GGroenendaal
commited on
Commit
•
081d5bf
1
Parent(s):
8fe5a80
move preprocessing to dependency injection
Browse files- base_model/evaluate.py +18 -20
- base_model/string_utils.py +20 -0
base_model/evaluate.py
CHANGED
@@ -1,29 +1,27 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
"""Preprocesses the sentence string by normalizing.
|
3 |
|
4 |
Args:
|
5 |
s (str): the sentence
|
6 |
|
7 |
Returns:
|
8 |
-
string: normalized
|
9 |
"""
|
10 |
-
import string, re
|
11 |
-
|
12 |
-
def remove_articles(text):
|
13 |
-
regex = re.compile(r"\b(a|an|the)\b", re.UNICODE)
|
14 |
-
return re.sub(regex, " ", text)
|
15 |
-
|
16 |
-
def white_space_fix(text):
|
17 |
-
return " ".join(text.split())
|
18 |
-
|
19 |
-
def remove_punc(text):
|
20 |
-
exclude = set(string.punctuation)
|
21 |
-
return "".join(ch for ch in text if ch not in exclude)
|
22 |
|
23 |
-
|
24 |
-
return text.lower()
|
25 |
|
26 |
-
return
|
27 |
|
28 |
|
29 |
def compute_exact_match(prediction: str, answer: str) -> int:
|
@@ -36,7 +34,7 @@ def compute_exact_match(prediction: str, answer: str) -> int:
|
|
36 |
Returns:
|
37 |
int: 1 for exact match, 0 for not
|
38 |
"""
|
39 |
-
return int(
|
40 |
|
41 |
|
42 |
def compute_f1(prediction: str, answer: str) -> float:
|
@@ -49,8 +47,8 @@ def compute_f1(prediction: str, answer: str) -> float:
|
|
49 |
Returns:
|
50 |
boolean: the f1 score
|
51 |
"""
|
52 |
-
pred_tokens =
|
53 |
-
answer_tokens =
|
54 |
|
55 |
if len(pred_tokens) == 0 or len(answer_tokens) == 0:
|
56 |
return int(pred_tokens == answer_tokens)
|
|
|
1 |
+
from typing import Callable, List
|
2 |
+
|
3 |
+
from base_model.string_utils import lower, remove_articles, remove_punc, white_space_fix
|
4 |
+
|
5 |
+
|
6 |
+
def normalize_text(inp: str, functions: List[Callable[[str], str]]):
|
7 |
+
for fun in functions:
|
8 |
+
inp = fun(inp)
|
9 |
+
return inp
|
10 |
+
|
11 |
+
|
12 |
+
def normalize_text_default(inp: str) -> str:
|
13 |
"""Preprocesses the sentence string by normalizing.
|
14 |
|
15 |
Args:
|
16 |
s (str): the sentence
|
17 |
|
18 |
Returns:
|
19 |
+
string: normalized with default parames
|
20 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
steps = [remove_articles, white_space_fix, remove_punc, lower]
|
|
|
23 |
|
24 |
+
return normalize_text(inp, steps)
|
25 |
|
26 |
|
27 |
def compute_exact_match(prediction: str, answer: str) -> int:
|
|
|
34 |
Returns:
|
35 |
int: 1 for exact match, 0 for not
|
36 |
"""
|
37 |
+
return int(normalize_text_default(prediction) == normalize_text_default(answer))
|
38 |
|
39 |
|
40 |
def compute_f1(prediction: str, answer: str) -> float:
|
|
|
47 |
Returns:
|
48 |
boolean: the f1 score
|
49 |
"""
|
50 |
+
pred_tokens = normalize_text_default(prediction).split()
|
51 |
+
answer_tokens = normalize_text_default(answer).split()
|
52 |
|
53 |
if len(pred_tokens) == 0 or len(answer_tokens) == 0:
|
54 |
return int(pred_tokens == answer_tokens)
|
base_model/string_utils.py
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import re
|
2 |
+
import string
|
3 |
+
|
4 |
+
|
5 |
+
def remove_articles(text):
|
6 |
+
regex = re.compile(r"\b(a|an|the)\b", re.UNICODE)
|
7 |
+
return re.sub(regex, " ", text)
|
8 |
+
|
9 |
+
|
10 |
+
def white_space_fix(text):
|
11 |
+
return " ".join(text.split())
|
12 |
+
|
13 |
+
|
14 |
+
def remove_punc(text):
|
15 |
+
exclude = set(string.punctuation)
|
16 |
+
return "".join(ch for ch in text if ch not in exclude)
|
17 |
+
|
18 |
+
|
19 |
+
def lower(text):
|
20 |
+
return text.lower()
|