IDEAS

Essential Interaction Elements of a Search Engine

23 de octubre de 2006David de Prado

The importance of information retrieval systems and the fact that search engines are among the most dynamic businesses in the Internet, has made everything related to them to be permanently a current issue. From the point of view of the interaction design, and most surely as a consequence of the abovementioned, search engines present a carefully studied framework that is being constantly reviewed. This unusual feature makes its development a fascinating phenomenon to witness, connecting us directly at the same time with many of the most innovative trends in the Internet.

Lee este artículo en español: Elementos de interacción esenciales de un buscador

 

This article will focus on global Internet search engines, however many of these things can be applied to other types of applications involving information retrieval systems.

WHAT DO THE USERS SEARCH ON THE INTERNET AND HOW

Before analysing the interaction elements upon which this article is focused, we must briefly go over some of the basic concepts that can help us to understand what do the users search on the internet and how.

Shneiderman, Byrd and Croft (Clarifying Serach: A User-Interface Framework for Text Searches) point out that the searching process can be divided into four stages:

1. Formulation of the searching strategy.
2. Action.
3. Review of results.
4. Query refinement.

 

According to these authors, the formulation, action, review and refinement of the queries done by the users follow an interactive process. They repeat this process over and over until they obtain relevant results that meet their needs. This process can take more or less time, but normally the number of users decrease as the number of refinements increase (according to Nielsen, up to 51% has success in their first query, only 32% in their second try and 18% in the third one). This situation gets worse due to the fact that generally the users of this type of applications tend to use scarce search terms (between two and three), so that refinement becomes essential.

Broder (A Taxonomy of Web Search -.pdf, 166 KB-) has made a taxonomy of web search that is widely used. He distinguishes three types of web search:

• Informational: the intent is to acquire information on a particular issue.
• Transactional: the intent is to locate a site in which the interaction consists in a transaction.
• Navigational: the intent is to locate a particular site the user has in mind or knows of its existence.

 

HOME INTERFACE

The basic challenge addressed in this point is to build an interaction system that must be consistent with the interface the user is going to see while the list of results is displayed. The mania for filling up empty spaces and the fact that normally demands are related to the need of showing the abilities and collections of the search engine can become important enemies.

We have to bear in mind that the homepage must make the start of the web search as easier as possible, since it is possible that the formulation of the search strategy has already been made. This means to:

• Provide the strictly necessary information.
• Introduce only relevant graphic elements.
• Place the focus (at an aesthetic level and a programming level) in the search box.
• Enable the start of the search by pressing the enter key.

 

The interaction elements that are strictly necessary consist of:

• The search box.
• The search button.

The search box

The search box normally has a long shape to encourage users to introduce several search terms. By end of 2004, Gary Price presented the data of 2,8 words as the average size of the query in the Blog de Search Engine Watch. We must remember that the positive experience of a user with a search engine will depend in the end of one's finding what one is searching for, and this becomes more possible if one can formulate a query with four or five terms rather than just one. Thus, it is essential for designing search interfaces to consider this fact and to encourage the formulation of queries using several search terms.

Belkin (Interface Techniques for Making Searching for Information More Effective -.pdf, 101KB -) even suggests the use of not only long boxes, but boxes allowing more than one line. A survey carried out in 2002 found out that using a box allowing five lines of text caused searches with more terms than if the box used was of one single line. However, none of the major search engines with market dominance have implemented this solution.

The Search Button

The search button must be clearly labelled ("search the Web", for example) and regarding its size and location the Fitt Law must be considered. This law affirms that the required time for reaching a target depends on its distance and size. Therefore, some search engines place the search button underneath or left to the search box, giving it a considerable size.

 

LIST OF SEARCH RESULTS

The list of search results is the space where the user, due to the interactive nature of this process, spends most of the time before finding the longed resource.

Thus, it is essential for the list of search results to:

• Inform clearly what is being searched for and what results are being visualised.
• Provide search alternatives for alternative spelling or queries that need a further formulation.
• Allow the scan of results.
• Allow the user to move easily through results.
• Allow query reformulation.
• Reduce as possible the need to remind all the search process (with a query history, for example).

