Configuration

Structures

attribute1=value1
attribute2=value2
attribute3=value3
...

Attributes

1. maxIters: Maximum number of optimization iterations (default = 100).
2. numthreads: Number of parallel threads to run (default = 1). Running with more thread is faster, but requires more memory.
3. invSigmaSquare: Inverse of sigma squared (regularization parameter) (default = 1.0).
4. epsForConvergence: Convergence parameter (default = 0.001).
5. maxSegment: Maximum length of a segment (value = 1 or -1). If maxSegment = 1, the trained model is linear-CRF. If maxSegment = -1, the trained model is semi-CRF. The maximum length of the semi-CRF segments will be determined automatically (default = -1).

Examples

• punc.conf
• ref.conf