I think I know the answer to this, but in case anyone has done something
clever...

Is there a way to get raw #s from the recognizer as to confidence, or
are we stuck with Poor, Intermediate & Strong?

I haven't played with the InkAnalyzer in .Net 3.0 yet, but if it's
available there, that's useful info as well (the class doesn't elaborate
on it).

You're stuck with it. And also it's only available on US English recognizer.
But if you're dealing with lines/paragraphs the confidence is pretty much
useless cause I think it just takes the least of the whole bunch. You would
have better luck dealing with confidence at the word level.

"Chris DiPierro" <cdipierr@mi-corporation.com> wrote in message
news:uqmBothyHHA.4476@TK2MSFTNGP06.phx.gbl...
>I think I know the answer to this, but in case anyone has done something
>clever...
>
> Is there a way to get raw #s from the recognizer as to confidence, or are
> we stuck with Poor, Intermediate & Strong?
>
> I haven't played with the InkAnalyzer in .Net 3.0 yet, but if it's
> available there, that's useful info as well (the class doesn't elaborate
> on it).

Yeah, that seems to mirror what we've found so far. Unfortunately, some
of our customers are pretty strict about confidence levels. Other
engines we've used in the past give a 0-100% score, which is nice. We'll
have to see how this flies.


Josh Einstein wrote:
> You're stuck with it. And also it's only available on US English
> recognizer. But if you're dealing with lines/paragraphs the confidence
> is pretty much useless cause I think it just takes the least of the
> whole bunch. You would have better luck dealing with confidence at the
> word level.
>
> "Chris DiPierro" <cdipierr@mi-corporation.com> wrote in message
> news:uqmBothyHHA.4476@TK2MSFTNGP06.phx.gbl...
>> I think I know the answer to this, but in case anyone has done
>> something clever...
>>
>> Is there a way to get raw #s from the recognizer as to confidence, or
>> are we stuck with Poor, Intermediate & Strong?
>>
>> I haven't played with the InkAnalyzer in .Net 3.0 yet, but if it's
>> available there, that's useful info as well (the class doesn't
>> elaborate on it).
>

Well you could fabricate more granular results at the line or paragraph
level by giving the word values some numeric value such as Unknown = 0, Poor
= 25, Intermediate = 50, Strong = 100 then average them together to get a
confidence for the whole line or paragraph. But this still probably isn't as
granular as you're looking for.

Josh

"Chris DiPierro" <cdipierr@mi-corporation.com> wrote in message
news:Oa0dyHtyHHA.5376@TK2MSFTNGP03.phx.gbl...
> Yeah, that seems to mirror what we've found so far. Unfortunately, some of
> our customers are pretty strict about confidence levels. Other engines
> we've used in the past give a 0-100% score, which is nice. We'll have to
> see how this flies.
>
>
> Josh Einstein wrote:
>> You're stuck with it. And also it's only available on US English
>> recognizer. But if you're dealing with lines/paragraphs the confidence is
>> pretty much useless cause I think it just takes the least of the whole
>> bunch. You would have better luck dealing with confidence at the word
>> level.
>>
>> "Chris DiPierro" <cdipierr@mi-corporation.com> wrote in message
>> news:uqmBothyHHA.4476@TK2MSFTNGP06.phx.gbl...
>>> I think I know the answer to this, but in case anyone has done something
>>> clever...
>>>
>>> Is there a way to get raw #s from the recognizer as to confidence, or
>>> are we stuck with Poor, Intermediate & Strong?
>>>
>>> I haven't played with the InkAnalyzer in .Net 3.0 yet, but if it's
>>> available there, that's useful info as well (the class doesn't elaborate
>>> on it).
>>