From 6607203eb2188e6fee98dd10b2874b012caaa9df Mon Sep 17 00:00:00 2001
From: Franck Dary <franck.dary@lis-lab.fr>
Date: Wed, 6 Jul 2022 16:41:24 +0200
Subject: [PATCH] Added scripts to launch experiments from my thesis
---
UD_any/thesis/chapter2/experimentBatch.py | 298 ++++++++++++++++++++++
UD_any/thesis/chapter3/machinesBatch.py | 32 +++
UD_any/thesis/chapter3/useEntropy.r | 167 ++++++++++++
3 files changed, 497 insertions(+)
create mode 100644 UD_any/thesis/chapter2/experimentBatch.py
create mode 100644 UD_any/thesis/chapter3/machinesBatch.py
create mode 100644 UD_any/thesis/chapter3/useEntropy.r
diff --git a/UD_any/thesis/chapter2/experimentBatch.py b/UD_any/thesis/chapter2/experimentBatch.py
new file mode 100644
index 0000000..a5ba176
--- /dev/null
+++ b/UD_any/thesis/chapter2/experimentBatch.py
@@ -0,0 +1,298 @@
+# To untrack the file do : git update-index --skip-worktree batches.py
+
+class T :
+ def reset() :
+ T.dropout = "InputDropout : 0.5"
+ T.mlp = "MLP : {3200 0.4 1600 0.4}"
+ T.loss = "Loss : crossentropy"
+ T.optim = "Optimizer : Adagrad {0.01 0.000001 0 0.0000000001}"
+
+ T.focusedBuffer = "0"
+ T.focusedStack = ""
+ T.contextualWindow = "-10 0"
+ T.contextualWindowSeq = "-10 10"
+ T.contextualTargets = "b.-3 b.-2 b.-1 b.0 s.0 s.1 s.2 b.0.0 s.0.0 s.0.-1 s.1.0 s.1.-1 s.2.0 s.2.-1"
+ T.contextualTargetsSeq = "b.-3 b.-2 b.-1 b.0 b.1 b.2 s.0 s.1 s.2 b.0.0 s.0.0 s.0.-1 s.1.0 s.1.-1 s.2.0 s.2.-1"
+ T.contextualTargetsFlat = "b.-3 b.-2 b.-1 b.0"
+ T.contextTargets = "b.-3 b.-2 b.-1 b.0 s.0 s.1 s.2 b.0.0 s.0.0 s.0.-1 s.1.0 s.1.-1 s.2.0 s.2.-1"
+ T.contextTargetsSeq = "b.-3 b.-2 b.-1 b.0 b.1 b.2 s.0 s.1 s.2 b.0.0 s.0.0 s.0.-1 s.1.0 s.1.-1 s.2.0 s.2.-1"
+ T.contextTargetsFlat = "b.-3 b.-2 b.-1 b.0"
+ T.rawInputLeft = "5"
+ T.rawInputRight = "10"
+ T.historyNb = "10"
+ T.historyMineNb = "2"
+
+ T.lstmOpt = "1 1 0.0 1"
+ T.lstmIn = "128"
+ T.lstmOut = "64"
+ T.stateNameSize = "64"
+
+ T.prefixSize = "5"
+ T.suffixSize = "5"
+
+ T.prefixBuffer = "0"
+ T.suffixBuffer = "0"
+ T.prefixStack = ""
+ T.suffixStack = ""
+
+ T.distThreshold = "15"
+
+ def refresh() :
+ T.templateIncr1Class = """Name : Tokenizer, Tagger, Morpho, Lemmatizer, Parser and Segmenter Machine
+Classifier : tokeparser
+{
+ Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_whole.ts lemmatizer_rules,data/lemmatizer_rules.ts lemmatizer_case,data/lemmatizer_case.ts parser,data/parser_eager_rel_strict.ts segmenter,data/segmenter.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Splitwords : data/splitwords.ts
+Predictions : ID FORM UPOS FEATS LEMMA HEAD DEPREL EOS
+Strategy
+{
+ Block : End{cannotMove}
+ tokenizer tagger ENDWORD 0
+ tokenizer tagger SPLIT 0
+ tokenizer tokenizer * 0
+ tagger morpho * 0
+ morpho lemmatizer_rules * 0
+ lemmatizer_rules lemmatizer_case * 0
+ lemmatizer_case parser * 0
+ parser segmenter eager_SHIFT 0
+ parser segmenter eager_RIGHT_rel 0
+ parser parser * 0
+ segmenter tokenizer * 1
+}
+"""%(T.dropout, T.mlp, T.optim, T.loss)
+
+ T.templateSeq1Class = """Name : Tokenizer, Tagger, Morpho, Lemmatizer, Parser and Segmenter Machine
+Classifier : tokeparser
+{
+ Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_whole.ts lemmatizer_rules,data/lemmatizer_rules.ts lemmatizer_case,data/lemmatizer_case.ts parser,data/parser_eager_rel_strict.ts segmenter,data/segmenter.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Splitwords : data/splitwords.ts
+Predictions : ID FORM UPOS FEATS LEMMA HEAD DEPREL EOS
+Strategy
+{
+ Block : End{cannotMove}
+ tokenizer tokenizer ENDWORD 1
+ tokenizer tokenizer SPLIT 1
+ tokenizer tokenizer * 0
+ Block : End{cannotMove}
+ tagger tagger * 1
+ Block : End{cannotMove}
+ morpho morpho * 1
