testar a pipeline de décupla validação cruzada

[leoalenc]

Testar:

• TestSuite.py
• CreateModels.sh
• TenFoldCrossVal.sh (sem e com a opção -t)
• ParseGoldTokResults.py > gold-tok-tags-results.txt
• ParseRawTextResults.py raw-text-results-*.txt
• CompareRawTextParsing.py
• CompareGoldTokParsing.py

[leoalenc]

CompareRawTextParsing.py issue874/exp2/raw-text-results.json issue876/exp2/raw-text-results.json 
Comparing:
  EXP1: issue874/exp2/raw-text-results.json
  EXP2: issue876/exp2/raw-text-results.json
Test: Mann–Whitney U (two-sided), alpha=0.05, folds=10

PART 1 — EXP1 significantly outperforms EXP2
- (none)

PART 2 — EXP2 significantly outperforms EXP1
- (none)

PART 3 — No statistically significant difference
- Tokenizer sentences sum / f1: mean1=62.546, mean2=61.550, U=61.000, p=0.427355
- Parsing sum / LAS: mean1=69.563, mean2=69.211, U=57.000, p=0.623046
- Tagging sum / lemmas: mean1=90.886, mean2=90.740, U=56.500, p=0.650025
- Tokenizer multiword sum / f1: mean1=87.695, mean2=86.401, U=56.000, p=0.677356
- Parsing sum / UAS: mean1=74.796, mean2=74.490, U=55.000, p=0.733730
- Tagging sum / xpostag: mean1=88.730, mean2=88.601, U=54.000, p=0.791260
- Tagging sum / feats: mean1=85.971, mean2=86.082, U=47.000, p=0.850107
- Tagging sum / upostag: mean1=89.517, mean2=89.429, U=53.000, p=0.850107
- Tokenizer words sum / f1: mean1=94.085, mean2=94.054, U=51.000, p=0.969850
- Tokenizer tokens sum / f1: mean1=94.243, mean2=94.232, U=50.000, p=1.000000

JSON comparison report saved to: comparison-exp2-vs-exp2.json

[leoalenc]

CompareGoldTokParsing.py issue874/exp2/gold-tok-tags-results.txt issue876/exp2/gold-tok-tags-results.txt 
Mann–Whitney U statistic: 53.0
Two-tailed p-value: 0.8501067391385259

JSON results saved to: comparison-exp2-vs-exp2-LAS.json

[leoalenc]

leonel@leonel-Inspiron-3584:~/complin/nheengatu/data/experiments$ grep -Eho "forms: [0-9]+" issue876/exp2/gold-tok-tags-results-[0-9]*.txt | awk -F: '{sum += $2} END {print sum}'
26848
leonel@leonel-Inspiron-3584:~/complin/nheengatu/data/experiments$ grep -Eho "forms: [0-9]+" issue874/exp2/gold-tok-tags-results-[0-9]*.txt | awk -F: '{sum += $2} END {print sum}'
26785
leonel@leonel-Inspiron-3584:~/complin/nheengatu/data/experiments$ grep -Ec "^# sent_id" issue874/exp2/all.conllu 
2834
leonel@leonel-Inspiron-3584:~/complin/nheengatu/data/experiments$ grep -Ec "^# sent_id" issue876/exp2/all.conllu 
2839