Markov analysis - Return and recursion role - python-3.x

Question

I am working on the solution of the Markov analysis in Think Python, but I do not understand the role of "Return" in the block code below.
As far as I known when the code reach return the function is cancel immediately, but isn't it unnecessary in this case, because there is a recursion here random_text(n-i) before the code reach the return statement, so the function will cancel only when the recursion is finish which mean when the for loop is over?? The question seem stupid but I am newbie in python and the recursion stuff is really confusing with me. I try to remove 'return' and it still run well.
def random_text(n=100):
start = random.choice(list(suffix_map.keys()))
for i in range(n):
suffixes = suffix_map.get(start, None)
if suffixes == None:
# if the start isn't in map, we got to the end of the
# original text, so we have to start again.
random_text(n-i)
return
word = random.choice(suffixes)
print(word, end=' ')
start = shift(start, word)
The full code is as below so you can understand what each function do.
from __future__ import print_function, division
import os
os.chdir(r"C:\Users\Hoang-Ngoc.Anh\Documents\WinPython-64bit 3.4.4.2\notebooks\docs")
import sys
import string
import random
# global variables
suffix_map = {} # map from prefixes to a list of suffixes
prefix = () # current tuple of words
def process_file(filename, order=2):
"""Reads a file and performs Markov analysis.
filename: string
order: integer number of words in the prefix
returns: map from prefix to list of possible suffixes.
"""
fp = open(filename)
skip_gutenberg_header(fp)
for line in fp:
for word in line.rstrip().split():
process_word(word, order)
def skip_gutenberg_header(fp):
"""Reads from fp until it finds the line that ends the header.
fp: open file object
"""
for line in fp:
if line.startswith('*END*THE SMALL PRINT!'):
break
def process_word(word, order=2):
"""Processes each word.
word: string
order: integer
During the first few iterations, all we do is store up the words;
after that we start adding entries to the dictionary.
"""
global prefix
if len(prefix) < order:
prefix += (word,)
return
try:
suffix_map[prefix].append(word)
except KeyError:
# if there is no entry for this prefix, make one
suffix_map[prefix] = [word]
prefix = shift(prefix, word)
def random_text(n=100):
"""Generates random wordsfrom the analyzed text.
Starts with a random prefix from the dictionary.
n: number of words to generate
"""
# choose a random prefix (not weighted by frequency)
start = random.choice(list(suffix_map.keys()))
for i in range(n):
suffixes = suffix_map.get(start, None)
if suffixes == None:
# if the start isn't in map, we got to the end of the
# original text, so we have to start again.
random_text(n-i)
return
# choose a random suffix
word = random.choice(suffixes)
print(word, end=' ')
start = shift(start, word)
def shift(t, word):
"""Forms a new tuple by removing the head and adding word to the tail.
t: tuple of strings
word: string
Returns: tuple of strings
"""
return t[1:] + (word,)
def main(script, filename='emma.txt', n=100, order=2):
try:
n = int(n)
order = int(order)
except ValueError:
print('Usage: %d filename [# of words] [prefix length]' % script)
else:
process_file(filename, order)
random_text(n)
print()
if __name__ == '__main__':
main(*sys.argv)