+ Block : End{cannotMove}
+ lemmatizer_rules lemmatizer_case * 0
+ lemmatizer_case lemmatizer_rules * 1
+ Block : End{cannotMove}
+ parser segmenter eager_SHIFT 0
+ parser segmenter eager_RIGHT_rel 0
+ parser parser * 0
+ segmenter parser * 1
+}
+"""%(T.dropout, T.mlp, T.optim, T.loss)
+
+ T.templateIncrNClass = """Name : Tokenizer, Tagger, Morpho, Lemmatizer, Parser and Segmenter Machine
+Classifier : tokenizer
+{
+ Transitions : {tokenizer,data/tokenizer.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Classifier : tagger
+{
+ Transitions : {tagger,data/tagger.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Classifier : morpho
+{
+ Transitions : {morpho,data/morpho_whole.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Classifier : lemmatizer
+{
+ Transitions : {lemmatizer_rules,data/lemmatizer_rules.ts lemmatizer_case,data/lemmatizer_case.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Classifier : parser
+{
+ Transitions : {parser,data/parser_eager_rel_strict.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Classifier : segmenter
+{
+ Transitions : {segmenter,data/segmenter.ts}
+ LossMultiplier : {}
+ Network type : Modular
+ %%s
+ %s
+ %s
+ End
+ %s
+ Type : classification
+ %s
+}
+Splitwords : data/splitwords.ts
+Predictions : ID FORM UPOS FEATS LEMMA HEAD DEPREL EOS
+Strategy
+{
+ Block : End{cannotMove}
+ tokenizer tagger ENDWORD 0
+ tokenizer tagger SPLIT 0
+ tokenizer tokenizer * 0
+ tagger morpho * 0
+ morpho lemmatizer_rules * 0
+ lemmatizer_rules lemmatizer_case * 0
+ lemmatizer_case parser * 0
+ parser segmenter eager_SHIFT 0
+ parser segmenter eager_RIGHT_rel 0
+ parser parser * 0
+ segmenter tokenizer * 1
+}
+"""%(T.dropout, T.mlp, T.optim, T.loss, T.dropout, T.mlp, T.optim, T.loss, T.dropout, T.mlp, T.optim, T.loss, T.dropout, T.mlp, T.optim, T.loss, T.dropout, T.mlp, T.optim, T.loss, T.dropout, T.mlp, T.optim, T.loss)
+
+ w2v = "FORM,../../cc.fr.300.vec"
+ T.words = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{FORM,data/FORM.w2v}"%(T.contextualTargets,"FORM",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.wordsSeq = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{FORM,data/FORM.w2v}"%(T.contextualTargetsSeq,"FORM",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.wordsFlat = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{FORM,data/FORM.w2v}"%(T.contextualTargetsFlat,"FORM",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.wordsPretrained = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{%s}"%(T.contextualTargets,"FORM",T.lstmOpt,300,T.lstmOut, w2v)
+ T.wordsSeqPretrained = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{%s}"%(T.contextualTargetsSeq,"FORM",T.lstmOpt,300,T.lstmOut, w2v)
+ T.wordsFlatPretrained = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{%s}"%(T.contextualTargetsFlat,"FORM",T.lstmOpt,300,T.lstmOut, w2v)
+ T.context = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.contextTargets,"ID EOS UPOS FEATS DEPREL",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.contextSeq = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.contextTargetsSeq,"ID EOS UPOS FEATS DEPREL",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.contextID = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.contextTargetsFlat,"ID",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.contextTagger = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.contextTargetsFlat,"ID UPOS",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.contextMorpho = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.contextTargetsFlat,"ID UPOS FEATS",T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.contextParser = "Context : Targets{%s} Columns{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.contextTargets,"ID UPOS FEATS DEPREL",T.lstmOpt,T.lstmIn,T.lstmOut)
