def levenshtein_method(cdr3, cutoff=1):
edgelist = set()
combos = [comb for comb in combinations(list(cdr3), 2)]
for combo in combos:
d = lev(combo[0], combo[1])
if d <= cutoff:
edgelist.add(combo[0] + "\t" + combo[1] + "\t" + str(d))
return edgelist
def have_same_root(word1, word2):
w1_stem = stemmer.stemWord(word1)
w2_stem = stemmer.stemWord(word2)
if w1_stem == w2_stem or lev(w1_stem, w2_stem) <= 2:
return True
comm = common_letters(w1_stem,w2_stem)
if comm < 4:
return False
if comm > 6:
return True
diff = abs(len(max(w1_stem,w2_stem))-comm)
if diff > comm or diff > 2:
return False
return True
def get_similar_code(processed_code, language, level):
filename = "coursedata/similar-code-files/" + language + "/level" + str(level) + ".csv"
shortest_distance = 75 # This is the threshold: when differ more than this value it's no longer similar code
similar_code = None
try:
with open(filename, mode='r', encoding='utf-8') as file:
csvFile = csv.reader(file, quoting=csv.QUOTE_MINIMAL)
for lines in csvFile:
distance = lev(processed_code, lines[1])
if distance < 1: # The code is identical, no need to search any further
similar_code = lines[0]
break
else:
if distance < shortest_distance:
shortest_distance = distance
similar_code = lines[0]
except:
similar_code = None
return similar_code
def getCloserMatch(word, wordlist):
"""
Search closest match (using Levenshtein distance)
Args:
word: The word to search in the list.
wordlist: Path to the wordlist to perform the search into.
Returns:
A tuple holding the matches array and Levenshtein distance between the
word variable and these matches.
Example:
>>> getCloserMatch("azerto", "wordlists/french_top1000.txt")
Levenshtein distance: 1
Match: azerty
Match: azert
"""
try:
f = open(wordlist, 'r')
passwords = f.read().splitlines()
f.close()
except:
print("Could not find required file")
exit(1)
min_distance = math.inf
matches = []
for password in passwords:
distance = lev(word, password)
if distance < min_distance:
min_distance = distance
matches = [password]
elif distance == min_distance:
matches.append(password)
return matches, min_distance
clean.append(poljeOrig[x - 1])
#stavljanje u string i ciscenje sil, uzdah, buka i greska segmenata
stringGlasova = ""
for x in range(len(clean)):
if str(clean[x]) != "sil" and str(clean[x]) != "uzdah" and str(
clean[x]) != "buka" and str(clean[x]) != "greska":
stringGlasova = stringGlasova + clean[x]
#ispis prepoznatog i originalnog stringa glasova
print "\nISPIS ORIGINAL I PREPOZNATO"
print stringLab
print stringGlasova
#racunanje preciznosti s levenshtein distancom
edit_dist = lev(stringLab, stringGlasova)
print "\nPRECIZNOST"
if odabirDistance == "M":
print "KORISTECI MAHALANOBISOVU UDALJENOST"
elif odabirDistance == "E":
print "KORISTECI EUKLIDOVU UDALJENOST"
print str(100 - float(edit_dist) / len(labData) * 100) + "%"
#formatiranje i zapisivanje u out.lab
start = 0
distribution = len(mfcc_feat) * 100000 / len(clean)
text_file = open("out/" + audioSignal.split(".")[0] + "Out.lab", "w")
text_file.write("")
text_file.close()
for x in range(len(clean)):
text_file = open("out/" + audioSignal.split(".")[0] + "Out.lab", "a")
#!/usr/bin/env python
from Levenshtein import distance as lev
def pairwise(iterable):
a = iter(iterable)
return zip(a, a)
with open("out.txt", "r") as _file:
errors = []
for line in _file:
if not line[:2] == "--":
errors.append(line[2:].strip())
_sum = 0
for transcribed, gold in pairwise(errors):
_sum = _sum + lev(transcribed, gold)
msg = f"""Out of {round(len(errors)/2)} errors, the average edit distance was {_sum/(len(errors)/2)}"""
print(msg)
def levenshtein(x, y, n=None):
if n is not None:
x = x[:n]
y = y[:n]
return lev(x, y) / (max(len(x), len(y)) if max(len(x), len(y)) > 0 else 1)
def extract_text_chat_screen(frame, frame_count=0):
# cv2.imshow("crop_img", crop_img)
global ChatLogArray
img = frame.copy()
# convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# threshold
ret, bin = cv2.threshold(gray, 245, 255, cv2.THRESH_BINARY)
# closing
kernel = np.ones((3, 3), np.uint8)
closing = cv2.morphologyEx(bin, cv2.MORPH_CLOSE, kernel)
