I can think of a solution using Pandorabots custom HTML. The custom HTML feature allows you to embed AIML
<template>
expressions in an HTML file. The server evaluates these AIML statements before serving the page as a normal HTML file.
You can use
<that index="1,1"/>
and
<that index="2,1"/>
to access the two sentences in the response.
You can set up something like this in the custom HTML file:
RESPONSE 1 = <template><that index="1,1"/></template>
<br/>
RESPONSE 2 = <template><that index="2,1"/></template>
<br/>
CONJUCTION = <template><srai>CONJUNCTION <that index="1,1"/> CONJOIN <that index="2,1"/></srai></template>
<br/>
The CONJUCTION category is:
<category>
<pattern>CONJUNCTION * CONJOIN *</pattern>
<template><star/> and <star index="2"/></template>
</category>
I’m using this CONJUNCTION category to eliminate the setence-splitter “.”.
The bot’s original response is:
Human: How tall are you ? What is your weight ?
ALICE: My anticipated body size is over 2 meters. As a software program, my weight is zero.
The custom HTML returns:
RESPONSE 1 = As a software program, my weight is zero.
RESPONSE 2 = My anticipated body size is over 2 meters.
CONJUCTION = As a software program my weight is zero and My anticipated body size is over 2 meters
This sketch of a solution is limited because it’s assuming that all the outputs have exactly two sentence responses. It seems straightforward to elaborate a variety of
<condition>
items that would take appropriate action depending on one, two, three or an arbitrary number of response sentences (would you really want the bot to say “I’m six feet tall and I weigh 200 pounds and I am 30 years old and I live in California and my favorite color is green”?—maybe not). Perhaps this solution may give you some ideas.
RSS AI Zone
