what exactly is max_edit_distance_dictionary about symspellpy HOT 3 CLOSED

It determines the content in _deletes dictionary. The keys are the edit candidates generated by only deletes which are within an edit distance of 1 to max_edit_distance_dictionary of the original words. The dictionary values are lists of suggested correction words.

For example, with a max_edit_distance_dictionary = 1, _deletes contains

'{te': ['the'], 'he': ['the'], 'the': ['the'], 'th': ['the']}

and when max_edit_distance_dictionary = 2, _deletes contains

'h': ['the'], 'e': ['the'], 'he': ['the'], 'th': ['the'], 't': ['the'], 'te': ['the'],
'the': ['the']}

"h", "e" and "t" were not previously included since they can only be obtained by making 2 deletes (edits) from the original word "the".

prefix_length also plays a part in generating _deletes since it limits the "original word" from which we make deletes. For example, when prefix_length = 7, "unhpi" can be mapped to "unhappiness" since it's 2 deletes from "unhappi".

So we enforce max_edit_distance <= max_edit_distance_dictionary since _deletes will contain any mapping with an edit distance greater then max_edit_distance_dictionary. There was also a mistake in _edits_prefix(),

if len(key) > self._max_dictionary_edit_distance:

should have been

if len(key) > self._prefix_length:

I'm sorry if that caused this confusion.

from symspellpy.

Thanks, it is very clear now. So _edits_prefix() will be fixed in the later version, right?

from symspellpy.

The changes has been pushed to the dev branch, I'll merge it to master later. Interestingly, it does not seem to affect the accuracy of the program since having a stop index larger then the string length is handled gracefully in python. For example

s = "abc"
print(s[: 10])

gives

abc

which is what _edits_prefix() wanted.

from symspellpy.