Adapt ContextD triggers

[lcreteig]

There are only 23 negation triggers that were actually present in the (evaluation subset) of the DCC dataset used in the ConTextD paper. 495 more triggers also seem to be available at this link from the paper. (But note that the paper states they had 509 triggers in total...)

Potential issues with the original ContextD triggers

• Some have spaces at the end, others do not (might not matter much in practice).
• "uitsluiten" is both marked as poss (possible/speculation) and neg (negation).
• Is the woord trigger in "trigger gebeurtenis voor " supposed to be there?
• Is the trigger "neg niet " correct (and if so, what does it mean?)
• "(negatief)" appears to have backslash-escaped parentheses, but not sure about this
• Triggers have either a "pre"/forward scope or a "post"/backward scope. Though I imagine that some words (e.g. geen) should have both (see [Trigger scopes] below). In medspacy this can be done by assigning a "BIDIRECTIONAL" scope
• pseudo/termination triggers currently apply to only one modality (e.g. negations, not speculations). This might be intentional, but can imagine that some are universal.
• [ ] There is no "historical experiencer" category by default in medspacy; needs to be added

Trigger scopes

For the 23 negation triggers that are reported in the ContextD paper, I checked whether the scope as defined in the trigger list matches my intuition. There are a few where the scope could be changed:

Trigger	Original scope	Alternative scope
gedaald	pre/forward	post/backward
kan niet	pre/forward	bidirectional
leek niet	pre/forward	bidirectional
niet	pre/forward	bidirectional
onbekend	pre/forward	bidirectional
evenmin	pre/forward	post/backward
geen	pre/forward	bidirectional

[bramiozo]

Good to have this noted for possible improvements :+1: