Writing Knowledgebases for the Web

Formatting Question Pages

The format of the web pages that ask questions of the user are controlled by four entries in your main knowledgebase. These are:

question_top	<HTML><HEAD></HEAD> <BODY bgcolor=#FFFFFF text=#000000> <H2><IMG SRC="/images/bullet.gif" HSPACE=10><FONT COLOR=#0000A0> Amzi! Internet Site Guide </FONT></H2> <FORM METHOD="POST" ACTION="/cgi-bin/kwcgibasic.exe%system(cgi_parameters)%">
question_separator	<P>
value_separator	<BR>
question_bottom	<P><INPUT NAME="submit" TYPE="submit" VALUE="Continue"></INPUT> </FORM> <BODY></HTML>

In question_top you need to specify the correct URL to run KnowledgeWright. Note the special system goal, system(cgi_parameters), this is used to pass the CGI parameters from one form to the next. The example above is for running the CGI interface. For, a Java Servlet, the <FORM> tag might read:

<FORM METHOD=""GET"" ACTION=""/amzi/KW"">

A question separator is inserted between each question and a value separator is inserted between each menu choice in a question.

Building Output Pages

The preferred way to dynamically build a web page is to create a text object that is the goal of your knowledgebase. Typically you set output_top and output_bottom in your main knowledgebase object to include your <HTML>, <HEAD> and <BODY> tags. For example:

output_top	<HTML><HEAD></HEAD><BODY bgcolor=#FFFFFF text=#000000>
output_bottom	</BODY></HTML>

Then in your text object you would include the dynamically generated sections along with fixed text and HTML tags. For example, the Amzi! Website Advisor has a text object with the following contents:

text

<H2><IMG SRC="/images/bullet.gif" HSPACE=10>
<FONT COLOR=#0000A0> Amzi! Internet Site Guide</FONT></H2>
<HR><P><FONT SIZE=3><I><B>
We recommend these articles, demos, newsgroups, and other sites for you...
</B></I></FONT><P>

%resource%

<HR><P><FONT SIZE=3><I><B>
We recommend these tools, products and services for you...
</B></I></FONT><P><UL>

%tool%

</UL><P>All <A HREF="/company/index.htm">Amzi!</A> products include 
free tech support (via internet, phone and fax), are not copy 
protected and are royalty-free.</P>

Note, the %resource% and %tool% in the text. These are sub-goals. So the reasoning engine will try to determine values for these objects. For the Website Advisor, resource and tool are rule_sets. The values of the rules are text that have embedded HTML tags. For example, here are some entries in the resource rule_set:

conditions

value

exploreProlog = "yes" and 
prologFamiliarity \= "Very"

"To understand more about Prolog's unique 
backtracking and unification algorithms 
see the PC AI article 
<A HREF=""/articles/prolog_under_the_hood.htm"">
Prolog Under the Hood</A>. For more information on 
<B>Prolog</B>, see the 
<A HREF=""http://fas.sfu.ca/0/cs/people/ResearchStaff/jamie/personal/prolog-faq.1"">
Prolog FAQ</A> and the 
<A HREF=""news:comp.lang.prolog"">Prolog newsgroup</A>. 
Find lots of source code in 
<A HREF=""http://users.ox.ac.uk/~popx/"">Jocelyn Paine's 
public domain library</A> and the 
<A HREF=""http://www.cs.cmu.edu/Groups/AI/html/repository.html"">
Carnegie Mellon Repository</A>."

embeddedDevelopment = "yes" and 
languagesAndTools include "Java"

"See the PC AI article on 
<A HREF=""/articles/java_class.htm"">Java Meets Prolog</A> 
which describes the <B>Amzi! Java Class</B>. If you have 
the Sun JDK, you can run the 
<A HREF=""http://www.amzi.com/download/demos/java_ship.zip"">
shipping advisor demo</A>."

Note, that these text values could also include further sub-goals in percent signs. This provides alot of flexibility for generating and formatting your documents.

If you choose to have multiple goals in your knowledgebase, then the HTML in output_continue is used to continue the reasoning process and to proceed onto the next goal. In this case the user will see a series of questions followed by some results, then another series of questions and more results.