Thursday, 25 August 2011

Thoughts on eRaUI

After some consideration I have come up with a possible architecture for the eRaUI tool, which I think might be a strong candidate for fulfilling the aims of the proposal. It would also allow us to create an open-source tool of significant practical value. I shall summarise the system by means of bullet points:
  • eRaUI would be able to be plugged into any web-based system or website. It would not replace the existing system's interface, but would reside on top of it. In practical terms, this would probably mean that eRaUI would reside as a helper widget in the corner of a brower window, rather like the Google Translation widget and other similar tools.
  • eRaUI would firstly provide a search box, rather similar to Google's site search. Putting a query into the box would cause suggestions to be dynamically generated and displayed beneath the box. Submitting the contents of box could conceivably bring up a larger window for in detail search.
  • eRaUI would learn from previous searches, and the behaviour of users having made searches.
  • eRaUI would also allow the user to choose content by means of links beneath the search box. These links would by dynamically generated according to the contents of the box, the activity of the user and the user's cookie info, IP and other factors.
  • eRaUI would be receptive to conditions such as IP address (resolving to a particular country), browser http headers etc.
  • eRaUI could show previews of web content by means of cascading windows or some similar mechanism.
  • eRaUI would be based around jQuery on the client side for maximum browser acceptance. On the server side, eRaUI would employ PHP (for non-persistent requests) and Java.
  • eRaUI would be configurable to particular types of system, but would also work 'out of the box' for novice installers.
  • eRaUI back end activity could be handled by a single centralised server. This could enable machine-learning across a wide variety of systems.
If implemented would have the following positive aspects:
  1. Generic - can work not only with NaCTeM and MapTube but with any system.
  2. Practical - could be installed with minimal difficulty on top of any existing interface and improve the usability of an otherwise hard to navigate system.
  3. The system can make extensive and practical use of machine-learning algorithms.
Please let me know your thoughts!


  1. Providing and easy to install widget should help encouraging others to explore eRaUI, so this is something I find very attractive. It also delivers added benefit beyond just using it on NaCTem and MapTube.

    From what you outline here, it seems to me that this is mainly directed at learning from the behaviour of users on a particular site to make the search and browsing on that site more effective. However, you also mention individual users and cookies. How are your planning to address the new "cookie legislation" in this context?

    If users have to accept a cookie whenever they want to search a site with eRaUI a site owner may have second thoughts on implementing it if this adds another dialogue.

    Another aspect to consider is ease of implementation. Many CMS have decent search functionality by now, so you would have to present the benefits of eRaUI over integrated search and Google search very clearly to get people to try it. This relates more to "marketing", of course, but should be considered.

    But then I guess a particular value would be to use this on content-rich websites that also hold collections, databases etc. I assume this is what you mean by "configurable to particular types of system".


  2. Cookies do present something of a problem - but I think this is not something which would interfere with the core functionality of the widget. Even without any use of cookies, we would still be able to offer predictive search and all other proposed features. The main obstacle would be if we wanted to track individual users over multiple visits to remember their preferences/search tendencies. There would be additional problems if we attempted to track users across different domains / systems. Hopefully we could make the use of cookies entirely seemless (non-evident) while complying with regulations.

    It's true that a lot of CMS and of course Google already offer quite advanced site search capabilities. Our search would have a key advantage - results would be tailored to an individual site, not to the web as a whole. This tailoring would be tweaked by user-interaction with the site in a way in which a generic search engine could never achieve (due to lack of inside-site interaction data). As yet, I'm not aware of any CMS which offers a predictive (auto-completing / suggesting) search. Our search would go a step further by not only guessing the keyphrase being typed, but by offering suggestions which would be different to the search phrase but nonetheless of interest to the user.

    Indeed - when I wrote "configurable to particular types of system" - I had in mind that eRaUI would be particularly effective for facilitating the easy navigation of sites with a somewhat sprawling structure - particularly if users tend to go in predictable directions once entering the site. It would be less effective for CMS sites which already tend to be quite organised and have their own search facility.

  3. Thanks for this, Eamonn. The cookie law will turn out to be problematic for a range of applications and force developers to think about other ways to achieve what they want to do. So it is good to hear that the core functionality of eRaUI will not be affected. The approach you will take should be of interest to others who are looking for a way to offer personalisation under the new regime!

  4. Yes - hopefully the effects of the cookie legislation will not be too prohibitive anyway - but yes if it does make things difficult then eRaUI will still be able to function - albeit possibly with some missing functionality.