oracle prune graph data under 3 1st-pass tuning conditions

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@449 ec762483-ff6d-05da-a07a-a48fb63a330f

35 files changed, 214 insertions, 0 deletions

diff --git a/report/prune_results/2pass-hdt b/report/prune_results/2pass-hdt
new file mode 100644
index 00000000..d2610931
--- /dev/null
+++ b/report/prune_results/2pass-hdt
@@ -0,0 +1,13 @@
+0.03	5.45
+0.05	6.72
+0.08	8.89
+0.12	11.51
+0.17	14.44
+0.25	17.11
+0.3	19.27
+0.4	20.20
+0.6	20.68
+0.8	20.53
+1	20.39
+2	20.51
+3	20.38
diff --git a/report/prune_results/2pass-hdt.pdf b/report/prune_results/2pass-hdt.pdf
new file mode 100644
index 00000000..6f3f43a0
--- /dev/null
+++ b/report/prune_results/2pass-hdt.pdf
diff --git a/report/prune_results/2pass-hdt.png b/report/prune_results/2pass-hdt.png
new file mode 100644
index 00000000..9c054fbb
--- /dev/null
+++ b/report/prune_results/2pass-hdt.png
diff --git a/report/prune_results/2pass-hdt0 b/report/prune_results/2pass-hdt0
new file mode 100644
index 00000000..1bc44f23
--- /dev/null
+++ b/report/prune_results/2pass-hdt0
@@ -0,0 +1,13 @@
+0.03	4.58
+0.05	6.68
+0.08	10.19
+0.12	13.87
+0.17	16.76
+0.25	19.52
+0.3	19.91
+0.4	20.37
+0.6	20.39
+0.8	20.46
+1	20.39
+2	20.41
+3	20.47
diff --git a/report/prune_results/2pass-hdt0.pdf b/report/prune_results/2pass-hdt0.pdf
new file mode 100644
index 00000000..70c5a65b
--- /dev/null
+++ b/report/prune_results/2pass-hdt0.pdf
diff --git a/report/prune_results/2pass-hdt0.png b/report/prune_results/2pass-hdt0.png
new file mode 100644
index 00000000..9d872af1
--- /dev/null
+++ b/report/prune_results/2pass-hdt0.png
diff --git a/report/prune_results/2pass-hdt1 b/report/prune_results/2pass-hdt1
new file mode 100644
index 00000000..13b7e8c0
--- /dev/null
+++ b/report/prune_results/2pass-hdt1
@@ -0,0 +1,13 @@
+0.03	12.09
+0.05	12.72
+0.08	13.64
+0.12	14.94
+0.17	15.93
+0.25	16.89
+0.3	17.60
+0.4	18.90
+0.6	19.59
+0.8	19.87
+1	20.03
+2	20.33
+3	20.43
diff --git a/report/prune_results/README b/report/prune_results/README
new file mode 100644
index 00000000..d448ce10
--- /dev/null
+++ b/report/prune_results/README
@@ -0,0 +1,21 @@
+tuned on urdu baselines (hiero) devtest, test on mt09
+
+hdt: single pass (no pruning of -LM forest).  these weights were used for the
+final +LM rescoring in all cases. (tune BLEU 22.2)
+
+2pass-hdt: 1st pass weights same as final weights (bad) (tune BLEU ???)
+
+2pass-hdt0: 1st pass weights tuned without any lm (same final weights as hdt). (tune BLEU 14.7)
+
+2pass-hdt1: 1st pass weights tuned with unigram lm (same final). (tune BLEU 16.2)
+
+data files / graphs: 2nd column is BLEU
+
+2pass-hdt - first column is beam alpha scaled per source word.
+
+times.2pass-hdt - first column is +LM rescoring time
+
+space.2pass-hdt - first column is portion of -LM forest edges kept
+
+It would be nice to graph these with the same scale, or preferably as 3 lines in
+the same graph.
