How Web Users Search
How users search is well documented. They are lazy searchers by nature, prone to using short queries and the default search mode. To determine whether some search results are relevant, they don't look at their full text, but at the first 10 titles, URLs, and short descriptions (1).
To reformulate their queries, they prefer to do so by adding, deleting, modifying, repositioning, or recycling search terms. In general, they don't consult lists of terms to expand or modify queries, nor they are prone to using other search modes (2 - 13).
To do the latter, they would need to know which modes are available and how to use these. Search modes are called through boolean operators or commands that are included in a query or invoked through an advanced search interface.
Most advanced interfaces are machine-centered where design is driven by database architectural requirements. Many of these force users to do multiple tasks to accomplish a single search, overwhelming them with filters, selection menus, checkboxes, radio buttons, and tiny text. Adding insult to injury, some advanced search interfaces are placed in obscure secondary pages so users must search for these in the first place!
Why build a Match Previews Interface
When we built the Minerazzi platform, we took into consideration all these drawbacks. In addition, one of our primary goals was to examine the following scenario.
Suppose that users querying a search engine are given the power to know in advance the number of matches from each available search mode.
- Could they be induced to:
- reformulate search modes?
- reformulate search terms?
- Could they have a search experience similar to that of:
- an "expert guiding a novice"?
- a "novice following an expert"?
In 1.a and 1.b, we investigate whether user behaviors and lazyness can be modified by design.
In 2.a, the search engine provides qualified search alternatives or paths to follow. In 2.b, a user with limited search skills is compelled to follow a qualified search path.
Does the outcome of these scenarios hold for users of vertical and horizontal search engines? Could their search experience improve?
So when we developed our platform, we wanted to address these questions, which we believe pose a new search paradigm in information retrieval. We also wanted to build a search interface that would allow users to reformulate search modes, preview match counts, and retrieve results with one or few clicks. As a result, we ended up building our Match Previews search interface, which is depicted in Figure 1. Match counts correspond to a search for [ digital marketing ] using the default AND mode.
Figure 1. Match Previews.
Users can view/hide the interface by toggling the "Match Previews" blue link near the top of the search results page section. The interface lists 8 types of search modes in two states: matches and nonmatches, for a total of 16 unique modes. Because these states complement each others, nonmatches are also called complement or complementary modes.
When a user accesses the interface, the label of the mode being used changes to a blue link pointing to a help resource describing that mode. For instance, if the user submitted an AND search and accesses the interface, the label of said mode changes to a link that reads "About AND". Clicking this link retrieves a document that explains what AND matches and nonmatches are.
Using the interface is a straightforward process. A user submits a query, and Minerazzi returns search results plus match counts from each available search mode. Clicking on a match count resubmits the query in that mode. It is that easy! You no longer need to memorize search modes to do advanced searches.
If you prefer to search in the old fashion way of including search modes in your query, you can do that, too. Just submit a query using the following format
query = search operator:search terms
where [ search operator ] is the search mode being used and [ search terms ] are your query terms. Notice that these fragments are delimited by a colon (":").
However, if using complement modes, these must start with the letter "N" to indicate that these find nonmatches. For example, typing NAND:[ search terms ] retrieves AND nonmatches. The only exception to this "N rule" is the XNOR mode (the complement of XOR) which for historical reasons is written in that way. As Minerazzi searches are case-insensitive, you can include search modes in lower or upper cases. Again, you can avoid all this hassle by using our Match Previews interface.
The interface layout was thoroughly thought out to reflects the fact that the accuracy of the results increases as we use search modes at the right of the OR mode. On the other hand, the exclusiveness of the results increases as we use search modes at the left of the OR mode. With more than two terms, however, sometimes the exclusiveness of the results does not follow this ordering (i.e., XOR can be more exclusive than NOT).
In general, our layout offers several benefits. First, users know which search modes are available. Second, they can realize which search mode returns no results, few results, or a lot of results. Third, they can make an educated choice about the next query mode to use so self-generate feedback is used as a search strategy.
We have tried other interface layouts involving objects like sliders, pull-down menus, knobs, and so forth, but these require of multiple clicks and new learning curves to accomplish a single search. We are open to design suggestions as long as these do not violate the Principle of Least Effort (15).
Allowing users to adopt a search strategy by running multimodal searches (universal queries) and then previewing match counts from all available search modes introduces an interesting search paradigm.
Multimodal searching is not the same as multimodal information retrieval (MMIR), an expression used in the IR literature to describe the retrieval of information in any modality like text, image, audio, video... (14). Also, universal queries is not the same as universal searches, an expression adopted by Google to describe the blending of horizontal and vertical search results.
We believe that giving users the power to previewing match counts can help them to develop a search strategy, reformulate search modes, and spend more time interacting with a search interface.
We look forward to more studies to assess whether previewing search counts with a knowledge domain-specific search engine provides a search experience similar to that of an "expert guiding a novice" or a "novice following an expert".
We welcome any research or collaborative work along these lines from any interested third party. Our platform has the necessary architecture to accomplish these and similar data mining studies.
- Agichtein, E., Brill, E., and Dumais, S. (2006). Improving Web Search Ranking by Incorporating User Behavior Information.
- Wikipedia, (Accessed on 9-9-2013). Web Search Query.
- Huang, J., and Efthimiadis, E. N. (2009). Analyzing and Evaluating Query Reformulation Strategies in Web Search Logs.
- Teevan, J., Adar, E., Jones, R., and Potts, M. (2006). History Repeats Itself: Repeat Queries in Yahoo's Logs.
- Popovic, V. (2003). Expert and Novice Users Models and their Application to the Design Process.
- Buzikashvili, N. (2002). A myth, reality and extremal principles of seeking in different search environments.
- Rieh, S. Y., and Xie, H. (2001). Patterns and Sequences of Multiple Query Reformulations in Web Searching:A Preliminary Study.
- Spink, A., and Gunar, O. (2001). E-commerce Web queries: Excite and Ask Jeeves study.
- Jansen, B. J. Spink, A., and Saracevic, T. (2000). Real Life, Real Users, and Real Needs: A Study and Analysis.
- Lazonder, A. W., Biemans, H. J. A., and Wopereis, I. G. J. H. (2000). Differences between Novice and Experienced Users in Searching Information on the World Wide We.
- Silverstein, C., Henzinger, M., Marais, H., and Moricz, M. (1998). Analysis of a very large AltaVista Query Log.
- Dillon, A., and Song, M. (1997). An empirical comparison of the usability for novice and expert searchers of a textual and a graphic interface to an art-resource database.
- Bruza, P. D., and Dennis, S. (1997). Query Reformulation on the Internet.
- Bokhari, M. U., and Hasan, F. (2013). Multimodal Information Retrieval: Challenges and Future Trends.
- Wikipedia (Accessed on 9-9-2013). Principle of Least Effort.