The home page and query form for the BioText Search Engine, as used in the usability study described in this article. To help ease users into a novel search interface, it can be useful to provide hints about how the interface works in a simple manner on the home page of the website. The BioText search engine employs two such techniques. The home page shows links to sample queries as a low-effort way to entice users who are unsure how to start into interacting with the system. The home page also provides reduced-size renderings of three different search results views. At the top of each view within the interface appears the query entry form. It provides: 1. Links to information about the search engine and the collection it indexes. 2. The entry form for entering the query, along with a search button (hitting Return on the keyboard is equivalent to selecting this button). 3. A radio button for selecting among the different subcollections and views (full text with figure view, table text with table view, etc). 4. A drop-down menu selector for choosing the search results ordering (one of Relevance, Descending Date, Ascending Date). 5. A label indicating the number of retrieved documents, and hyperlinks allowing the user to select which page of results to view. Additionally, the full text with figure display view shows a set of checkboxes that dynamically determine which parts of the documents surrogates to show or omit (abstract, full text excerpts, and figures). By default, all three are selected, but if the user unselects any of the checkboxes, that selection is retained into subsequent searches. The footer of each search results page also shows the total number of hits, along with the selector for other pages of hits. At the time of the study, selecting a radio button did not automatically re-run the query and change the results view; rather, the user had to select the Search button to activate the change (as noted in the study results below, the automated behavior is expected).

The figure captions list view shows the results of searching over the text of the article's title, the authors, and the figure captions, so the items retrieved can differ from that of full text figure view (Figure 2). The figure for each document is shown as a larger-sized thumbnail (compared to the full text view) and is shown to the right of the title and caption text, to emphasize the difference with the full text figure view. It was thought it would be important to signal this difference because the search results differ when searching over caption versus full text. As in the full text search view, clicking on the figure's thumbnail shows the full size image along with its caption in a new window.

The table view searches over the text within tables as well as the table captions and the corresponding articles' titles. The matching tables are displayed together with their captions and the title, authors and citation of the article they originate from. The results are shown as a vertical list consisting of the article's title and other metadata followed by the caption followed by an HTML rendering of the table, so that all tables in the interface have one consistent appearance. If more than one table occurs within a given article, the other information is currently repeated for each table. Tables longer than 8 rows long are truncated and a link is shown that allows the user to expand the table to its full length. Additionally, a checkbox is shown in the query form area that allows the user to see all tables fully expanded; the default is for this to be unselected.

The full text search with figure display searches over, and consecutively displays, the article's title, authors, abstract, and the full text from the body of the article. This is the primary (and default) view for the system. For each retrieved document, a display is shown on the left-hand side that consists of a vertical list of textual information from the article. This list consists of the document's metadata (title, authors, journal, publication date), the document's abstract, and excerpts from the full text of the document that contain query term hits. Additionally, for each retrieved document, small thumbnail versions of the first six figures from that article are shown on the right-hand side of the display, along with a link to see the document summary view labeled “View all K figures and captions” where K is the actual number of figures found in the document. Clicking on the figure itself produces a new page that shows the full-size figure along with its caption text. Throughout the entire BioText interface, each text area type is assigned a background color, and this color is kept consistent throughout the interface (e.g., yellow is the background color for caption text in all of the interface views). As mentioned above, the user can choose to reduce the amount of information displayed, either for the current query or for the entire search session. For example, the user can choose to view only titles and figures. Additionally, if a text area exceeds a predefined length threshold (approximately 500 characters), the text is cut off, and an ellipsis is shown along with a link to “Show Full Abstract” or “Show Full Excerpts.” If the user clicks on the link, the text is expanded in place, and at the end of the text a link is shown that allows the user to reverse the procedure (“Shorten Abstract” or “Shorten Excerpts”). Query terms are highlighted in the text areas via boldface font. In the title area, the title text is all in boldface, and so a yellow highlight background is applied to the query term hits within the title. In the full text as well as all the other search results views, a hyperlink is shown that allows the user to view the article directly as HTML or PDF, or in the summary version provided on the BioText site.

The figure captions search and figure grid view searches in the same manner as in the figure captions list view (Figure 3), but shows the matching figures as thumbnails arranged in a 4×5 grid layout together with some citation information. This view is intended to be similar to that of web image search interfaces. Beneath each figure is shown a link to view the full-sized figure and its caption and another link to go to the article summary view. The first 100 (approximately) characters of the caption are shown, followed by an ellipsis. A mouse hover over the figure shows more complete metadata about the figure, namely its title, authors, journal and publication date.

In each of the four results views (Figures 2–5), the user can click on a link that brings up the full text of the paper, either in HTML or PDF format (there is no need to first go through PubMed because the collection consists entirely of Open Access documents). Alternatively, the user can select a link labeled “View All Figures and Tables from this Article,” which is shown alongside search hits. This produces a view in HTML showing the article title (along with authors, journal, and date), abstract, and a vertical display view of all of the figures and tables alongside their captions. During the study, there was a problem with this view in that it retained the query form at the top of the page but removed the query that produced the search results listing that led to the article. A better solution is either to remove the query form from this view, or provide a link back to the search results for the previously issued query.