diff --git a/report/prune_results/cdec-2pass-hdt.ini b/report/prune_results/cdec-2pass-hdt.ini
new file mode 100644
index 00000000..c0d034c3
--- /dev/null
+++ b/report/prune_results/cdec-2pass-hdt.ini
@@ -0,0 +1,9 @@
+cubepruning_pop_limit=200
+feature_function=LanguageModel /export/ws10smt/data/urdu-english/lm/u2e.en.lm.gz -o 3
+feature_function=WordPenalty
+feature_function=ArityPenalty
+add_pass_through_rules=true
+formalism=scfg
+grammar=test.grammar.gz
+weights=weights-hdt
+prelm_weights=weights-hdt
diff --git a/report/prune_results/cdec-2pass-hdt0.ini b/report/prune_results/cdec-2pass-hdt0.ini
new file mode 100644
index 00000000..177cf5b0
--- /dev/null
+++ b/report/prune_results/cdec-2pass-hdt0.ini
@@ -0,0 +1,9 @@
+cubepruning_pop_limit=200
+feature_function=LanguageModel /export/ws10smt/data/urdu-english/lm/u2e.en.lm.gz -o 3
+feature_function=WordPenalty
+feature_function=ArityPenalty
+add_pass_through_rules=true
+formalism=scfg
+grammar=test.grammar.gz
+weights=weights-hdt
+prelm_weights=weights-hdt0
diff --git a/report/prune_results/cdec-2pass-hdt1.ini b/report/prune_results/cdec-2pass-hdt1.ini
new file mode 100644
index 00000000..da8791c0
--- /dev/null
+++ b/report/prune_results/cdec-2pass-hdt1.ini
@@ -0,0 +1,10 @@
+cubepruning_pop_limit=200
+feature_function=LanguageModel lm.gz -o 3
+feature_function=LanguageModel lm.gz -o 1 -n Unigram
+feature_function=WordPenalty
+feature_function=ArityPenalty
+add_pass_through_rules=true
+formalism=scfg
+grammar=test.grammar.gz
+weights=weights-hdt
+prelm_weights=weights-hdt1
diff --git a/report/prune_results/cdec-hdt.ini b/report/prune_results/cdec-hdt.ini
new file mode 100644
index 00000000..ca00bc7e
--- /dev/null
+++ b/report/prune_results/cdec-hdt.ini
@@ -0,0 +1,8 @@
+cubepruning_pop_limit=200
+feature_function=LanguageModel lm.gz -o 3
+feature_function=WordPenalty
+feature_function=ArityPenalty
+add_pass_through_rules=true
+formalism=scfg
+grammar=test.grammar.gz
+weights=weights-hdt
diff --git a/report/prune_results/cdec-hdt1.ini b/report/prune_results/cdec-hdt1.ini
new file mode 100644
index 00000000..a58b7ba5
--- /dev/null
+++ b/report/prune_results/cdec-hdt1.ini
@@ -0,0 +1,8 @@
+cubepruning_pop_limit=200
+feature_function=LanguageModel lm.gz -o 1 -n Unigram -m 3
+feature_function=WordPenalty
+feature_function=ArityPenalty
+add_pass_through_rules=true
+formalism=scfg
+grammar=test.grammar.gz
+weights=weights-hdt1
diff --git a/report/prune_results/hdt b/report/prune_results/hdt
new file mode 100644
index 00000000..1926d429
--- /dev/null
+++ b/report/prune_results/hdt
@@ -0,0 +1 @@
+	20.42
diff --git a/report/prune_results/space.2pass-hdt b/report/prune_results/space.2pass-hdt
new file mode 100644
index 00000000..96347fd1
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt
@@ -0,0 +1,13 @@
+0.012619212257812	5.45
+0.015353377349888	6.72
+0.02327939994978	8.89
+0.04440807192522	11.51
+0.091294153035714	14.44
+0.201633255	17.11
+0.29858187198103	19.27
+0.4530871268415	20.20
+0.68171495345982	20.68
+0.81462782204241	20.53
+0.88406101526228	20.39
+0.9881657092634	20.51
+0.9980128671875	20.38
diff --git a/report/prune_results/space.2pass-hdt.pdf b/report/prune_results/space.2pass-hdt.pdf
new file mode 100644
index 00000000..d88acd12
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt.pdf
diff --git a/report/prune_results/space.2pass-hdt.png b/report/prune_results/space.2pass-hdt.png
new file mode 100644
index 00000000..79087d8b
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt.png
diff --git a/report/prune_results/space.2pass-hdt0 b/report/prune_results/space.2pass-hdt0
new file mode 100644
index 00000000..b5b833f9
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt0
@@ -0,0 +1,13 @@
+0	20.47
+0.0221153941406	4.58
+0.04510728694754	6.68
+0.11049838877232	10.19
+0.23648305700335	13.87
+0.4020261202009	16.76
+0.61663757265625	19.52
+0.71281294006697	19.91
+0.83940257511161	20.37
+0.9465962484375	20.39
+0.98112857600446	20.46
+0.9926947912946	20.39
+0.9992315530134	20.41
diff --git a/report/prune_results/space.2pass-hdt0.pdf b/report/prune_results/space.2pass-hdt0.pdf
new file mode 100644
index 00000000..d7426bbd
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt0.pdf
diff --git a/report/prune_results/space.2pass-hdt0.png b/report/prune_results/space.2pass-hdt0.png
new file mode 100644
index 00000000..8e778926
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt0.png
diff --git a/report/prune_results/space.2pass-hdt1 b/report/prune_results/space.2pass-hdt1
new file mode 100644
index 00000000..903a1d08