+
+ T.prefix = "Focused : Column{prefix%s:FORM} NbElem{%s} Buffer{%s} Stack{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.prefixSize,T.prefixSize,T.prefixBuffer,T.prefixStack,T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.suffix = "Focused : Column{suffix%s:FORM} NbElem{%s} Buffer{%s} Stack{%s} LSTM{%s} In{%s} Out{%s} w2v{}"%(T.suffixSize,T.suffixSize,T.suffixBuffer,T.suffixStack,T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.rawInput5_10 = "RawInput : Left{5} Right{10} LSTM{%s} In{%s} Out{%s}"%(T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.rawInput5_5 = "RawInput : Left{5} Right{5} LSTM{%s} In{%s} Out{%s}"%(T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.rawInput5_15 = "RawInput : Left{5} Right{15} LSTM{%s} In{%s} Out{%s}"%(T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.rawInput5_2 = "RawInput : Left{5} Right{2} LSTM{%s} In{%s} Out{%s}"%(T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.history = "History : NbElem{%s} LSTM{%s} In{%s} Out{%s}"%(T.historyNb,T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.historyMine = "HistoryMine : NbElem{%s} LSTM{%s} In{%s} Out{%s}"%(T.historyMineNb,T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.splitTrans = "SplitTrans : LSTM{%s} In{%s} Out{%s}"%(T.lstmOpt,T.lstmIn,T.lstmOut)
+ T.stateName = "StateName : Out{%s}"%(T.stateNameSize)
+ T.dist = "Distance : FromBuffer{} FromStack{0 1 2} ToBuffer{0} ToStack{} Threshold{%s} LSTM{%s} In{%s} Out{%s}"%(T.distThreshold,T.lstmOpt,T.lstmIn,T.lstmOut)
+
+T.machines = []
+T.reset()
+T.refresh()
+
+T.machines.append(["tokeparser_incr_1class_window5_10", T.templateIncr1Class%"\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.stateName,T.dist])])
+T.machines.append(["tokeparser_seq_1class_window5_10", T.templateSeq1Class%"\n ".join([T.wordsSeq,T.contextSeq,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.stateName,T.dist])])
+
+T.machines.append(["tokeparser_incr_Nclass_sup_window5_10", T.templateIncrNClass%("\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]))])
+T.machines.append(["tokeparser_incr_Nclass_sup_window5_5", T.templateIncrNClass%("\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]))])
+T.machines.append(["tokeparser_incr_Nclass_sup_window5_15", T.templateIncrNClass%("\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]))])
+T.machines.append(["tokeparser_incr_Nclass_sup_window5_2", T.templateIncrNClass%("\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]))])
+
+T.machines.append(["tokeparser_incr_Nclass_nosup_window5_10", T.templateIncrNClass%(
+"\n ".join([T.wordsFlat,T.contextID,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Tokenizer
+"\n ".join([T.wordsFlat,T.contextTagger,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Tagger
+"\n ".join([T.wordsFlat,T.contextMorpho,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Morpho
+"\n ".join([T.wordsFlat,T.contextMorpho,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Lemmatizer
+"\n ".join([T.words,T.contextParser,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans,T.dist]), # Parser
+"\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]))]) # Segmenter
+
+
+T.machines.append(["tokeparserPre_incr_1class_window5_10", T.templateIncr1Class%"\n ".join([T.wordsPretrained,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.stateName,T.dist])])
+T.machines.append(["tokeparserPre_seq_1class_window5_10", T.templateSeq1Class%"\n ".join([T.wordsSeqPretrained,T.contextSeq,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.stateName,T.dist])])
+
