* Transfer Engine

  The transfer engine is written in c++.  It uses the Genkit feature unification package.  Genkit uses templates heavily and so is sensistive to g++ version.  I can only get it to compile with g++ 3.3, so that is what I use to compile all the code and the support libraries.  I don't recommend recompiling any of the support libraries unless you really have to.

Avenue/Transfer - main code files
 transfer.hpp
    Stores the class header for all of the transfer engine classes.

 transfer.cpp
    Code for the older version of the transfer process.  Deprecated but still functional.

 transfer3.cpp
    Code for the modern bottom-up, simaltaneous parse and transfer system.

 transfer-support.cpp
    Includes helper functions to translate files, produce lattices.

 transfer-decoder.cpp
    Include main decoder function, decodeFast, relies on data stored in rangeBeams

 transfermain.cpp
    Simple front end to transfer engine with main command to read command line arguments

 UnicodeTools.{hpp,cpp}
    Some simple code I wrote to handle upper and lower casing for UTF-8.  Probably better to convert to ICU at some point.

 language.cpp/hpp, chinese.cpp/hpp, english.cpp/hpp
    Some files I wrote to do number handling for English and Chinese.  Now handled by morph servers.

 Makefile
    Location of libraries and code, for Avenue and Barrow


* ANTLR files (first converted to c++)

Avenue/Transfer

 transfer.g 
   Reads in transfer grammar and lexicons and processes to internal format.
   Produces TransferGrammarLexer.{hpp,cpp}, TransferGrammarParser.{hpp,cpp}

 fstruct-cpp.g
   Reads in feature structures and processes to internal format.
   Produces FStructLexer.{hpp,cpp}, FStructParser.{hpp,cpp}

 parsetree-cpp.g 
   Reads in a parse tree as output by Stanford parser
   (or similar treebank format) and extracts constituent boundaries
   and types
   Produces ParseTreeLexer.{hpp,cpp}, ParseTreeParser.{hpp,cpp}


* Support libraries:

  These support libraries are compiled on both Avenue and Barrow in
 their 32-bit and 64-bit versions, respectively.  Some of the
 pathnames may vary due to this difference, check the Makefile.

GENKIT 
  Unification library, written by Ben Han, include UKernel, Toolkit
  Very sensitive to g++ version, currently requires 3.3
  /shared/code/genkit

ANTLR
  Similar to LEX, YACC, used for reading in grammars, lexicons, feature structures
  /shared/code/antlr-2.7.6/lib/cpp/src

SRILM
  SRI language modeling library
  /shared/code/srilm1.5/

SALM
  Suffix Array Language modeling toolkit
  /shared/code/SALM

STTK
  Calculate translation probabilities
  

* Scoring Scripts

  I've put all the common necessary code for doing BLEU and METEOR scoring in 
barrow:/usr8/eepeter/decoder/scoreutils.pl.  Use the method scoreFile($hypfile, $reference)
Just require this file, and call scoreFile


* Morphology

English
 Avenue/Transfer/English
  anamorphEnglish.pl - Relies on morpha from University of Sheffield
  genmorphEnglish.pl - Relies on morphg from University of Sheffield

Chinese
 barrow:/shared/code/segmenter/segserver5.pl


* Weights
 Avenue/Transfer/weights/optimizenbest
 
 - extracthyps.pl:  Extract training file from n-best output