Home » Venn Search Interface

Venn Search Interface

Venn Search Interface is a functional prototype I built for a course I took in Sprint 99', CMU; advised by Suguru Ishizaki (suguru@andrew.cmu.edu).

Note: This relies on Altavista's search engine. However, they have modified the syntax of the query string. A live version is no longer provided.

Question...

What do we need to find what we want on the web? Good search engines? Good search skills? Good luck?

What if...

Thinking of a situation where you're doing a research on the relationship between Clinton and NATO. You're not sure about which direction you're going yet. You think it might be useful to find news about Clinton and NATO, but not about recent actions in Kosovo. What would you do?

To utilize the power of search engines, one way to describe your query will be: ("Bill Clinton" AND "NATO" AND (NOT "Kosovo"))

It's not trivial for most people; moreover, if the number exceeds 5 or 6, I, as an engineer, would need a piece of paper to figure it out and would need to type in a long query string.

Problems

There are couple of problem within current search engine interfaces:

• Current search engines use boolean operators (AND, OR, etc.) for uses to describe their queries. However, boolean operators confuse people in several ways.(e.g. in English, OR usually indicates one of two but not both). And to form a complex query using boolean logic is usually difficult.
• Phrases: As you can see, I used quotes on "Bill Clinton" which represents that it's a phrase. Is it a common sense?
• Inconsistence: the format of query strings differs from search engines to search engines. It's absolulte painful to memorize it.
• It's always frustrating to see "No result found" when you type in too many keywords and "100000 results found" when you eliminate some of them. You just don't know where to go and keep trying your luck.

What is Venn Search Interface?

Venn Search Interface is a visual search interface working on top of Altavista™; it helps users to describe and refine their query.

Design issues:

• Minimal design: Show only the interface that is necessary. For example, in the start-up screen, there's only a textbox for users to input text.
• Efficient design: I don't want this interface to be one that requires more keystrokes than a general search interface. For example, what we usually do is to input a word, space, and another word. Now all you need is to press ENTER.
• Instant feedback: As I mentioned above, it's frustrating that I input lots of keywords and try to be as specific as possible, but it turns out to be "no result".
• Refinement: As indicated in Saito's paper, Cognitive Model for Searching Ill-defined Targets on the Web, users refine their search strategies along with searching on the web. Thus, Venn Search Interface works like an assistant where your search results are shown aside.
• Navigation: In the next version of VSI, history panel will be implemented. In this way, not only the small navigation on the top-left corner of your window, but the history panel will help the users to navigate among their search histories.

Screen shots

Here contains the screenshots for my first prototype. Since I'm writing a paper about it for CHI 2000, the prototype would need work. Some functionality needs to be added, too.

Basically, it starts as an applet resides in a web page.

After you click on it, Venn Search Interface appears.

Following the example we talked about, we want information about "Bill Clinton" and NATO, but not about Kosovo. We filled in the first query string like this..

Within seconds (It does a query to Altavista, thus the time needed depends on the networking), you got a idea about how many matches there are for "bill clinton". You continues type in NATO and Kosovo respectively. You'll get a result like this:

From this, you got an idea about the amount of information you're gonna get from a search. Since we want "Bill Clinton", "NATO", but Kosovo, we select the area marked "5106".

The area will be "selected" and another browser window will be opened, which executes the query, shows you the information so that you can see how far you've got. The Venn Search Interface stays on the top of the window. like ...

From the pages returned, you go through some of the articles and find that the topic, "NATO and Clinton's altitute toward Iraq and Pakistan", is interesting. You press "Refine".

The query result you've made will become a circle so that you can put other query strings into the other 2 circles. The previous one will be shown on the left-top so that you can go back.

Future work

As I mentioned, I'd like to write a CHI paper for this. For the interface, there's still something to improve; for the functionality, the history panel should be placed underneath so that you can overlook the process. The direction of the CHI paper is not decided yet.

Last Updated : 7/19/2005 by Bo-chieh Yang (bcyang@alumni.cmu.edu)