Finding multiple substrings in a string without iterating over it multiple times

One way I could think of to improve is:

• Get all unique lengths of the words in _legal
• Build a dictionary of words from line of those particular lengths using a sliding window technique. The complexity should be O( len(line)*num_of_unique_lengths ), this should be better than brute force.
• Now look for each thing in the dictionary in O(1).

Code:

line = 'thing1 thing2 456 xxualt542l lthin. dfjladjfj lauthina '
_legal = ['thing1', 'thing2', 'thing3', 'thing4', 't5', '5', 'fj la']
ul = {len(i) for i in _legal}
s=set()
for l in ul:
    s = s.union({line[i:i+l] for i in range(len(line)-l)})
print(s.intersection(set(_legal)))

Output:

{'thing1', 'fj la', 'thing2', 't5', '5'}