+T.machines.append(["tokeparserPre_incr_Nclass_sup_window5_10", T.templateIncrNClass%("\n ".join([T.wordsPretrained,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]))])
+T.machines.append(["tokeparserPre_incr_Nclass_sup_window5_5", T.templateIncrNClass%("\n ".join([T.wordsPretrained,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_5,T.history,T.splitTrans,T.dist]))])
+T.machines.append(["tokeparserPre_incr_Nclass_sup_window5_15", T.templateIncrNClass%("\n ".join([T.wordsPretrained,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_15,T.history,T.splitTrans,T.dist]))])
+T.machines.append(["tokeparserPre_incr_Nclass_sup_window5_2", T.templateIncrNClass%("\n ".join([T.wordsPretrained,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]), "\n ".join([T.words,T.context,T.prefix,T.suffix,T.rawInput5_2,T.history,T.splitTrans,T.dist]))])
+
+T.machines.append(["tokeparserPre_incr_Nclass_nosup_window5_10", T.templateIncrNClass%(
+"\n ".join([T.wordsFlatPretrained,T.contextID,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Tokenizer
+"\n ".join([T.wordsFlatPretrained,T.contextTagger,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Tagger
+"\n ".join([T.wordsFlatPretrained,T.contextMorpho,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Morpho
+"\n ".join([T.wordsFlatPretrained,T.contextMorpho,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans]), # Lemmatizer
+"\n ".join([T.wordsPretrained,T.contextParser,T.prefix,T.suffix,T.rawInput5_10,T.historyMine,T.splitTrans,T.dist]), # Parser
+"\n ".join([T.wordsPretrained,T.context,T.prefix,T.suffix,T.rawInput5_10,T.history,T.splitTrans,T.dist]))]) # Segmenter
+
+
+templatesExperiments = [
+ {
+ 'mode' : 'txt',
+ 'expName' : '%s'%machine[0],
+ 'template' : 'templates/tokeparser_incr',
+ 'arguments' : '-n 40 --lockPretrained --devScore --trainStrategy 0,ExtractGold,ResetParameters:4,ExtractDynamic,Save --machine \"%s\"'%machine[1],
+ 'pretrained' : 'FORM',
+ 'evalArguments' : ''
+ } for machine in T.machines
+]
+
+langs = [
+ "UD_French-GSD_0",
+ "UD_French-GSD_1",
+ "UD_French-GSD_2",
+ "UD_French-GSD_3",
+ "UD_French-GSD_4",
+ "UD_French-GSD_5",
+ "UD_French-GSD_6",
+ "UD_French-GSD_7",
+ "UD_French-GSD_8",
+ "UD_French-GSD_9",
+]
+
+repRange = [0]
+
diff --git a/UD_any/thesis/chapter3/machinesBatch.py b/UD_any/thesis/chapter3/machinesBatch.py
new file mode 100644
index 0000000..7928d95
--- /dev/null
+++ b/UD_any/thesis/chapter3/machinesBatch.py
@@ -0,0 +1,32 @@
+# To untrack the file do : git update-index --skip-worktree batches.py
+
+tagExpPre = [
+ {
+ 'mode' : 'tsv',
+ 'expName' : '%s'%exp,
+ 'template' : 'templates/'+exp,
+ 'arguments' : '-n 20 --trainStrategy 0,ExtractGold,ResetParameters:4,ExtractDynamic,Save --devScore --lockPretrained',
+ 'pretrained' : '',
+ 'evalArguments' : ''
+ } for (exp) in [("tagparser_incr"), ("tagparser_seq")]
+]
+
+tagExpNoPre = [
+ {
+ 'mode' : 'tsv',
+ 'expName' : '%s'%exp,
+ 'template' : 'templates/'+exp,
+ 'arguments' : '-n 20 --trainStrategy 0,ExtractGold,ResetParameters:4,ExtractDynamic,Save --devScore',
+ 'pretrained' : 'FORM',
+ 'evalArguments' : ''
+ } for (exp) in [("tagparser_incr_nopretrained"), ("tagparser_seq_nopretrained")]
+]
+
+templatesExperiments = tagExpPre + tagExpNoPre
+
+langs = [
+ "UD_English-EWT",
+]
+
+repRange = [0]
+
diff --git a/UD_any/thesis/chapter3/useEntropy.r b/UD_any/thesis/chapter3/useEntropy.r
new file mode 100644
index 0000000..ec39b02
--- /dev/null
+++ b/UD_any/thesis/chapter3/useEntropy.r
@@ -0,0 +1,167 @@
+## Opening useful libraries
+
+library(data.table)
+library (dplyr) # Library for ddply
+library(lmerTest) # pvalue
+library (lme4)
+library(sjmisc) #str_contains
+library(piecewiseSEM) # R squared
+library(ggplot2)
+
