Why search can be weird and frustrating (and fun!)

From CNet, we see that Craigslist bests MySpace as top search term. The article contains a graph from hitwise showing the relative popularity of searches for craigslist, myspace, and facebook.

This is weird to me, because these are the actual domain names of the things that the users are (in all likelihood) searching for. Typing them into the location bar in Firefox takes you right to the appropriate place in each case. But a search engine is how you find things right?

It seems pretty obvious to me that anyone searching for one of these is unlikely to want to find all the occurrences of the term craigslist and it seems like the obvious thing to do with such a query is to just have the search engine issue a redirect to the appropriate resource. Of course, Google lives by showing you ads, so they're not going to do that, but they should (and do!) have the site come up at the top of the listings.

Not to give away too much, but if you look at the top searches for Sun's intranet, you see the same phenomenon: a number of the top searches in any given month are the actual host names for internal tools (the one you use to book vacation, the one you use to reserve a conference room, etc.) as well as for things like java.

Properly speaking, this is a search problem, but not a search engine problem, per se. A search engine can easily find all the places where a given word (or domain name) occurs, but it takes someone who understands the user community of the search engine and the purposes to which a search interface is going to be put to make sure that useful things happen.

It seems to me that if you're going to offer a search interface to a wide community, you need to be sure that you're watching your query logs and noticing things like the fact that people are search for the internal tools and then making sure that queries for those things are being handled in the appropriate way.

For example, behind the firewall, a search for a particular Web-based application should simply re-direct to that application (perhaps with a delay for the one-in-a-million person who actually wanted to search for that term.) A search for a person's name (or a close spelling) should generate a results page with just their information from the corporate directory.

In my experience, however, search seems to start cycling around how much stuff is being crawled and what document formats are supported for indexing, rather than on how it can be made truly useful.

At least it means there's lots of low-hanging fruit for a dedicated search guy!

Comments:

Post a Comment:
Comments are closed for this entry.
About

This is Stephen Green's blog. It's about the theory and practice of text search engines, with occasional forays into recommendation and other technologies that can use a good text search engine. Steve is the PI of the Information Retrieval and Machine Learning project in Oracle Labs.

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today