RULES

RULE INPUT FORMATS

Rule types and formats (one rule per line in *.rules file)

GENERAL FORM

SampWeight-StepWeight-RuleType-[rule parameters]
  SampWeight = rule weight for purposes of stochastic gradient term sampling
  StepWeight = rule weight for purposes of stochastic gradient step size
  RuleType = type of rule (seed, sentincl, sentexcl, cl, ml, ...)
  [rule parameters] = rest of arguments vary depending on rule type


SPECIFIC RULES

[seed]
1-50-seed-2-0 = assign all occurrences of word 2 to topic 0
1-50-seed-2,7-0 = assign all occurrences of word 2 or 7 to topic 0
1-50-seed-2 7-0 = assign all occurrences of bigram 2 7 to topic 0

[sentincl]
100.0-100.0-sentincl-6-0 = allow topic 0 only if co-sentenence w/ topic 6

[sentexcl]
1.0-10.0-sentexcl-0-7 = dis-allow co-sentence topic 0 and topic 7

[cl]
1.0-10.0-cl-3-5 = word 3 and word 5 should not be assigned to the same topic

[ml]
1.0-10.0-ml-2-4 = word 2 and word 4 should be assigned to the same topic

[doc]
1.0-10.0-doc-1-0,1,2 = assign words in docs labeled 1 to topic 0,1, or 2