From 38520c3b30c31f339633ebbf5ae1d864b17358a3 Mon Sep 17 00:00:00 2001
From: Franck Dary <franck.dary@lis-lab.fr>
Date: Tue, 26 Nov 2019 15:21:15 +0100
Subject: [PATCH] Updated print_result.py script to compute standard deviation
---
UD_any/print_results.py | 34 +++++++++++++++++++++++++++++++---
1 file changed, 31 insertions(+), 3 deletions(-)
diff --git a/UD_any/print_results.py b/UD_any/print_results.py
index e4ac9ca..7b1abe3 100755
--- a/UD_any/print_results.py
+++ b/UD_any/print_results.py
@@ -2,29 +2,57 @@
import glob
import sys
+import math
if __name__ == "__main__" :
metrics = ["LAS","UAS","Tokens","Sentences","UPOS","UFeats","Lemmas"]
output = []
+ outputByModelScore = dict()
for pathToFile in glob.iglob("" + '*stdout') :
- model = pathToFile.split(".")[0].split("_UD_")[0]
- corpus = pathToFile.split(".")[0].split("_UD_")[1]
+ model = pathToFile.split("_UD_")[0]
+ corpus = pathToFile.split("_UD_")[1].split('.')[0]
for line in open(pathToFile, "r") :
for metric in metrics :
if metric in line and metric[0] == line[0]:
splited = line.strip().replace("|","").split()
- output.append([corpus, splited[0], splited[3], model])
+
+ model = model.split('.')[0]
+
+ if model not in outputByModelScore :
+ outputByModelScore[model] = dict()
+ if splited[0] not in outputByModelScore[model] :
+ outputByModelScore[model][splited[0]] = []
+
+ outputByModelScore[model][splited[0]].append([corpus, splited[0], splited[3], model])
+
+ for model in outputByModelScore :
+ for metric in outputByModelScore[model] :
+ score = 0.0
+ standardDeviation = 0.0
+ for exp in outputByModelScore[model][metric] :
+ score += float(exp[2])
+ score /= len(outputByModelScore[model][metric])
+ for exp in outputByModelScore[model][metric] :
+ standardDeviation += (float(exp[2])-score)**2
+ standardDeviation /= len(outputByModelScore[model][metric])-1
+ standardDeviation = math.sqrt(standardDeviation)
+ score = "%.2f[±%.2f]%%"%(score,standardDeviation)
+ output.append(outputByModelScore[model][metric][0])
+ output[-1][2] = score
maxColLens = [0 for _ in range(len(output[0]))]
+ output = [["Corpus","Metric","F1.score","Model"]] + output
+
for line in output :
for i in range(len(line)) :
maxColLens[i] = max(maxColLens[i], len(line[i]))
+ output = output[1:]
output.sort()
output = [["Corpus","Metric","F1.score","Model"]] + output
--
GitLab