lets preview to another approach as well, simulating more on how a human would handle
The idea:
- the word of interest is known
- we start with the first char and ommit anything that is not a letter
- then the n+1 char is checked if it is same as the n+1 char in the searched word.
- yes: continue with n+2…
- no: abort
A strategy like take first char and last char, extracting a part result spanning up this within the text will work for your word but not for a word like “houses” as the ending char also occurs within the token.