From 8ec11dd5e2419cffdbddb7720892b3e61953f16b Mon Sep 17 00:00:00 2001
From: "Vriezen, E.C. (Emma)" <e.vriezen@student.ru.nl>
Date: Sun, 1 Dec 2019 20:03:50 +0100
Subject: [PATCH] The repetition checking works now (in mistake_finder.py). Is
run after the anchor algorithm.
---
documentation/MistakeDefinitions.txt | 1 +
umbra/filereader.py | 2 +-
umbra/mistake_finder.py | 54 +++++++++++++++++++++++-----
umbra/words.py | 15 ++++++--
4 files changed, 60 insertions(+), 12 deletions(-)
diff --git a/documentation/MistakeDefinitions.txt b/documentation/MistakeDefinitions.txt
index 91c8cdc8..96319ebc 100644
--- a/documentation/MistakeDefinitions.txt
+++ b/documentation/MistakeDefinitions.txt
@@ -1,4 +1,5 @@
Stuttering/repetition mistakes:
+(These are all implemented in mistake_finder.py)
- 2 or more times the same word in a row in the shadow, while it only appears
once in the source. Example: 'en' and 'en' in file 7 (2nd 'en' is repetition)
diff --git a/umbra/filereader.py b/umbra/filereader.py
index b35c2589..9ce3f469 100644
--- a/umbra/filereader.py
+++ b/umbra/filereader.py
@@ -162,7 +162,7 @@ class CSVWriter(FileWriter):
info = "Total " + str(info[0]), "Shadowed " + str(info[1]), "Skipped"\
" " + str(info[2])
sc = []
- for entry in source.words:
+ for entry in source:
sc.append([entry.word, entry.onset, entry.offset, entry.shadowed])
with open(path+'.csv', 'w') as f:
writer = csv.writer(f)
diff --git a/umbra/mistake_finder.py b/umbra/mistake_finder.py
index 447f2282..1d8ba502 100644
--- a/umbra/mistake_finder.py
+++ b/umbra/mistake_finder.py
@@ -6,22 +6,25 @@ class MistakeFinder:
"""Finds the mistakes in already aligned lists of Words"""
def __init__(self):
- self.source = None
- self.shadow = None
+ self._source = None
+ self._shadow = None
def start(self, source, shadow):
"""Find all the mistakes and classify them."""
- self.source = source
- self.shadow = shadow
+ self._source = source
+ self._shadow = shadow
# Loop over the shadow:
- for index, word in enumerate(self.shadow):
- if not word.correct:
+ for index, word in enumerate(self._shadow):
+ if not word.correct and word.mistake is None:
self._determine_mistake(index, word)
# Loop over the source:
- for word in self.source:
+ for word in self._source:
if not word.shadowed and word.mistake is None:
# If not yet marked as mistake, then it is skipped:
word.mistake = Mistake.SKIPPED
+ for word in self._shadow:
+ if word.mistake == Mistake.REPETITION:
+ print(word)
def _determine_mistake(self, index, word):
if not self._check_repetition(word, index): # If not a repetition,
@@ -31,8 +34,41 @@ class MistakeFinder:
word.mistake = Mistake.RANDOM
def _check_repetition(self, word, index):
- assert 0 <= index < len(self.shadow)
- # print(index)
+ assert 0 <= index < len(self._shadow)
+ found = self._check_pre_repetition(word, index)
+ if not found:
+ if index < len(self._shadow)-1 and self._shadow[index+1].word.find(
+ word.word) == 0:
+ word.mistake = Mistake.REPETITION
+ return True
+ return found
+
+ def _check_pre_repetition(self, word, index):
+ found_before = False
+ stop = False
+ diff = 0
+ i = index
+
+ while not found_before and not stop and i > 0:
+ i -= 1
+ stop = self._shadow[i].is_anchor()
+ found_at = self._shadow[i].word.find(word.word)
+ if found_at >= 0 and found_at == len(self._shadow[i].word)-\
+ len(word.word):
+ found_before = True
+
+ chain = found_before
+ while chain and index-i > 1:
+ diff += 1
+ i += 1
+ chain = self._shadow[i].word == self._shadow[index+diff].word
+
+ if chain:
+ for j in range(index, index+diff+1):
+ self._shadow[j].mistake = Mistake.REPETITION
+ else:
+ return False
+ return True
# TODO: WordNet integration. Function below should be usable already!
diff --git a/umbra/words.py b/umbra/words.py
index 689fe325..de047c62 100644
--- a/umbra/words.py
+++ b/umbra/words.py
@@ -10,7 +10,7 @@ class Word:
self._mistake = None
def __str__(self):
- return "%s | %f | %f" % (self._word, self._onset, self._offset)
+ return "%s | %s | %f | %f" % (self._word, self._mistake, self._onset, self._offset)
def __len__(self):
return 1
@@ -92,7 +92,11 @@ class Word:
Args:
anchor: Word instance to anchor to
"""
- self._anchor = anchor
+ self._set_anchor(anchor)
+
+ def _set_anchor(self, anchor):
+ """Anchor setter. Has to be overridden in the subclass."""
+ raise NotImplementedError
def get_difference(self, other):
"""Get the difference between the onset of this word and the other.
@@ -164,6 +168,10 @@ class ShadowWord(Word):
assert mistake != Mistake.SKIPPED
self._mistake = mistake
+ def _set_anchor(self, anchor):
+ self._anchor = anchor
+ self._source = anchor
+
class SourceWord(Word):
def __init__(self, word, onset, offset):
@@ -197,6 +205,9 @@ class SourceWord(Word):
assert mistake != Mistake.RANDOM
self._mistake = mistake
+ def _set_anchor(self, anchor):
+ self._anchor = anchor
+
class Sentence(list):
def __init__(self, words):
--
GitLab