+setwd("/home/franck/oculometry/provoNoTok")
+
+################################################################################################
+rescaleVariables = function(dataset)
+{
+ dataset$FREQ = scale(log(dataset$FREQ))
+ dataset$WD_LEN = scale(dataset$WD_LEN, center = TRUE, scale = TRUE)
+
+ for (colName in colnames(dataset))
+ {
+ if (colName != "SENT_ID" & (str_contains(colName, "ENT_") | str_contains(colName, "SUR_")))
+ {
+ dataset[colName] = scale(dataset[colName], center = TRUE, scale = TRUE)
+ }
+ }
+
+ return(dataset)
+}
+################################################################################################
+
+################################################################################################
+computeCorrelation = function(dataFile)
+{
+ data = na.omit(read.table(dataFile, header=T, sep="\t", quote="", nrows=-1))
+ data = subset(data, FIRST_TIME != 0)
+ data = rescaleVariables(data)
+
+ print(paste("CORR ", dataFile, "FREQ"))
+ print(cor(data$ENT_CUR_MEAN_TAGGER, data$FREQ), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_MORPHO, data$FREQ), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_LEMMATIZER_RULES, data$FREQ), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_PARSER, data$FREQ), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_SEGMENTER, data$FREQ), method="pearson")
+ print("")
+
+ print(paste("CORR ", dataFile, "WD_LEN"))
+ print(cor(data$ENT_CUR_MEAN_TAGGER, data$WD_LEN), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_MORPHO, data$WD_LEN), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_LEMMATIZER_RULES, data$WD_LEN), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_PARSER, data$WD_LEN), method="pearson")
+ print(cor(data$ENT_CUR_MEAN_SEGMENTER, data$WD_LEN), method="pearson")
+ print("")
+}
+################################################################################################
+
+
+################################################################################################
+globalSummary = function(targetCol, dataFile)
+{
+ data = na.omit(read.table(dataFile, header=T, sep="\t", quote="", nrows=-1))
+ data = subset(data, FIRST_TIME != 0)
+ data = rescaleVariables(data)
+
+ sink(paste(dataFile, targetCol, "analysis", sep="."), append=T)
+
+ vanillaModel = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + (1 | TEXT_ID/SUBJECT)",sep="")), REML=FALSE, data=data)
+
+ print("#SUMMARY")
+
+ for (ent in c("ENT", "SUR"))
+ for (meth in c("CUR"))
+ for (mod in c("ADD"))
+ for (level in c("TAGGER", "MORPHO", "LEMMATIZER_RULES", "SEGMENTER"))
+ {
+ measure = paste(ent,meth,mod,level,sep="_")
+ formulaStr = paste(targetCol," ~ FREQ + WD_LEN + ",measure," + (1 | TEXT_ID/SUBJECT)",sep="")
+ model = lmer(formula(formulaStr), REML=FALSE, data=data)
+ print(summary(model))
+ }
+
+ for (ent in c("ENT", "SUR"))
+ for (meth in c("CUR", "ATT", "TGT"))
+ for (mod in c("ADD", "MEAN", "MAX"))
+ for (level in c("PARSER"))
+ {
+ measure = paste(ent,meth,mod,level,sep="_")
+ formulaStr = paste(targetCol," ~ FREQ + WD_LEN + ",measure," + (1 | TEXT_ID/SUBJECT)",sep="")
+ model = lmer(formula(formulaStr), REML=FALSE, data=data)
+ print(summary(model))
+ }
+
+ sink()
+ print(paste("SUMMARY", targetCol, dataFile))
+}
+################################################################################################
+
+################################################################################################
+globalAnova = function(targetCol, dataFile)
+{
+ data = na.omit(read.table(dataFile, header=T, sep="\t", quote="", nrows=-1))
+ data = subset(data, FIRST_TIME != 0)
+ data = rescaleVariables(data)
+
+ sink(paste(dataFile, targetCol, "analysis", sep="."))
