Question

I’m using NHunspell to check a string for spelling errors like so:

var words = content.Split(' ');
string[] incorrect;
using (var spellChecker = new Hunspell(affixFile, dictionaryFile))
{
    incorrect = words.Where(x => !spellChecker.Spell(x))
        .ToArray();
}

This generally works, but it has some problems. For example, if I’m checking the sentence “This is a (very good) example”, it will report “(very” and “good)” as being misspelled. Or if the string contains a time such as “8:30”, it will report that as a misspelled word. It also has problems with commas, etc.

Microsoft Word is smart enough to recognize a time, fraction, or comma-delimited list of words. It knows when not to use an English dictionary, and it knows when to ignore symbols. How can I get a similar, more intelligent spell check in my software? Are there any libraries that provide a little more intelligence?

EDIT:
I don’t want to force users to have Microsoft Word installed on their machine, so using COM interop is not an option.

Answer 1

If your spell checker is really that stupid, you should pre-tokenize its input to get the words out and feed those one at a time (or as a string joined with spaces). I’m not familiar with C#/.NET, but in Python, you’d use a simple RE like \w+ for that:

>>> s = "This is a (very good) example"
>>> re.findall(r"\w+", s)
['This', 'is', 'a', 'very', 'good', 'example']

and I bet .NET has something very similar. In fact, according to the .NET docs, \w is supported, so you just have to find out how re.findall is called there.