acoli-repo / conll-transform

seamless conversion between different CoNLL/TSV dialects

conll-transform

seamless conversion between different CoNLL/TSV dialects

generates transformation workflows for CoNLL-RDF/Fintan (currently: bash scripts, upcoming: JSON configurations)

open issues:

• call with CoNLLCorpusManager JSON config file
• deprecate bash output OR provide proper escaping

Synopsis

  ./transform.sh [-help]
  ./transform.sh SRC TGT [OWL]
    -help list all supported formats
    SRC   source format
    TGT   target format
    OWL   CoNLL ontology, TTL format, 
	  defaults to ./conll-rdf/owl/conll.ttl

read CoNLL data from stdin, write to stdout transform from SRC format to TGT format according to OWL

Sample call

$ ./transform.sh CoNLL-99 CoNLL-03

Log:

  loading CoNLL-RDF ontology ./conll-rdf/owl/conll.ttl

  building bash script

  1. configure preprocessing
    warning: unsupported input encoding "iobEncoding"
    warning: unsupported input encoding "iobEncoding"
    warning: unsupported input encoding "iobEncoding"

  2. configure extraction

  3. configure update
  generalization: VCHUNK => CHUNK
  warning: no mapping for target format property NER
  warning: no mapping for source format property NCHUNK
  warning: no mapping for source format property CLAUSE

  4. configure formatter

  5. configure postprocessing
  warning: unsupported output encoding "iobEncoding"

  6. writing script

This generates

java -classpath bin org.acoli.conll.transform.Preprocessor -col " " | \
./run.sh CoNLLStreamExtractor http://ignore.me/ WORD POS NCHUNK CLAUSE VCHUNK | \
./run.sh CoNLLRDFUpdater -updates \
	"PREFIX conll: <http://ufal.mff.cuni.cz/conll2009-st/task-description.html#>
	 INSERT { ?a conll:CHUNK ?b } 
	 WHERE { ?a conll:VCHUNK ?b};" | \
./run.sh CoNLLRDFFormatter -conll WORD POS CHUNK NER | \
java -classpath bin org.acoli.conll.transform.Postprocessor -col " "