Add support for RST parsing
enoriega opened this issue · 4 comments
It would be very useful to expose discourse parsing.
Can you clarify a bit?
Dependency parses are included for each Sentence
contained in the Document
returned by a call to an annotate
endpoint.
Ah, I see the issue name has RST. Do you mean discourse parsing?
Yeah, sorry
processors-server
keeps two processor instances in memory: FastNLPProcessor
and BioNLPProcessor
. When initializing both using the withDiscourse
parameter, memory consumption exceeds 3GB (I think the server probably needs 5GB+ to function when discourse is enabled).
Because that sort of memory requirement is prohibitive for many users, maybe the solution is to make this aspect of the server configurable, so that one can easily run the server with or without support for discourse parsing.