AutoHotkey Community

[BoBo¨]

[PhiLho]

On the phones I have/had, I always deactivate this feature... Perhaps I just don't use it correctly, but it seems to me it gets in a way instead of helping.
But hey, if somebody finds this interesting to use...
_________________
vPhiLho := RegExReplace("Philippe Lhoste", "^(\w{3})\w*\s+\b(\w{3})\w*$", "$1$2")

[BoBo¨]

[JDN]

They may be correct. But my experience with these kinds of systems is that you spend far more time looking and checking to see what word the system predictied than you save by not typing the word.

There have been many similar systems produced during the past 20 years. And listening to the marketing hype for these systems, you would think they had discovered the greatest thing since sliced bread. But these systems never seem to catch on. They all just seem to die a quiet death. You have to wonder why that is. If they are so fabulous, why have they all just died on the vine. Gosh! I sure wish I could remember the names of some of the predecessors to this. I remember one huge marketing splash for a product like this based somewhere in Massachusetts.

Why do I seem so critical of this thing? Well, every time FastWrite makes an error, you have to delete the word or partial word they incorrectly produced. Plus you have to then type the correct word from scratch.

Correcting an error in that kind of way, very quickly kills any savings. The reason for that is that you have to spend three times as much time dealing with an error than you do just typing a normal word. For every error, you have to:

1) take the time to look and check to see what word it produced

2) take the time to delete that word or part of the word

3) take the time to type the correct word or allow the system to predict and produce the word again.

[BoBo¨]

In the short instance I guess you're right. But as T9 acts as a 'self teaching'-system (with every word it 'learns'/keeps), its benefit will/should increase, right? That might be less obvious for the English language but for German its different because Germans are using much longer words, (eg. "Lohnsteuerhilfeverein","Bushaltestelle","Autoreparaturwerkstatt", ...) which could be a pain if typed on such tripple-char-keys (don't know the English word for those, sorry ) of a mobile.

Therefore I assume the main thing to think about is how to teach T9 its vocabulary/dictionary (btw. a quite popular request if you Google for T9). Once that's done it should be an interesting feature.

[JDN]

Hi Bobo,

I'm not sure I know exactly what you mean by "T9". Do you mean the package shown in the first post of this thead named "FastWrite"?

If so, please consider the following method for German that I believe would be very superior to the approach outlined in "FastWrite". But let me see if I can describe it breifly. That will be a challenge. I should say that I don't know the German language. I once created a word frequency list for German text and was shocked there appeared to be 4 or 5 words that all meant "the" - such as Die, Der, Das, Dat, etc. That word "the" is the most frequently occurring word in English. Typically it comprises close to 5 percent of all English document. By that, I mean to say, the five keystrokes " the " (including the leading and trailing blank) make up 5 percent of all the keystrokes in most any English document.

But here is my suggestion for a superior approach. If you know - with certainty - a sequence to produce your long word, then that will be much faster than using an approach in which you have to check and see if the word was produced correctly. Since a huge proportion of the time you spend typing German is spent on a comparitively small set of words, then it makes sense to learn some of these words (as many as you are comfortable with) and then produce those words with some shorthand sequences - with certainty. By certainty, I mean ...

Suppose you know the sequence <Semicolon - L> produces the word "Lohnsteuerhilfeverein" and <Semicolon - B> produces the word "Bushaltestelle" and <Semicolon - A> produces the word "Autoreparaturwerkstatt".

Now, when I say the sequence <Semicolon - L>, I mean you press the Semicolon and the L simultaneously. I call this a "multi-press". The user manual for AutoHotkey (AutoHotkey.chm) says it has an equivalent facility that can do this. But I've not tried it yet.

By the way, I have written a small utility program called JDN WORD FREQUENCY. This program reads a text file. It can handle huge text files, by the way. It is not afraid of files as large as 100 megabytes. I have not uploaded it to AutoHotkey.net. If there appears to be much interest, I can always upload a version of this utility.

[BoBo¨]

[JDN]

I have uploaded two utility programs relevant to this thread. They are very similar but produce somewhat different outputs.

They both input a text document and count all the unique words. Then they output a report showing all the words in the document together with a count of how many times each word occurred and how many keystrokes were used to type each word. The idea is to determine which words occur most frequently and exactly how many keystrokes were made to produce each word.

But this is only a general idea of how they work. You really should read the readme files to see exactly how they differ. They are both very, very tiny files and will only take a second to read.

1) The first progam is called WC.EXE (Word Count) and was written by Branko Radovanovic of CROATIA

WORD COUNT - http://www.autohotkey.net/~JDN/PD/WORDCOUNT.ZIP

2) I wrote the second program and I think it has a more extensive output report. It's called JDNWC.

JDN WORD COUNT - http://www.autohotkey.net/~JDN/PD/JDNWC.ZIP

[PhiLho]