brendano’s gists

brendano / gist:893449

Created March 29, 2011 22:11

	acl_2006 applications
	acl_2006 asian_language_processing
	acl_2004 conversational_spoken_language_processing
	acl_2006 coreference
	acl_2004 coreference_and_anaphora
	acl_2005 corpus_annotation
	acl_2006 dialogue
	acl_2004 dialogue_systems
	acl_2005 discourse
	acl_2006 discourse

brendano / gist:868901

Created March 14, 2011 08:41

	~/tdf/exper % Rscript ../sigtest.r .test.

	*** Results with bootstrapped CI's
	f mae ci.conf ci.V2 ci.V3
	1 meta.test.last.prediction.4.000e-02 353.8108 0.95 314.0861 394.0680
	2 meta.test.single.prediction.4.000e-02 334.8699 0.95 297.5722 373.5800
	3 meta.test.singlenogap.prediction.4.000e-02 335.2703 0.95 299.4178 370.9729
	4 meta.test.ts.prediction.1.600e-03 335.3140 0.95 298.3792 370.9963
	5 text.test.last.prediction.2.000e-01 342.7077 0.95 305.0201 380.5768
	6 text.test.single.prediction.4.000e-02 295.0461 0.95 261.3029 329.7087

brendano / gist:868450

Created March 13, 2011 21:13

	dataset \| NBER \| ACL \|
	response \| log(#downloads) \| 1{ #citations >= 5 } \|
	model (GLM variant) \| gaussian/squared loss \| linear logistic loss \|
	\| aka linear regression \| aka MaxEnt \|
	interp of feat weight=1 \| e times more downloads \| e times higher prob of high-citation class \|
	metric1 \| MAE \| Acc (BalAcc?) \|
	metric2, maybe dont use \| tau \| tau \|

brendano / gist:841747

Created February 24, 2011 04:23

325    !! !!! !!!! !!!!! !!!!!! !!!!!!! !!!!!!!! !!!!!!!!! !!!!!!!.. !!.. !!.... !!: !!?? !" !' !. !... !: !? ", ". ": #1 #2 #2010 #39 #4 #8217 #ui #ww $1 $10 $100 $1000 $188 $2 $20 $200 $25 $32 $379 $5 && '" '' ($149 ($169 (( ((( (((((((((((((((((((((((((((((((( (: (= (@ (^_^) (¬_¬ )( ))) ): *)) ** **awwyyy *] +22 ," ,... -& -- --- ----- ------> ----> ---->>> ---> --> -6 -> -_- -__- -___- .! ." .' ., .. ... ..." .... ..... ...... ....... ........ ............ ...: ..: .: .?!! 0-1 0.00 00 04:45 09 1,000 1-0 1/2 10 10.27 10.4 10/26 10/27- 10/27/2010 10/30/10 100% 100,000 10093 101 101.1 106 107.5 109a 10:45 10:55 10¢ 11 11/01/10 11:30 12 12.99 1200 1221 13 13% 13.94 14:14 15 15.7 15/30 16 161 17 17% 175 1793 17:27 18 1895 18:1 1980 1995 2+3 2.0 2.3 2.5 2/3 20 20% 2008 2010 2011 2014 2020 21 2221 23 23.0 257 26 27 28 29 29.676 30 30% 300 31 31.1 33.1 330 35 360 3:00 4,900 4-8 4.25 40 40% 401 45% 465.00 48 5'1 50 500 5000 516 52 53 55 56.3 57 5o 6-8 6.05 6.95 60-80% 63 640 6501 67 6:30 6:40 70 7046614311 75 76 7

