CogniSieve is the first example of a user programmable Cognitron for Mac OS PowerPC systems. Cognitrons are robotic or software components that perform recognition tasks. This shareware version of CogniSieve contains a sieve to recognize email addresses and telephone numbers. However, because the associational rules for recognition can be edited by the user, CogniSieve can be customized to extract any items of interest such as names, addresses, a particular column of numbers in a table of share prices, or even complex semantic tokens such as all sentences and phrases announcing parties (as in an occasion for fun). Unlike keyword based methods, it would not be confused by sentences that contain the word party used in another sense, such as "involvement" (to be a party to) or political entities.
Cognitrons are not something totally new. The Newton's Smart Assist feature which recognises and "understands" simple commands such as "dinner with John" is an early example. The latest example is Apple's recently announced Data Detectors. (See http://www.research.apple.com/research/tech/AppleDataDetectors/Default.html.) A lot more of them will surface with the rise of agent software. They are essential for smarter User/Agent communication.
How to Use CogniSieve
First make sure you have three files, the CogniSieve Player, CogniSieve Manual, the Read Me and the Phone/Email/URL Sieve folder. Make a back-up copy of Phone/Email/URL Sieve folder, as it may get corrupted during usage.
CogniSieve files come in two flavors. There is one Rules file (regular document icon) where all the associational rules are stored, and there can be one or more associated Sieve files (magnifier icon) which provide a choice of ways for one to view or use the rules.
Open a CogniSieve file by double clicking its icon. The first time you open a CogniSieve Sieve file, after copying it to a new location, you will get a open File dialog asking you to locate its linked Rule file. When you get this dialog, just navigate to the requested Rule file and double click on it. Then the Sieve file will open.
When the user opens the Phone/Emial Sieve provided, he should see a window with two panes separated by a blue line in the middle. The lower pane is the Input pane where one could type or paste text to be parsed. The upper pane with item headers in the top row, is the Output pane where the results of sieving will be displayed.
( Before parsing any text, you may wish to edit or rearrange the order of the items headers in the Output Pane by selecting them and using the usual Cut or Paste commands. If you delete the item headers, make sure you have saved a copy of the Sieve file with headers. The output will be neater if the order of the item headers correspond to the order of the items in the text to be parsed. )
I. Paste some text into the Input (lower) pane or type into it. After a two second delay, the program will automatically start to scan the text. You can force it to rescan at any time by choosing menuitem Process Input under the Parser Menu (Command-3). If anything is recognised in the text, it will appear in one of the columns in the upper (Output) pane.
II. Copy the results of sieving in the Output pane and paste into other programs such as Excel. Since the sieved items are separated by tabs and returns, when you paste them into a spreadsheet each item will be placed into individual cells.
When you quit, the program may ask if you want to save changes. Saving will remember any changes you made to the window size, position as well as all the text that you have processed. (There is a bug that may occur when a lot of text is saved. The progress bar window may not close after all the saving is done. Just click on the Stop button to dismiss it.)
If you wish to see how the parsing is done, open the middle pane by dragging the blue line apart and choose menuitem Process Input. To see the processing take place stepwise, choose the menuitem Tokenise (Command-1) followed by repeated Do Single Step (Command-2) actions.
Phone Number Sieve
The rules for this sieve will allow it to recognise all telephone numbers with the following variants. We realise this may not cover every country.
7 digit local numbers
#######
### ####
###-####
8 digit local numbers
########
#### ####
####-####
#-local_no
###-local_no
(###) local_no
#-###-local_no
phone_no ( alphanumeric string )
e.g. 123-2123 (fax)
Email Sieve
The Email sieve is somewhat restrictive. Addresses will be recognised if they are of the form:
id string @ address string
The id string is an alphanumeric or numeric string. Symbols such as #, - or period may occur in the string, but other symbols will prevent recognition.
The address string is of the form of multiple repeats of (alpha-numeric).(alpha-numeric).(alpha-numeric) etc.
Despite these restrictions, we think most of the common email addresses should be recognised.
URL Sieve
Web addresses are recognised if they are of the form:
http:// path name or ftp:// path name
The path name is any repeated sequence of alphanumeric strings separated by periods or slashes.
Shareware Fee and Registration
This Player version of CogniSieve is shareware. If you find it useful and would like to continue to use it after a trial period of one month, you must register your copy of the software by using the included Register application. Payment is handled through a company called Kagi Software.
Shareware fee : $10.00
CogniSieve has the following pricing.
Single user licenses, $10 per user
A Site License costs $270 (roughly equal to 27 users) and covers all locations for your organization within a 160 kilometer radius of your site (100 miles). One big advantage of a Site License is that you do not need to keep track of how many people at your site are using the software.
A World-Wide License costs $1000 and it covers all locations for your organization on the planet earth.
Paying for CogniSieve is fairly simple. Open the Register program that accompanies CogniSieve. Enter your name, your email address, and the number of single user licenses you desire for each program you wish to purchase (or Site or Word-Wide licenses). Save or Copy or Print the data from the Register program and send the data and payment to Kagi. More specifics on the Register program to follow. Kagi handles my payment processing.
If paying with Credit Card or First Virtual, you can email or fax the data to Kagi. Their email address is sales@kagi.com and their fax number is +1 510 652-6589. You can either Copy the data from Register and paste into the body of an email message or you can Save the data to a file and you can attach that file to an email message. There is no need to compress the data file, it's already pretty small. If you have a fax modem, just Print the data to the Kagi fax number.
Payments sent via email are processed within 3 to 4 days. You will receive an email acknowledgement when it is processed. Payments sent via fax take up to 10 days and if you provide a correct internet email address you will receive an email acknowledgement.
If you are paying with Cash or USD Check you should print the data using the Register application and send it to the address shown on the form, which is:
Kagi
1442-A Walnut Street #392-2WX
Berkeley, California 94709-1405
USA
Customized Sieve Service
This shareware version does not support editing of the rules. If you are interested in a version that allows editing, or in our Customized Sieve Service, please contact us by email for further details. You should make a list of the general characteristics of the patterns that you would like to be able to recognise and have some examples. Including a few examples of items that might seem similar but should not get recognised will also be useful.
Contacting Us
Visit our web site for the latest news and versions of shareware:
http://home1.pacific.net.sg/~mchen/xanatech/
To contact us it's best to use email:
mchen@pacific.net.sg
but if you really must use snail mail, then you can write to:
