FrameNet Japanese MT Project

This directory contains resources developed by the Japanese machine translation project in the NSF funded Framenet online effort.

The directories and their contents are described.

1. bin: Tool used in Japanese frames creation effort. Described in doc directory and in http://bulba.sdsu.edu/enoue/vgrep_help.html http://bulba.sdsu.edu/enoue/vgrep_manual.html
2. doc: Brief manual style documentation of the tools in bin.
3. english_frames: Hand created English crime frames, elaborated for inferencing.
4. english_lexicon: English lexicon built from framenet, passives undone, preposition identities added. Separated into frame specific files, one per frame.

   Passive undoing. In contrast to the usual representation, valence info is a 4-tuple consisting of frame element, Syntactic category, surface grammatical function and logical grammatical function (the new one). For example the verb "cloak" in the Adornment frame has the following two lex entries:

   1. instantiated_lex_entry('cloak', 'cloak 1', 'V', features:[], 'Adornment', [('theme', 'NP', 'Ext', 'Ext'), ('location', 'NP', 'Obj', 'Obj')]).
   2. instantiated_lex_entry('cloak', 'cloak 2', 'V', features:[], 'Adornment', [('location', 'NP', 'Ext', 'Obj'), ('theme', 'pp(by)', 'Comp', '')]).
   The location as subject occurs only in passives, and this is shown by the fact that when it shows up with the external grammatical function, it is always a logical object.
5. format: Tools for the english lexicon creation effort. The top level script is mk_lex.prl.
6. frames_to_kado: Automatically generated connections between kado ids and frames. Method: Find all the kado ids that correspond to translations of the English predicates from a given frame. Choose the ids that come up most often.
7. japanese_frames: Frames and lex entries for japanese frame building effort.