From ac6c34002ac87a7999ab9c85c7f20a840f50b32a Mon Sep 17 00:00:00 2001
From: Charlotte Mauvezin <charlotte.mauvezin@irht.cnrs.fr>
Date: Tue, 4 Jan 2022 12:08:00 +0000
Subject: [PATCH] Adding a line with the mean total for the precision, the
recall and the f1...
---
nerval/evaluate.py | 22 +++++++++++++++++++---
1 file changed, 19 insertions(+), 3 deletions(-)
diff --git a/nerval/evaluate.py b/nerval/evaluate.py
index ab11dde..2f0070d 100644
--- a/nerval/evaluate.py
+++ b/nerval/evaluate.py
@@ -514,7 +514,6 @@ def run(annotation: str, prediction: str, threshold: int, verbose: bool) -> dict
Each measure is given at document level, global score is a micro-average across entity types.
"""
-
# Get string and list of labels per character
annot = parse_bio(annotation)
predict = parse_bio(prediction)
@@ -570,6 +569,10 @@ def run_multiple(file_csv, folder, threshold, verbose):
if os.path.isdir(folder):
list_bio_file = glob.glob(str(folder) + "/**/*.bio", recursive=True)
+ count = 0
+ precision = 0
+ recall = 0
+ f1 = 0
for row in list_cor:
annot = None
predict = None
@@ -582,11 +585,24 @@ def run_multiple(file_csv, folder, threshold, verbose):
predict = file
if annot and predict:
+ count += 1
print(os.path.basename(predict))
- run(annot, predict, threshold, verbose)
+ scores = run(annot, predict, threshold, verbose)
+ precision += scores["All"]["P"]
+ recall += scores["All"]["R"]
+ f1 += scores["All"]["F1"]
print()
else:
- raise f"No file found for files {annot}, {predict}"
+ raise Exception(f"No file found for files {annot}, {predict}")
+ if count:
+ print(
+ "Average scores in all corpus (mean of final files scores)\n"
+ f" * Precision: {round(precision/count, 3)}\n"
+ f" * Recall: {round(recall/count, 3)}\n"
+ f" * F1: {round(f1/count, 3)}\n"
+ )
+ else:
+ raise Exception("No file were counted")
else:
raise Exception("the path indicated does not lead to a folder.")
--
GitLab