3 ms·
> We need a lot of data to do this well. Yup. And once you've got a sufficiently large tagged error corpus, spelling correction will be as simple as a lookup:
by tgv 7y ago
> We need a lot of data to do this well.
Yup. And once you've got a sufficiently large tagged error corpus, spelling correction will be as simple as a lookup: almost all misspellings (in terms of frequency) will be present in the corpus, and you can drop the (rather simplistic) algorithmic part.
- techbio 7y agoBy way of word2vec, or ?
- colejohnson66 7y agoBut what about “they’re”, “there”, and “their”? And other homonyms people frequently misspell?