+
+ vanillaModel = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + (1 | TEXT_ID/SUBJECT)",sep="")), REML=FALSE, data=data)
+
+ EntCurAdd = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_CUR_ADD_TAGGER + ENT_CUR_ADD_MORPHO + ENT_CUR_ADD_LEMMATIZER_RULES + ENT_CUR_ADD_PARSER + ENT_CUR_ADD_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ EntCurMean = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_CUR_MEAN_TAGGER + ENT_CUR_MEAN_MORPHO + ENT_CUR_MEAN_LEMMATIZER_RULES + ENT_CUR_MEAN_PARSER + ENT_CUR_MEAN_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ EntCurMax = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_CUR_MAX_TAGGER + ENT_CUR_MAX_MORPHO + ENT_CUR_MAX_LEMMATIZER_RULES + ENT_CUR_MAX_PARSER + ENT_CUR_MAX_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+
+ EntAttAdd = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_ATT_ADD_TAGGER + ENT_ATT_ADD_MORPHO + ENT_ATT_ADD_LEMMATIZER_RULES + ENT_ATT_ADD_PARSER + ENT_ATT_ADD_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ EntAttMean = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_ATT_MEAN_TAGGER + ENT_ATT_MEAN_MORPHO + ENT_ATT_MEAN_LEMMATIZER_RULES + ENT_ATT_MEAN_PARSER + ENT_ATT_MEAN_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ EntAttMax = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_ATT_MAX_TAGGER + ENT_ATT_MAX_MORPHO + ENT_ATT_MAX_LEMMATIZER_RULES + ENT_ATT_MAX_PARSER + ENT_ATT_MAX_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+
+ EntTgtAdd = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_TGT_ADD_TAGGER + ENT_TGT_ADD_MORPHO + ENT_TGT_ADD_LEMMATIZER_RULES + ENT_TGT_ADD_PARSER + ENT_TGT_ADD_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ EntTgtMean = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_TGT_MEAN_TAGGER + ENT_TGT_MEAN_MORPHO + ENT_TGT_MEAN_LEMMATIZER_RULES + ENT_TGT_MEAN_PARSER + ENT_TGT_MEAN_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ EntTgtMax = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + ENT_TGT_MAX_TAGGER + ENT_TGT_MAX_MORPHO + ENT_TGT_MAX_LEMMATIZER_RULES + ENT_TGT_MAX_PARSER + ENT_TGT_MAX_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+
+ SurCurAdd = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_CUR_ADD_TAGGER + SUR_CUR_ADD_MORPHO + SUR_CUR_ADD_LEMMATIZER_RULES + SUR_CUR_ADD_PARSER + SUR_CUR_ADD_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ SurCurMean = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_CUR_MEAN_TAGGER + SUR_CUR_MEAN_MORPHO + SUR_CUR_MEAN_LEMMATIZER_RULES + SUR_CUR_MEAN_PARSER + SUR_CUR_MEAN_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ SurCurMax = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_CUR_MAX_TAGGER + SUR_CUR_MAX_MORPHO + SUR_CUR_MAX_LEMMATIZER_RULES + SUR_CUR_MAX_PARSER + SUR_CUR_MAX_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+
