This is a study of chatbot interpreter routines coded in PHP, which includes open source chatterbot knowledgebases converted to pure PHP.  It is a small PHP script with no XML, nor SQL used.  This Chatbot study in PHP is entirely coded in all PHP.

Think of this experiment as like a cheap GPS which has recorded speech audio, but has no text to speech.  Although a cheap GPS does not pronounce the street names, like the more expensive GPS systems do, it still surprisingly works very well.  Is a similar phenomenon possible with chatbots?  What will be revealed if we peal back advanced routines, like recursion or wildcard matching?

The question that this study is attempting to answer, is whether a chatbot simply needs at least several thousand records for its robot brain to cross a certain threshold to gain the quality of being believable?

http://elizabot.com/study/

Suggestion:  Please include small paragraphs while testing such as, “Hello. I am testing you. Hopefully you will not mind.  Is that OK?”  or anything else is also fine, “I am here. You are there. We are where?”  It does not have to be anything special.  The “Chatbot study in PHP” parses punctuation including periods, question marks, commas, etc.  With that suggested, please feel free to do whatever comes natural testing… This is simply a suggestion. Thanks!

I tried a few things with it. It doesn’t seem to handle apostrophes very well and some of the output seemed redundant, however there was usually something sensible buried in the responses. I’ll postpone my vote until you’ve tweaked it a bit more as I believe it looks promising.

Thank you Andrew!  You have no idea how helpful your comments are. Apostrophes should be parsed now. The corrections you suggested sort of spread across the whole knowledge base. With your help and the help of others, I think the outputs are more closely matching the inputs… So the response seems to have gotten more coherent. Inevitably, more gaps in the conversion process will be revealed as this experiment progresses. 

Special thanks to those who have helped anonymously. If you would like to help, you are welcome to do so anonymously. When I said, “small PHP script”, I was referring to the “Chatbot Study in PHP” interpreter which a small number of lines of code. The total number of lines of PHP code in the “Chatbot Study in PHP” is considerably higher due to the sizes of the open source knowledge bases being converted to PHP.  So whenever, unformatted responses are located, an update may be applied globally.

Thanks again for helping.

Reference: http://elizabot.com/study