# invert black/white
inv = cv2.bitwise_not(closing)
# cv2.imshow("img_outline", closing)
# cv2.waitKey()
height = 1080
# a boolean which indicated if the messages from the
# person playing the game should also be included
include_player_messages = True
# because the iterator might need to change, don't use for loop
y = 0 # start at the top
while y < 750: # stop when the chat input box is reached
y = y + 1
if y > 750: # reached the end of chat. So stop here
break
x = 288 # in the case of 1920 x 1080
is_message_box = False # Check if there is a message box at this y
if closing[y, x] == 255 and closing[y + 100, x] == 255:
is_message_box = True
if is_message_box:
end_y = y + 10
while closing[end_y, x] == 255:
end_y = end_y + 1
crop_img_text = closing[y + 50:end_y - 2, 420:1253]
if crop_img_text.size == 0:
print("failure in grabbing text from chat")
y = end_y + 10
continue
OCR_result = get_text_from_image(crop_img_text)
# how many pixels to move downwards to find the color of the person
color_person_y = 50
# get the color values of the player saying the word
b, g, r = frame[y + color_person_y, x + 100]
# get the color name based on the RGB values
color_name = utils.get_color_name(r, g, b)
# variable which indicates if a chat message is already logged
# This is done to prevent duplicates
already_logged = False
for chat in ChatLogArray:
if color_name == chat.colorName:
if chat.message == OCR_result:
already_logged = True
# Also check if the levenshtein distance == 1 in case of OCR errors
# if the distance is only 1, then it very likely saw the same message,
# which is usually caused by an OCR error reading a character the wrong way
levenshtein_distance = lev(chat.message, OCR_result)
if levenshtein_distance == 1:
already_logged = True
# if the chat entry doesn't exist yet, create it
if not already_logged:
# now get the name through OCR
crop_img = closing[y:y + 50, 420:1253]
OCR_result_name = get_text_from_image(crop_img)
# now put it all in the chat logger
new_log = ChatLog(len(ChatLogArray), r, g, b, color_name, frame_count, OCR_result_name, OCR_result)
ChatLogArray.append(new_log)
y = end_y + 10
# now do the same for the player's chat messages.
y = 0
while (y < 750):
y = y + 1
if y > 750: # reached the end of chat. So stop here
break
x_player = 1375 # player messages in the case of 1920 x 1080
is_message_box_player = False
if closing[y, x_player] == 255 and closing[y + 100, x_player] == 255 and include_player_messages:
is_message_box_player = True
if is_message_box_player:
end_y = y + 10
while closing[end_y, x_player] == 255:
end_y = end_y + 1
crop_img_text = closing[y + 50:end_y - 2, 420:1253]
if crop_img_text.size == 0:
print("failure in grabbing text from chat")
y = end_y + 10
continue
OCR_result = get_text_from_image(crop_img_text)
# how many pixels to move downwards to find the color of the person
color_person_y = 50
# get the color values of the player saying the word
b, g, r = frame[y + color_person_y, x_player - 40]
color_name = utils.get_color_name(r, g, b)
already_logged = False
for chat in ChatLogArray:
# and chat.r == r and chat.g == g and chat.b == b
# check if this crewmate already said this (don't log double)
if color_name == chat.colorName:
if chat.message == OCR_result:
already_logged = True
# Also check if the levenshtein distance == 1 in case of OCR errors
# if the distance is only 1, then it very likely saw the same message,
# which is usually caused by an OCR error reading a character the wrong way
levenshtein_distance = lev(chat.message, OCR_result)
if levenshtein_distance == 1:
already_logged = True
# if chat.r == r and chat.g == g and chat.b == b:
# already_logged = True
# if the chat entry doesn't exist yet. Create it
if not already_logged:
# now get the name through OCR
# crop_img = closing[ y+50:y+100, 420:1253]
crop_img = closing[y:y + 50, 420:1253]
OCR_result_name = get_text_from_image(crop_img)
# now put it all in the chat logger
new_log = ChatLog(len(ChatLogArray), r, g, b, color_name, frame_count, OCR_result_name, OCR_result)
ChatLogArray.append(new_log)
# print(ChatLogArray)
# cv2.floodFill()
# cv2.imshow("img_outline", crop_img)
y = end_y + 10