Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 38 additions & 0 deletions frequency_each_word.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,38 @@
"""
Write an algorithm to find the frequency of each word in a given text.
Your algorithm should keep track of each word it encounters and the number
of times it has appeared. (You can ignore case sensitivity.)
"""

def func(text : str):
word_array = []
current_word =""
word_check =""
for char in text:
if char.isalnum():
current_word +=char
elif char.isspace():
if current_word:
word_array.append(current_word)
current_word = ""
elif char in ".,-_*?\;/}]{[&%+^'!":
if current_word:
word_array.append(current_word)
current_word = ""
word_array.append(char)
if current_word:
word_array.append(current_word)


for element in set(word_array):
word_counter =0
word_check = element
for j,element in enumerate(word_array):
if word_check == word_array[j]:
word_counter+=1
print_elements =[]
temp_element = f"{word_check} : {word_counter}"
print_elements.append(temp_element)
print(print_elements)
TEXT = """Magnus Carlsen is a chess player. He has become world! champions! many times..."""
func(TEXT)