1. Features

[+wb]:  only the line boundary symbol #, automatically placed by the program at the end of a line.
  The # is underspecified for all other features (0 value).
[+segment]:  born by either real syllables or the dummy symbols LB2-LB5, explained below
[+accent]:  some degree of stress; in raw data files this is 2, 3, or 4.
[+rise]:  next syllable is stronger in stress.  Auto-assigned by the program ConvertStressesAndJuncturesToSingleSymbols.exe.
[+fall]:  next syllable is weaker in stress.  So [-rise,-fall] denotes a level stress sequence.
[+J2]:  This syllable ends a simplex word.
[+J3]:  This syllable ends a Clitic Group.
[+J4]:  This syllable ends a Phonological Phrase.
[+J5]:  This syllable ends an Intonational Phrase.
[+RealSyl]:  borne by real syllables (there are 10 or 11, depending on extrametricality)
[-RealSyl]:  borne by the four dummy symbols LB2, LB3, LB4, LB5.  These mean that the last syllable of the *preceding* line has
juncture level 2, 3, 4, or 5.  This juncture level is needed for some constraints, like inversion constraints.
Look at a sample training data file to see how this works.  

2. Symbols in input files

The symbols of the data files can be understood in terms of these features.  Thus the
first syllable of "Of" in "Of fairest creatures we desire increase" is

MwUuJ2 

which means:  

"in metrical Weak position" (Mw)
"unstressed" (U)
"first syllable of a rising stress sequence ("fair" is stronger)" (u for up)
"word final, but non-final in its Clitic Group"

Metrical Strong is Ms.
Stress is S
Falling stress is d (for down)

3. Making your own constraints

The essential trick is never to use any featural expression that is not in the file NaturalClassMemberships.txt.
If you use something not in this file, the program will crash.


