Simulation of Beeferman metric
1 = Propose none
2 = Propose all
3 = Random unknown # boundaries
4 = Random known # boundaries

5 = TextTiling with default parameters

My implementation of DotPlot, maximization, cosine sim, sentence based, stem, stopwords removed
6 = Normal
7 = Smoothing
8 = Noise reduction
9 = Thresholding
10 = Hard ranking 3x3 mask
11 = Hard ranking 5x5 mask
12 = Hard ranking 7x7 mask
13 = Hard ranking 9x9 mask
14 = Hard ranking 11x11 mask (i.e. C99)
15 = Hard ranking 13x13 mask
16 = Hard ranking 11x11 mask with Spread activation
17 = Hard ranking 11x11 mask, rank used as weight for similarity matrix

Block based DotPlot, maximization, dot density as sim
18 = Stem reiteration without stopword removal
19 = Stem reiteration with stopword removal
20 = Spread activation

Exact implementation of DotPlot
21 = Maximization
22 = Minimization

Segmenter
23 = Fixed distance model (has memory usage problems)
24 = Adaptive distance model

C99 (reimplementation of 10 with 11x11 mask)
25 = Known # boundaries
26 = Unknown # boundaries

TextTiling (Hearst's original implementation)
27 = Default parameters
28 = Parameters from Hearst's paper, k=6, w=20

More simulation for Beeferman metric
29 = Chop document into regular sized chunks to check whether
     the data is biased, since the mean segment size should
     be a constant (given infinite data). Better check, just in case.

Mask size testing with C99
30 = 1x1 mask, i.e. all rank=0, see if this will bias the evaluation
	 metric to a good score. Need to check to ensure the evaluation data
	 is reliable.
31 = 3x3 mask
32 = 5x5 mask
33 = 7x7 mask
34 = 9x9 mask
35 = 11x11 mask
36 = 13x13 mask
37 = 15x15 mask
38 = 17x17 mask

Min-Yen Kan's Segmenter (original implementation)
39 = Segmenter 1.6 with default databases and parameters

Further testing with C99
40 = Context vector frequencies weighted by entropy
     calculated from document term frequency
41 = Context vector frequencies weighted by inverse term
     frequency from document term frequency
42 = C99 with 3x3 mask, automatic termination
