Keyword Concerns

Adding the human touch to search engines could allow users to retrieve the information they want much more quickly, according to a report to be published in a forthcoming issue of IJIPT.

Every day, millions of us use search engines to look for information. We enter some text into a search box and retrieve page after page of results, most of which will be useless. It is not a particular efficient way of finding information, when you think about it.

According to Darius Pfitzner, Kenneth Treharne, and David Powers of the School of Computer Science, Engineering and Mathematics, at Flinders University of South Australia, this reflects, not a lack of skill on the search engine users’ part, but a lack of penetration of key developments in Human Computer Interaction (HCI) designed to make document retrieval easier.

They point out that in the context of document search, which is the vast majority of searching that takes place on the World Wide Web, the value of textual language is self-evident for searching natural language documents. However, they suggest that the human touch has in most cases not been factored into the development of search engine tools.

The researchers suggest that general user preferences, such as the number of words we prefer to use to describe or search for a document, should be embedded into search tools. While keyword identification has been studied in depth and is familiar to computer scientists and SEO hackers the world over, there is a lack of basic research into the human angle regarding how to construct a search query. In other words, the way search tools work has been engineered for the computers themselves rather than the people using them.

It would be much better if text search could draw on our brains’ powerful and innate pattern recognition abilities, the researchers say, this would also allow us to exploit the way the brain can organize knowledge in memory and so solve the search problem in a much more efficient manner. This concept could turn the search engines upside down.

Taking this approach would balance the computer processing overheads for search tasks with the user’s skills, or lack thereof, to create the optimal search string, and perhaps targeting the actual results they need rather than generating page after page of garbage, so avoiding the GIGO paradigm (Garbage In, Garbage Out). The team is now working on developing just such a system. Part of their preliminary experiments suggest that presenting results graphically as clusters of documents each with a short description is a more effective way of displaying results.

Darius Pfitzner, Kenneth Treharne, David Powers (2008). User keyword preference: the Nwords and Rwords experiments International Journal of Internet Protocol Technology, 3 (3) DOI: 10.1504/IJIPT.2008.020947