• View
• Edit
• History
• Print

(:pmgraphviz -- digraph { subgraph cluster_public_code{ public_code [shape=record,label="APIs|libraries|mashups|publications|meta_repositories"]; } subgraph cluster_seedea{ seedea [shape=record,label="<discovery>discovery\n(intelligence)|<integration>integration\n(hooks)|suggestion\n(stats/ML)|evaluation\n(stats/ML)|prediction\n(CAEE)"]; } subgraph cluster_client{ client [shape=record,label="<working_platform>content|user_preferences\n(explicit feedback)|user_history\n(implicit feedback)"]; } seedea:discovery -> public_code [label="categorisation"]; seedea:integration -> client:working_platform [label="adaptation"]; } :)

Suggestion mechanism

• MS_n = n_th micro-service
• MS_n+1 = new micro-service added thanks to the discovery mechanism, no history of usage
• U_m = m_th user
• U_m+1 = new user with no history of activity

See also Recked: A Night of Recommendation Technologies, January 2009

External data-mining

initially the principle was to use only data from the local Seedea instance (aka here but after reflection:

• the mining of public pmwikis
  • more generally other wikis (MediaWiki, ...)
    • even more generally other services providers

make a lot of sense. Adapters (see Yahoo! Execute or RunMyProcess for potential help) would have to be used so eventually a balance of effort/information would decide if it's sufficiently valuable but the opportunity must be studied, especially to bootstrap the system.

To do

• formulate the processes in Business Process Modeling Notation (BPMN)