This package is a distribution of the Neural Network for Semantic Extraction.
Please see the attached report and slides for more information on semantic labeling, as well as:
        http://ml.nec-labs.com/software.php?project=senna 
To run it on two demo sentences you type:

./testme

Alternatively you could type the contents of the testme script which are:

export LD_LIBRARY_PATH=torch/lib
lua senna.lua test.txt

where test.txt is the filename of a list of sentences, one per line.
The output is a sentence number, followed by the parts-of-speech and
the PropBank ARG-labels of that sentence with respect to any verbs found.
You should see output something like:

1
The cat sat on the mat.
The|DT cat|NN sat|VBD on|IN the|DT mat|NN .|.
The|ARG1 cat|ARG1 sat|rel on|ARG2 the|ARG2 mat|ARG2 .|-

2
NEC can do semantic extraction with the aid of this algorithm.
NEC|NNP can|MD do|VB semantic|JJ extraction|NN with|IN the|DT aid|NN of|IN this|DT algorithm|NN .|.
NEC|ARG0 can|ARGM-MOD do|rel semantic|ARG1 extraction|ARG1 with|ARGM-MNR the|ARGM-MNR aid|ARGM-MNR of|ARGM-MNR this|ARGM-MNR algorithm|ARGM-MNR .|-

===Included Files=============

This package requires some libraries, which we've also included:

lua	        - Lua scripting language required to run SENNA


senna.luac	- The main script for running semantic extraction
data/*          - Data files for the trained Neural Network model
torch/*         - Torch 5 library files for Lua, 
		- [underlying code for Neural Networks]. 
		- A full installation of Torch can be found at:
		-    http://sourceforge.net/projects/torch5 


Here are descriptions of the types of tags.

Part-of-Speech tags:
====================
The descriptions of the tags can be found in the attached file POS_tags.htm

Semantic tags.
==============

 1) The argument label is one of {rel, ARGA, ARGM} + { ARG0, ARG1, ARG2,
   ... }.  The argument labels correspond to the argument labels in the frame 
   descriptions from the PropBank database.  
   ARGA is used for causative agents, ARGM for
   adjuncts of various sorts, and 'rel' refers to the surface string of
   the verb. Typically ARG0 is an actor and ARG1 is an object, 
   although if the verb is something which a subject affects themselves,
   such as "sat" you will typically see ARG1 as the actor. 

 2) ARGM modifier labels can have optional extra arguments.
   Labeled features follow:

   EXT - extent
   DIR - direction
   LOC - location
   TMP - temporal
   REC - reciprocal
   PRD - predication
   NEG - negation
   MOD - modal
   ADV - adverbial
   MNR - manner
   CAU - cause
   PNC - purpose not cause.
   DIS - discourse