brendano / gist:841746

Created February 24, 2011 04:22

325    !! !!! !!!! !!!!! !!!!!! !!!!!!! !!!!!!!! !!!!!!!!! !!!!!!!.. !!.. !!.... !!: !!?? !" !' !. !... !: !? ", ". ": #1 #2 #2010 #39 #4 #8217 #ui #ww $1 $10 $100 $1000 $188 $2 $20 $200 $25 $32 $379 $5 && '" '' ($149 ($169 (( ((( (((((((((((((((((((((((((((((((( (: (= (@ (^_^) (¬_¬ )( ))) ): *)) ** **awwyyy *] +22 ," ,... -& -- --- ----- ------> ----> ---->>> ---> --> -6 -> -_- -__- -___- .! ." .' ., .. ... ..." .... ..... ...... ....... ........ ............ ...: ..: .: .?!! 0-1 0.00 00 04:45 09 1,000 1-0 1/2 10 10.27 10.4 10/26 10/27- 10/27/2010 10/30/10 100% 100,000 10093 101 101.1 106 107.5 109a 10:45 10:55 10¢ 11 11/01/10 11:30 12 12.99 1200 1221 13 13% 13.94 14:14 15 15.7 15/30 16 161 17 17% 175 1793 17:27 18 1895 18:1 1980 1995 2+3 2.0 2.3 2.5 2/3 20 20% 2008 2010 2011 2014 2020 21 2221 23 23.0 257 26 27 28 29 29.676 30 30% 300 31 31.1 33.1 330 35 360 3:00 4,900 4-8 4.25 40 40% 401 45% 465.00 48 5'1 50 500 5000 516 52 53 55 56.3 57 5o 6-8 6.05 6.95 60-80% 63 640 6501 67 6:30 6:40 70 7046614311 75 76 7

brendano / gist:836221

Created February 20, 2011 19:15

	Gold no_suffixes/model.pred ==> model_base/m.pred
	@ @ciaranyree @ @
	O it O O
	V was V V
	P on P P
	N football N N
	N wives N N
	, , , ,
	IMPROVE $ one O $
	P of P P

brendano / gist:836220

Created February 20, 2011 19:12

	Gold then model_base/m.pred then include_noah/model.pred
	@ @ciaranyree @ @
	O it O O
	V was V V
	P on P P
	N football N N
	N wives N N
	, , , ,
	REGRESS $ one $ O
	P of P P

brendano / gist:836211

Created February 20, 2011 18:59

	+-zsh:400> paste ../blitz2011/data/newsplit/dev.goldtags.tab no_suffixes/model.pred model_base/m.pred
	+-zsh:400> tabawk '!$1{print ""; next} {change=($4!=$6); improve=$2==$6; regress=$2==$4; print !change ? "" : improve ? "IMPROVE" : regress ? "REGRESS" : "", $2,$1,$4,$6}'
	@ @ciaranyree @ @
	O it O O
	V was V V
	P on P P
	N football N N
	N wives N N
	, , , ,
	IMPROVE $ one O $

brendano / gist:836210

Created February 20, 2011 18:58

	+-zsh:399> paste ../blitz2011/data/newsplit/dev.goldtags.tab no_suffixes/model.pred model_base/m.pred
	+-zsh:399> tabawk '!$1{print ""; next} {change=($4!=$6); improve=$2==$6; regress=$2==$4; print !change ? "" : improve ? "IMPROVE" : "REGRESS", $2,$1,$4,$6}'
	@ @ciaranyree @ @
	O it O O
	V was V V
	P on P P
	N football N N
	N wives N N
	, , , ,
	IMPROVE $ one O $

brendano / gist:836207

Created February 20, 2011 18:57

	+-zsh:398> paste ../blitz2011/data/newsplit/dev.goldtags.tab no_suffixes/model.pred model_base/m.pred
	+-zsh:398> tabawk '!$1{print ""; next} {change=($4!=$6); improve=$2==$6; print !change ? "" : improve ? "IMPROVE" : "REGRESS", $2,$1,$4,$6}'
	@ @ciaranyree @ @
	O it O O
	V was V V
	P on P P
	N football N N
	N wives N N
	, , , ,
	IMPROVE $ one O $

	dataset \| NBER \| ACL \|
	response \| log(#downloads) \| 1{ #citations >= 5 } \|
	model (GLM variant) \| gaussian/squared loss \| linear logistic loss \|
	\| aka linear regression \| aka MaxEnt \|
	interp of feat weight=1 \| e times more downloads \| e times higher prob of high-citation class \|
	metric1 \| MAE \| Acc (BalAcc?) \|
	metric2, maybe dont use \| tau \| tau \|

Brendan O'Connor brendano