise-words, search scopes etc. For example we recently discovered that making a default search only consider the document created in the last 3 months made a big increase in user satisfaction. I strongly believe get better search results by paying for someone to manage search actively than paying $$$ for a specialised search package.

Posted: Nov 22 2004, 04:40 PM by stevenc | with 6 comment(s)
Filed under:

Comments

TrackBack said:

# November 22, 2004 8:43 AM

stevenc said:

Steven,

I received your feedback to my blog article about SharePoint search limitations.
http://www.sharepointblogs.com/bmixon/archive/2004/11/21/940.aspx

To be honest, we have looked at Verity and it too does not impress me. Autonomy is one that we have not looked at but may do so in the near future.

Keep up the great work; I know I appreciate it!

Bob Mixon
# November 23, 2004 5:07 AM

stevenc said:

Steven,

Fundamentally I agree with most of your points…

The issue with Enterprise Search is simply that there haven’t been any breakthroughs in this space comparable to what’s been done in the Internet Search space. Google’s Page Rank algorithm dramatically improved relevance. The real break through here was that this was done without requiring additional work from users. Most attempts at relevance feedback require input from users and it’s been demonstrated time and time again that users have no tolerance for this.

WRT your point about personalized search…Enterprise Search technology hasn’t even gotten to the point where ranking algorithms adjust to the unique needs of companies yet without manual tweaking. In your case, you adjusted the rank based on the needs of your company. But clearly the same adjustment wouldn’t work in a patent office, or law firm. There is still much work to do in this area.

Finally, In my view, the biggest problem with Enterprise Search is the poor user experience. On Google, you can move quickly from page to page. Contrast that with what happens in the Enterprise. When a user clicks a link, he first has to download the document, open it in the native application, and then take on the role of the search engine by searching through the document. With documents ranging in size from a one page memo to a fifty page market study, this inconsistent experience is arduous at best. Something like surfing the web on a 28k modem.

I’ve blogged about this here if you’re interested:
http://www.ba-insight.net/Default.aspx?tabid=30&mid=364&ctl=ViewEntry&EntryID=9

Great post Steven. Thanks!
# December 1, 2004 11:59 AM

stevenc said:

I get what you're saying about search vendors, and I agree that you will get better results by managing it yourselves.

However, perhaps search isn't the right way to go? After all, search is fine when you know what you want and the meta-data used to tag it, but sucks otherwise. Often, you want the user to know where to START looking for something and then lead them there as quickly as possible.

There is a tool called BrainEKP (Enterprise Knowledge Platform) that uses the well-known Brain mind-mapping GUI to link together documents, database records and web pages from across your enterprise. You can see it here:
http://www.thebrain.com/

It's a promising add-on to an enterprise portal (although I'm unsure of its SharePoint friendliness) and would allow your users to browse through connected bits of data however they like. One nice feature is that the connections are very fast to follow, and therefore it brings some speed back to the process.

For example, you might find the IT Dept Expense Form by following a link from an IT Dept item, or by going in through Employee Expenses, or perhaps Company Forms.
# January 6, 2005 10:47 AM

stevenc said:

Hi Angus,

I've seen similar search visualisation tools to the Brain thing before, the most funky was from Inxight (http://www.inxight.com/products/sdks/st/index.php) called the Star Tree which you can see here http://www.inxight.com/images/st_ebay_large.gif

Now to me that sort of thing is very clever, but I don't think my users will see it in the same light. My users want a search that works like Google, it seems to manage to return pretty much what they want when they give it a few words without much fuss, they want my intranet search to do the same.

The other point which I don't think I mentioned in the article is that no one way works for everyone. Some users will not search however great you made it, they like to browse through hierarchies. We need to accept that we need to have as many different ways of finding data as possible and allow users to use the one that suits them best.
# January 6, 2005 9:09 PM

stevenc said:

Hi,

You might want to look at a new product on the market called the iBox from a vendor called Interse (www.interse.com). The iBox is an add-on product for SharePoint and is unique in that it uses metadata to create a powerful search which allows users to do things like filter, browse through hierarchies, free text search - it caters for different styles of people - I agree with your point that no one way works for everyone.

Here is some information from Interse:

The iBox is an add-on software product for MS SharePoint that dramatically enhances the way professionals reuse unstructured information. The iBox integrates with SharePoint to provide an intuitive and high quality search interface for employees.

SharePoint search is not effectively able to use structures, including synonyms, hierarchies and relations, from backend systems, such as SAP, as metadata sources for easy and effective metadata tagging. The iBox adds value to your SharePoint investment through effective metatagging and search capabilities. The iBox provides users with a high quality search capability, and improved categorisation of content. With the iBox, the process of reusing unstructured information is enhanced by a factor 20-100 in terms of effort and time spent.

The iBox can be integrated into other technology platforms in a similar way as the combined iBox and MS SharePoint.


Hope this helps...
# June 24, 2005 9:28 AM
Leave a Comment

(required) 

(required) 

(optional)

(required)