Visual Bracketing for Web search Result Visualization Jonathan C. Roberts and Edward Suvanaphen Computing Laboratory, University of Kent j.c.roberts@kent.ac.uk, es45@kent.ac.uk Abstract This paper introduces a novel visual bracketing method that provides detail-in-context views where the inner part contains the Focus bracketed by the context information at a lower semantic level We demonstrate two designs for web search result visualization that obviate the need to frequently select the next button for viewing multiple search results. Keywords: Focus+context, detail in context, greeking, bracketing, multiple views. 1. Introduction Search engines are an important part of information gathering, and serve as road signs for the Internet; they point the users in the right direction enabling them to find specific sites or general information about a topic. Traditional search engines (e.g. Google, Yahoo and Excite) visualize the results in the form of a text-based list. The actual data contained within each returned list varies from search engine to search engine, but in general, the title of the page is returned, a short extract of some text from the page is included along with a URL link to the page. However one problem with search engines is that they return too much information, this coupled with a traditional text representation means that there is not enough space on the screen to display all the results. The search engine may find 100 or 100,000 pages of information relevant to the query. Conventionally the result list is divided into smaller portions and returned 10, 50 or 100 at a time, but it is only feasible to view 5 to 10 results at any one time on the screen without scrolling which this forces the user to click on the next button, to view the next set of results Some critics may argue that search engines are becoming more effective at finding the ‘correct’ answer. While this may be true, it is often more beneficial to allow the user to select and decipher what they believe to be a correct answer rather than displaying a reduced set of results. Certainly, from our experiences browsing a larger set of results is useful, because the users develop and change their understanding of what and how they are searching based on the returned results. Thus, we believe it is more prudent to visualize a large number of the results. That may be achieved through distortion or focus+context methods that will provide a hierarchical view of the data. Our aim is to design a visualization that (1) utilizes Focus+Context techniques enabling the user to more effectively browse through the results and quickly drill down to relevant information, while gaining an understanding of the context of the selected information; (2) uses multiple views to provide different levels of detail and additional information; (3) to display the results in a way that would be innately familiar to traditional web users. We achieve this, first by using a text representation that will be familiar to the users and second by a bracketing technique where lower ‘level of detail’ brackets the regular representation to provide the detail in context. 2. Data Gathering & Visualization This research focuses on visualization techniques rather than the data gathering process. Thus, we use the Google API to provide the required data, which provides a convenient way to gather web search result data and integrate it with an application. 2.1. Search results The Google API is well-written and simple to use. For example, the search term string can be submitted to Google using the API, and the results are subsequently returned in a list of elements each named GoogleSearchResultElement’s (SRE’s). One SRE contains nine variables concerning one url that matches the keyword search. In this work we focus on four of the main variables (title, URL, paragraph of text and page size). We use a model-view-controller design- pattern in our implementation, and thus store each of the SRE’s in a list that can be visualized on demand. Web searching provides a rich data source and there are many variables that may be gathered and displayed. Indeed, some search engines such as Google and Proceedings of the Seventh International Conference on Information Visualization (IV’03) 1093-9547/03 $17.00 © 2003 IEEE