Wishlist items and future development road map
Clone this wiki locally
Development roadmap up to the 'end of summer' re-announcement
Automated deployment to live
QA (quality assurance) process to ok pushing code up to live, including: UX review against standards and vision; design review against guideslines/samples; code/functionality review.
Documentation to enable others to add their APIs and to enable a one-page document for QA testing
Zotero integration with 'your sources as input' completed and any quirks documented
'Save results into Zotero' final code and UX plan agreed, integrated into the results screen for browser plugin users; team agrees is good to go live
'Save results into Zotero' final code and UX plan agreed, integrated into the results screen for Zotero standalone users; team agrees is good to go live
'Save results' as text final code and UX plan agreed, integrated into the results screen; team agrees is good to go live
Multilingual (mixed languages within input as well as languages other than English) code and UX plan agreed, integrated into the 'magic query box'; team agrees is good to go live
Review the randomness of the query generation algorithms - where does it need tweaking or is it good as it is?
Add functions to re-run the query with the same input text for additional serendipitous discovery (see also https://github.com/chnm/serendipomatic/issues/26)
Other wishlist items for possible 'end of summer' roadmap
(partly so we can remove some things from the issues queue that are better here)
Build in 'rearrange by' (group by via masonry filter function) on item attributes to be determined (based on minimum data quality) - item type, temporal, spatial fields would seem the most sensible but we may need to tweak depending on whether we can get data at a similar level of granularity (e.g. city/region/country, day, year, period, etc). See also https://github.com/chnm/serendipomatic/issues/29
Build the 'hint' function? i.e. 'If the results are a little too random, give us a date or place and we'll try again'. Is there a way to use hints about disciplines or is that not supported by the information we can get from APIs currently? And what impact does this have on serendipity?
Add a topic modelling function for parsing the text and test whether that gives us useful additional terms for the API queries
Pull in Zotero item abstracts and tags
Pull in an entire Zotero library (or change any text that promises that)
Pull in selected collections from a Zotero library
Add instructions for using items exported from other reference management systems (may also have a dependency on html stop words and coping with xml)
Other wishlist items
Change the front page layout on viewing after first successful go at the results screen to make the Zotero call to action a bit more prominent? The idea being that people are more likely to give Zotero a go after seeing how it works with other text. We could set a cookie when they've seen results to determine whether people see the 'on-boarding' or 'on-board' version.
Brian's twitter bot idea https://github.com/chnm/serendipomatic/issues/20.
API for serendipomatic itself so that other people can make use if it and integrate it in ways we haven't thought of? (farther in the future)
Make it possible to link to the "stash" results page with a seed url (e.g. wikipedia page) to allow sharing similar results (need to have text somewhere that makes it clear results won't be exactly the same every time)-- this might setup a simple way for the twitter bot to 'generate' / link to content; if we go this route, we should implement a twitter gallery card for linkable results pages, so the twitter feed will be able to display some of the results. (Note: use the twitter card validator before making this live)
Stable URLs for results obtained from Serendip-o-matic
DigitalNZ APIs http://www.digitalnz.org/developers
Are there any Asian/Middle Eastern/African national aggregators we can add as sources? Or for the moment is it a matter of adding subject-specific APIs like the Reciprocal Research Network API and checking the way place entities are recognised and set in queries to improve global coverage?