We explain briefly the web application implementation of Exploratory Parsing support. We intend to preserve the server-side mechanisms while replacing the client-side with more easily shared and managed federated wiki pages.
Exploratory Parsing asks that one understand some aspect of a corpus by developing a grammar that finds, counts, and catalogs features in that corpus.
Web Application
The web application supports this process through numerous features each with a built-for-purpose mechanism.
Exploratory Parsing web application display grammar changes between two runs.
It manages incrementally improved versions of the grammar.
It initiates runs where a grammar is applied to the corpus.
It monitors in progress runs by retrieving partial results.
It collects representative subsets of datasets for quick turnaround tests.
It identifies grammar changes within each run.
It graphs counts of gramatical elements.
It displays samples associated with gramatical elements.
it downloads datasets of extracted features.
Wiki Replacement
We found that projects typically involved hundreds of runs where dozens seemed worth keeping. These runs assumed a trial-and-error branching structure that was lost. The tool had a diff-on-hover feature partially recovered this experiment structure until grammars became too long to view without scrolling.
Our approach will be to capture significan branches for any project in a hypertext of concerns. The grammar on any given page will run to localize content of interest in the corpus. Subsequent pages will explore this local, possibly many different ways.
The webapp's drop-down for selecting datasets will be replaced by wiki pages.
The webapp's edit window will be replaced by Code plugins on experimental branches.
The webapp's run button will be replaced by the Parse plugin which clones itself as each run is initiated.
The webapp's grammar diff will be replaced by side-by-side viewing of branch pages.
The webapp's list of recent runs will be replaced by wiki pages.
The webapp's login dialog will be replaced by cluster credentials stored (and shared) in wiki pages.
The webapp's extracted feature downloads will be replaced by generated Data plugins feeding the wiki's client-side computation and visualization facilities.