Search Interface

The list of search results must have a search interface that enables the query refinement. The same features ruling the home interface should also be applied to the interface displaying results, but here with added challenges - we must use less space (the results are the kings), the fact of having less space should not mean that both interfaces have been designed inconsistently.

Search Status

The information provided by the search status is necessary to put the search experience into context, allow to consult efficiently the results and to make a query reformulation, if necessary. The status must inform about

• What is being searched for (the search term(s))
• Where it is being searched for (the search index being consulted)
• Which results have been found (size of the search universe).
• Which results are being visualised in a specific page.

This is the typical informative element that is valued or considered only when you miss it.

The status display varies from one search engine to another: some of them concentrate the information in one single element and in one single place (msn), other divide the information in two segments that are placed separately but with a gestalt relationship (Google, Ask).

In my opinion, one should tend to concentrate the information in one single place, though it is true that in the abovementioned, some pieces of information in the status are more important than others (the search term is more important than the total number of retrieved results), so it might become interesting to highlight through graphic elements some pieces against others, or even distribute them throughout the working space adapting their size to the availability of such space.

The place or places containing the status information must be necessarily at the beginning of the list of search results: in this way, the users find it useful when they are dealing with those results and, in case of confusion, they can check what is being looked for and where.

Spellchecking and Suggestions

The spellchecking is a very useful element and provides a positive search experience. It is true that users can make spelling or grammar mistakes when they are carrying out a search and the Spell-checking systems can respond to approximately 70% of these mistakes (Avi Rapapaport Search query Spellchecking 2003 -.doc, 71KB-)

Normally it is presented as an element with two parts:

• Identifying periphrasis "Did you mean:"
• Suggestion.

It is convenient for both parts to be differentiated; the suggestion must go highlighted, and it must have the same search engine link-type appearance.

The spellchecking must be placed in such a way that enables its quick access, which will allow an immediate query reformulation if it has not been the correct one. Normally it is placed at the beginning and at the end of the list of search results. The efficiency of placing query reformulation elements at the end of the list of search results has been proven. Very often it is precisely when we have finished consulting the list of search results that we wonder whether we could make a more productive query.

The suggestions or relevant Feedback respond to the fact that many times the users do not have a clear idea on how to formulate efficiently the query they have in mind. Search engines like Yahoo!, and Ask in a more complex way, offer this type of functionality. Depending on the complexity of its functionality, the taxonomy normally tends to be very similar to the Spellchecking taxonomy regarding shape and location.

List

The list must be easily scanned, from which we can draw the following three conclusions:

• Each one of the results must be clearly differentiated.
• Each one of the elements forming up a result (title and description at least) must be clearly identifiable.
• The fonts must have a size that is adequate for the users of the search engine or at least it should count with the possibility of adapting them in an easy way.

With regards to numbering results as some search engines do (Yahoo! or Clusty, for instance), I do not have yet an opinion. The main point against this practice, according to my view, is that it can suggest there is a hierarchy of relevance, and for me this would be a delicate issue. It is true, though, that numbering results facilitates the application of a reference to them. Also the users find easier to locate one result among others, and these two facts cannot be simply ignored.

Paging

The paging is an interface element that helps the users to move themselves along their search, so its importance is capital.
Practically all search engines use a very simple interface paging with links to the different pages of results (normally they run every 10 pages) and links of "previous" and "next".

All the major search engines place this element where it becomes necessary for the user, just after the list of search results and the suggestions/spellcheck. Some of them are centered (Yahoo!, Google, Ask), others go in the left corner (Clusty or Alltheweb) and other go in the right corner (msn).

Users scan the results starting from left to right, which in my opinion makes the two first options (left and center) as the best location option, as they do not imply an interruption in the reading direction of the users. I believe the option of msn intends to place the paging near the scroll bar so that the users have to move less their mouse if they want to see more results.

As you can see, this article has focused on essential interaction elements, but there are still many more of them. I encourage the readers to use abundantly the message system, to complete this article and leave critics, comments or interesting articles related to this issue.