DefiningImageAccess/Tool/Exhibit
From ImageWeb
Contents |
Exhibit
Exhibit
Exhibit (http://simile.mit.edu/exhibit/) is a lightweight structured data publishing framework that lets people who know only basic HTML create Web pages with rich, dynamic visualization of structured data and faceted browsing and sorting on that structured data [Exhibit WWW 2007]. This means that one can enjoy a rich user interface styled by Exhibit without going through a high technical barrier:
- no need for software installation, configuration, or maintenance: One simply needs to create the structured data file(s) and an HTML file, which embeds a reference to the Exhibit Web API, and put them on your web server.
- no need for profound coding knowledge: To start quickly, one can copy and paste the HTML code from either Exhibit online tutorial or other demonstration web sites, linked through Exhibit home page.
Exhibit Fly-TED
By following the Exhibit online beginners' tutorial, we had the following interface to start with:
The user interface contains two parts: a content view panel and a facet browse panel:
- The content view panel displays a collection of items, whose layout can be customised by different views, such as thumbnail view or detail view. Figure X is showing the thumbnail view of images.
- The facet browse panel contains the facets by which users can filter the images in the content panel. Figure is X showing all the images of the gene CG10694 from the strain wild type (wt).
Create the Exhibit
To realise the above interface, we created the structured data and the HTML page to present these data.
Creating the structured data
Exhibit takes data in its own format - JSON[1]. We created the data file by exporting the image metadata in our Jena RDF database to an N3 file, which then gets translated into JSON using the Web-based tool of Babel (http://simile.mit.edu.babel).
Creating the HTML page
The initial HTML code was copied from Exhibit's tutorial and then customirised based on the feedbacks from users.
Exhibiting the six requirements
1. (YES) R1: Browse images by the gene names. 2. (YES) R2: Find all its images of one gene, which contain the same expression pattern, regardless of their strains. 3. (Partial) R3: Find all its images of one gene, which contain the same set of expression patterns, regardless of their strains.
In Exhibit, a selection of multiple filtering conditions within one facet does not construct a conjunction relationship between them, but a disjunction one. This makes it impossible for users to search for images all showing the same set of expression patterns.
First, we can build two copies of the same facet, to enable users to build a conjunction condition from the two facets. However, without renaming the property used by the facet, we cannot filter the conditions shown in the second facet when users make selection in the first facet. A hacking solution is to make a copy of the property and give it a different name, and then create a facet on the copies the property.
I do not think this would work either. I am working on it. (Jun)
4. (Partial) R4: Classify the images of one gene by their strains and then expression patterns. As above.
5. (Partial) R5: Find all the images showing a certain set of patterns of a particular strain. As above.
6. (TODO) R6: Find all the images containing both somatic cells and gene expressions in the germ line.
7. (NO) R7: Find all the images NOT showing a certain set of patterns of a particular strain.
Performance of Exhibit
RDF triples of varied size were used to evaluate the performance of Exhibit loading and browsing these triples. An open source toolkit iMarcos was used in this evaluation. iMarcos is a plugin to Firefox and IE. It can record users' interaction in a browser, and then replay this record. Before replaying the recording, users can set some stop watches in order to gather information about performance by editing its Marco file.
Our experiment showed that when loading 1,100 image metadata records (~18,000 RDF triples), Exhibit became unresponsive.
