Spell Checking Question

[docmeizie]

Okay when spell check runs, it has to reference a database of all the words that are words. I really want to know what it is referencing so perhaps I can reference in a project I am working on myself. I know this can be done cause you can reference the spell checker on a form by going through Word. But I am trying not to have to go through Word. I am creating a game that compares the all the words a user comes up with to all viable words for a given set of letters mixed up.

If I take a peek in your Windows, to fix a problem, does that make me a "Peeping Tom"? Hmmmmmmmmmmv

 

[johnwm]

strongm

I agree that Dictioanry requires less coding but am I missing something obvious? Is there a way of loading a Dictionary quicker than using the .Add method?

To load the file and do a split into the basic array takes approx 1.4 seconds (as a complied .exe)

To load the Dictionary from the array using .Add takes an additional approx 7.3 seconds (also compiled)

I also got some interesting results on actually finding words. All these figures are averaged over 10000 repeats

[tt]
Using Dictionary Using BSA
abacus 24.9 microsec 10.8 microsec
monolith 18.0 10.9
zoo 9.1 10.5
[/tt]

BSA seems to be fairly consistent, but Dictionary appears to find the latest additions much faster than the early ones. Any thoughts on why that might be, and how I could load the Dictionary faster?

________________________________________________________________
If you want to get the best response to a question, please check out FAQ222-2244 first

'If we're supposed to work in Hex, why have we only got A fingers?'

 

[ESquared]

Looking up a single word 10,000 times will not properly demonstrate any speed differential. We could both get a lucky hit onto a word that sits right on a binary power multiple boundary.

Instead, I suggest that we submit the entire contents of the dictionary to the dictionary, or even submit the reversed-word contents of the dictionary. This will give us both the same set of words to time.

We should use a consistent method to time it, and I also propose 3 separate times: Dictionary load time (from file); index creation time (for you, zero, but put the timer in there anyway), then lookup time for all the reversed words. I suppose the total time can be included.

That is, we should use the same timeGetTime function or whatever.

What say you?

 

[ESquared]

Unfortunately I am back at work and can't work on my pet projects... that must wait until my own time. And also unfortunately, I am going out of the country for 2 weeks on Thursday, so my time is mostly eaten up by preparation. I may not be able to do this soon. :-(

I will see what I can do, and if necessary will revive this thread when I return.

-E

 

[strongm]

>Is there a way of loading a Dictionary quicker than using the .Add method?

Sadly not...however, I was going with ESquared's request that we "...not have its build-time at program startup counted"

Plus, can I confirm that you are using the Dictionary object's .Exists method rather than .Item property when doing your timings, as this a much closer match to your FindWord function?

 

[johnwm]

Esquared,

Regarding lucky hits on boundaries, I've just tested with 'loaves' (on the very first division) and I get 4.2 microsec . It's the first one I've found that gives significantly different times though.

________________________________________________________________
If you want to get the best response to a question, please check out FAQ222-2244 first

'If we're supposed to work in Hex, why have we only got A fingers?'

 

[ESquared]

The reason I want to program my own Dictionary object is that I found the speed of .Adding things to collections to be very slow. In my own code I will do work to try and optimize adding new words.

Last time I did this I used RtlMoveMemory for sorting, basically treating an array of strings as pointers and rearranging them at will.

I don't want to use the scripting object because of speed issues, and also I want additional functionality that I don't think it provides, like anagram serving. Admittedly, I haven't checked it out... hehe

 

[ESquared]

I can't build the index I had envisioned if there are nonalpha characters in the dictionary. Now what do I do?

Sigh.

 

[johnwm]

strongm,
Yes I was using the .Exists method

________________________________________________________________
If you want to get the best response to a question, please check out FAQ222-2244 first

'If we're supposed to work in Hex, why have we only got A fingers?'