--- /dev/null
+++ b/report/prune_results/space.2pass-hdt1
@@ -0,0 +1,13 @@
+0.01020272126674	12.09
+0.010919479288504	12.72
+0.01218286062221	13.64
+0.014458555416295	14.94
+0.018661148814732	15.93
+0.02947725258929	16.89
+0.03930440498326	17.60
+0.0660309891183	18.90
+0.14408529348214	19.59
+0.24025580074777	19.87
+0.339429688816965	20.03
+0.70468050518973	20.33
+0.86776233962053	20.43
diff --git a/report/prune_results/space.hdt b/report/prune_results/space.hdt
new file mode 100644
index 00000000..6fe9efd6
--- /dev/null
+++ b/report/prune_results/space.hdt
@@ -0,0 +1 @@
+0	20.42
diff --git a/report/prune_results/space.hdt.png b/report/prune_results/space.hdt.png
new file mode 100644
index 00000000..bf478e4e
--- /dev/null
+++ b/report/prune_results/space.hdt.png
diff --git a/report/prune_results/time.2pass-hdt b/report/prune_results/time.2pass-hdt
new file mode 100644
index 00000000..377ab05a
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt
@@ -0,0 +1,13 @@
+0.03330357	5.45
+0.11414621	6.72
+0.33327009	8.89
+0.695401786	11.51
+1.181362	14.44
+1.49443638	17.11
+1.62612723	19.27
+1.75272321	20.20
+1.7993192	20.38
+2.04206194	20.39
+2.1030971	20.68
+2.15327567	20.53
+2.32144531	20.51
diff --git a/report/prune_results/time.2pass-hdt.pdf b/report/prune_results/time.2pass-hdt.pdf
new file mode 100644
index 00000000..305e3bbc
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt.pdf
diff --git a/report/prune_results/time.2pass-hdt.png b/report/prune_results/time.2pass-hdt.png
new file mode 100644
index 00000000..92545300
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt.png
diff --git a/report/prune_results/time.2pass-hdt0 b/report/prune_results/time.2pass-hdt0
new file mode 100644
index 00000000..fef5bdaf
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt0
@@ -0,0 +1,13 @@
+0.2642299	4.58
+0.629174107	6.68
+1.1253906	10.19
+1.17707031	20.47
+1.21284598	20.41
+1.45111049	13.87
+1.57308036	16.76
+1.72605469	20.39
+1.79328125	20.46
+1.87117188	19.52
+1.96085379	20.37
+1.98935826	19.91
+2.26236328	20.39
diff --git a/report/prune_results/time.2pass-hdt0.pdf b/report/prune_results/time.2pass-hdt0.pdf
new file mode 100644
index 00000000..715be382
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt0.pdf
diff --git a/report/prune_results/time.2pass-hdt0.png b/report/prune_results/time.2pass-hdt0.png
new file mode 100644
index 00000000..78418b6d
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt0.png
diff --git a/report/prune_results/time.2pass-hdt1 b/report/prune_results/time.2pass-hdt1
new file mode 100644
index 00000000..63e0ba86
--- /dev/null
+++ b/report/prune_results/time.2pass-hdt1
@@ -0,0 +1,13 @@
+0.00753348	12.09
+0.01848214	12.72
+0.04334821	13.64
+0.10146763	14.94
+0.19296317	15.93
+0.37776228	16.89
+0.49594308	17.60
+0.940814732	18.90
+1.43468192	19.59
+1.59580357	19.87
+1.77708147	20.03
+2.150558	20.43
+2.23821987	20.33
diff --git a/report/prune_results/time.hdt b/report/prune_results/time.hdt
new file mode 100644
index 00000000..cce55d29
--- /dev/null
+++ b/report/prune_results/time.hdt
@@ -0,0 +1 @@
+2.37764246	20.42
diff --git a/report/prune_results/time.hdt.png b/report/prune_results/time.hdt.png
new file mode 100644
index 00000000..6835d047
--- /dev/null
+++ b/report/prune_results/time.hdt.png
diff --git a/report/prune_results/weights-hdt b/report/prune_results/weights-hdt
new file mode 100644
index 00000000..dcc4167a
--- /dev/null
+++ b/report/prune_results/weights-hdt
@@ -0,0 +1,10 @@
+Arity_0 2.11777
+Arity_1 4.17369
+Arity_2 0.306923
+Glue -0.570606
+LanguageModel 3.63341
+PassThrough -0.67976718
+PhraseModel_0 -1.88489
+PhraseModel_1 -0.644993
+PhraseModel_2 -0.877253
+WordPenalty -7.6251
diff --git a/report/prune_results/weights-hdt0 b/report/prune_results/weights-hdt0
new file mode 100644
index 00000000..40a18324
--- /dev/null
+++ b/report/prune_results/weights-hdt0
@@ -0,0 +1,9 @@
+Arity_0 2.29931287103145
+Arity_1 2.95468395455627
+Arity_2 2.41711561162438
+Glue 0.69069383210885
+PassThrough -1.24872442206905
+PhraseModel_0 -0.397063761969712
+PhraseModel_1 -0.414233658695665
+PhraseModel_2 0.0509197848828125
+WordPenalty -1.87078476361445
diff --git a/report/prune_results/weights-hdt1 b/report/prune_results/weights-hdt1
new file mode 100644
index 00000000..12cf5436
--- /dev/null
+++ b/report/prune_results/weights-hdt1
@@ -0,0 +1,10 @@
+Arity_0 1.70741
+Arity_1 -0.68519
+Arity_2 1.14986
+Glue -3.126714
+PassThrough -0.83092
+PhraseModel_0 -1.94859
+PhraseModel_1 -0.147535
+PhraseModel_2 -1.61482
+Unigram 5.44106
+WordPenalty -8.92751