CampusROAR: Improving our content

What makes for a good RSS feed?  Or more specifically, what makes an RSS feed look attractive when viewed in a ‘Personalised Magazine’ format.  That’s what I’ve been trying to find out and have come up with three (plus one control case) potential improvements.  These all utilise the ECS ePrints RSS feed which can be accessed here: http://eprints.ecs.soton.ac.uk/cgi/latest_tool?output=RSS2.

Control Case

This is the base case consisting of the default ePrints RSS feed.  Not much to say about this one: the body consists of just the authors and there are no embedded images.

Control Case - Unedited RSS feed.

http://eprints.ecs.soton.ac.uk/cgi/latest_tool?output=RSS2

Case 1: PDF Preview and Article Abstract

In Case 1, the ePrints generated PDF preview image is used and embedded in the RSS item.  Also, I have replaced the default list of authors with the abstract of the paper which is much more useful.

Case 1: PDF preview and article abstract.

+ Easy to find and insert preview for items where one is present.

– Preview not present for all items in feed.

– In many cases is not much more interesting than the control case.

http://users.ecs.soton.ac.uk/ad10g08/ab-prv.rss

Case 2: Article Abstract and Image Extracted from PDF Document

In this case I have again used the abstract as the body, but for the image I have used an image extraction tool to extract images from each PDF, and then hand-picked the most interesting one from each item, and embedded that.

Case 2: Article abstract and PDF extracted image.

+ Relevant to the item.

+ Can sometimes give a more attractive image than case 1.

– Not all items have a PDF present, and not all PDFs present contain extractable images.

– Many items do not contain attractive or interesting images.

– Hard to select which image is the best automatically.

http://users.ecs.soton.ac.uk/ad10g08/Abstract-pdfimg.rss

Case 3: Article Abstract and Keyword Image Searching

In this case I again used the abstract, and also added an image which was found using keyword search for creative commons images on Flickr.  The images were sorted by “interesting”, and the top image selected.  The keywords were selected by hand from primarily the keyword field if present, or if not by hand selecting an important word from the title.

Case 3: Article abstract and keyword searched image.

+ Images can be found for every item.

+ Images are more exciting and interesting in most cases.

– Hard to select keywords, especially in cases where keyword field not present.

– Images are the least relevant to the item of all the cases.

– Some selected images may be unsuitable for purpose.

http://users.ecs.soton.ac.uk/ad10g08/Abstract-keywordImage.rss

Leave a Reply

Your email address will not be published. Required fields are marked *

*