DocumentationParty

Morphological analyzer notes

Existing documentation:

• LkbSppp

• [http://svn.emmtee.net/trunk/upf/srg/freeling/README FreeLing SPPP README]

What's SPPP? In the same space as chart mapping, may become redundant. Predates chart mapping. Might also overlap with REPP, supported in LKB and soon in PET.

-> Send question to Rebecca: Will SPPP be part of extended support for PET?

LkbSppp

Current REPP processing does damage for English to the input text because TNT requires punctuation removed. That's lossy --- throwing spaces around a hyphen etc. Doesn't arise in terms of generation.

ERG generation -- some lisp code at the end provides some patching up like sentence-initial capitalization.

Ask Montse about generation What about Jacy? Chasen on the way out? BURGER very large fully inflected form dictionary

Chart mapping is fully supported in PET and planned for the LKB. But mostly for robustness over real phenomena real corpora. Other use is passing along morphological ambiguity to the parser. Montse could want this too in the LKB for Spanish. (por tanto] tiempo] v. [por [tanto tiempo)

Time involved in chart mapping is mostly in the rules to create more or fewer tokens.

No chart mapping in the LKB yet, so that takes us back to SPPP as the recommended connection in the parsing direction. Nothing in the generation direction yet.

Posssible MA/MS project:

• -- Hook up existing reversible morphophonological analyzer to existing grammar in the generation direction -- Do this in a way that it can be added to the customization system as an option.

Montse/Luis FreeLing might be good beta testers

OpenFST --- should be reversible SPRouT --- reversible?