+ SurAttAdd = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_ATT_ADD_TAGGER + SUR_ATT_ADD_MORPHO + SUR_ATT_ADD_LEMMATIZER_RULES + SUR_ATT_ADD_PARSER + SUR_ATT_ADD_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ SurAttMean = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_ATT_MEAN_TAGGER + SUR_ATT_MEAN_MORPHO + SUR_ATT_MEAN_LEMMATIZER_RULES + SUR_ATT_MEAN_PARSER + SUR_ATT_MEAN_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ SurAttMax = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_ATT_MAX_TAGGER + SUR_ATT_MAX_MORPHO + SUR_ATT_MAX_LEMMATIZER_RULES + SUR_ATT_MAX_PARSER + SUR_ATT_MAX_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+
+ SurTgtAdd = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_TGT_ADD_TAGGER + SUR_TGT_ADD_MORPHO + SUR_TGT_ADD_LEMMATIZER_RULES + SUR_TGT_ADD_PARSER + SUR_TGT_ADD_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ SurTgtMean = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_TGT_MEAN_TAGGER + SUR_TGT_MEAN_MORPHO + SUR_TGT_MEAN_LEMMATIZER_RULES + SUR_TGT_MEAN_PARSER + SUR_TGT_MEAN_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+ SurTgtMax = lmer(formula(paste(targetCol," ~ FREQ + WD_LEN + SUR_TGT_MAX_TAGGER + SUR_TGT_MAX_MORPHO + SUR_TGT_MAX_LEMMATIZER_RULES + SUR_TGT_MAX_PARSER + SUR_TGT_MAX_SEGMENTER + (1 | TEXT_ID/SUBJECT)",sep="")) , REML=FALSE, data=data)
+
+ print("#ANOVA")
+ print(anova(vanillaModel, EntCurAdd, EntCurMean, EntCurMax, EntAttAdd, EntAttMean, EntAttMax, EntTgtAdd, EntTgtMean, EntTgtMax, SurCurAdd, SurCurMean, SurCurMax, SurAttAdd, SurAttMean, SurAttMax, SurTgtAdd, SurTgtMean, SurTgtMax))
+
+ sink()
+ print(paste("ANOVA", targetCol, dataFile))
+}
+################################################################################################
+
+################################################################################################
+computeCorrelation("outputs/UD_English-EWT/incr_pretrained_times.tsv")
+computeCorrelation("outputs/UD_English-EWT/incr_nopretrained_times.tsv")
+
+
+globalAnova("FIRST_TIME","outputs/UD_English-EWT/incr_nopretrained_times.tsv")
+globalAnova("FIRST_TIME","outputs/UD_English-EWT/incr_pretrained_times.tsv")
+globalAnova("FIRST_TIME","outputs/UD_English-EWT/seq_nopretrained_times.tsv")
+globalAnova("FIRST_TIME","outputs/UD_English-EWT/seq_pretrained_times.tsv")
+
+
+globalAnova("TOTAL_TIME","outputs/UD_English-EWT/incr_nopretrained_times.tsv")
+globalAnova("TOTAL_TIME","outputs/UD_English-EWT/incr_pretrained_times.tsv")
+globalAnova("TOTAL_TIME","outputs/UD_English-EWT/seq_nopretrained_times.tsv")
+globalAnova("TOTAL_TIME","outputs/UD_English-EWT/seq_pretrained_times.tsv")
+
+
+globalSummary("FIRST_TIME","outputs/UD_English-EWT/incr_nopretrained_times.tsv")
+globalSummary("FIRST_TIME","outputs/UD_English-EWT/incr_pretrained_times.tsv")
+globalSummary("FIRST_TIME","outputs/UD_English-EWT/seq_nopretrained_times.tsv")
+globalSummary("FIRST_TIME","outputs/UD_English-EWT/seq_pretrained_times.tsv")
+
+
+globalSummary("TOTAL_TIME","outputs/UD_English-EWT/incr_nopretrained_times.tsv")
+globalSummary("TOTAL_TIME","outputs/UD_English-EWT/incr_pretrained_times.tsv")
+globalSummary("TOTAL_TIME","outputs/UD_English-EWT/seq_nopretrained_times.tsv")
+globalSummary("TOTAL_TIME","outputs/UD_English-EWT/seq_pretrained_times.tsv")
+
--
GitLab