E2 Reader- First interface rethinking and graph database implementation
The first throw of the nodegel navigation interface tried to replicate the inherent pattern of linking plus a few types of linking E2 does not provide. Feedback from the handful of invited testers showed that the interface was too complex, giving users power (choices) that may not be welcome or needed as a default.
Simplify the interface into three initial choices: New, Recommended, and Power Find. The basic split between Content and People is retained.
New presents a list of new writeups, as E2 New Writeups does, but adds options for filtering and lets you page back further into the past without changing contexts (i.e., clicking a link that take you to another page). It is the default starting point and the first item in the list is displayed for immediate reading.
Recommended presents a personalized list composed by using powerful recommendation techniques to present a short list of recommended articles. This list is easily expand in the same way as New. This function learns your preferences, including factors such as the degree of serendipity and controversy you favor.
Power Find reveals all the navigation avenues and options for using them, singly or in combination, with various filters. You can use this to modify how Recommended works for you or create a one-off list to explore or to search for something more specific.
The API E2 offers (and quite likely, the database structure that E2 uses) does not provide enough information for the E2 Reader to implement certain features of navigation and the recommendation function.
1) Create a graph database to store the required data with needed flexibility and make it fast to retrieve.
2) Add a new content evaluation function that is dedicated solely to the purpose of identifying user preferences for content. This differs from the E2 voting function in that it is not used directly for feediback to authors and is not made public as a 'reputation' number on the article.