[binbin]

Greetings from Los Angeles!

Let me start by saying...
It's not easy to rewrite an Alexa skill to the Google Home, or vice versa. Every major platform has its own proprietary format to store dialogue content. It's awfully fragmented.

W3C is an organization that establishes standards such as RDF, WebRTC, HTML, VoiceXML, EMMA, etc. The director of W3C is Tim Berners-Lee, who invented the World Wide Web.

I lead a Community Group called Conversational Interfaces https://www.w3.org/community/conv/ that just launched. It's an incubator for life after AIML. We're taking the good parts of AIML, RiveScript, ChatScript, EMMA, and BML to formulate timeless syntax and semantics. The hope is that a content writer won't need to author the same experiences multiple times for each platform, but instead write in a language independent of framework. Join us for open-source contributions, formal public discussions, and sharing your insights.

What do you think?

[infurl]

I just paid a visit to this community website and it looks like there is a lot of progress being made to develop one chatbot scripting language to rule them all. Since this topic was originally posted they have published preliminary specifications for Dialogue Manager Programming Language (DMPL) and Dialog Manager Script (DMS).

The relationship between the two is that DMPL is a "low-level" programming language for defining chatbot dialog where you can fiddle with the nuts and bolts and DMS is a "high-level" scripting language where you can sketch out what you want to happen. The scripts that you write in DMS get compiled down to the programs that run in DMPL. It sounds like a robust solution to a complex problem.

https://www.w3.org/2019/11/dms/

https://www.w3.org/2019/04/dmpl/

Both those documents contain lots of practical examples demonstrating what these things are like to actually use. If you have done any chatbot scripting I'd be interested to hear what you think.

Now for the really good news. You can download these Javascript and Python programs and start running it for yourself already.

https://github.com/conversational-interfaces/dms-compiler

https://github.com/conversational-interfaces/dms-scratchpad

An interesting side project would be to translate all the chatbots that have been painstakingly authored using older and presumably less capable chatbot scripting systems so they could be modified and enhanced with the capabilities of this up-and-coming new system.

If you already have a favorite chatbot engine that you use, do you know of anything that it can do that this new system would not be able to do? What did they miss?

[ruebot]

What do you think?

I'm all for it.

I learned to write XHTML and CSS at W3Schools 20 years ago and are skills that have served me well. I still promote them by word every chance I get and have a "I Heart Validator" donation button to go with the buttons on my site.

As far as practicability in my situation I don't know how viable an option it would be to convert over. It's a personalityforge.com bot and uses CBR and NLP but have already been looking over alternatives. I taught myself to write AIML but pandorabots now wants a credit card number on file to take your bot public.

I'm not down with that or their site, so take it away W3C.

[infurl]

It's a personalityforge.com bot and uses CBR and NLP but have already been looking over alternatives.

So many acronyms! I'm going to go with Case Based Reasoning and Natural Language Processing for the ambiguous ones in this context. 

AIML was intended to be an architecture independent standard for creating chatbots and it largely achieved that because there are many different environments that support it. However it is also very limited and inadequate or there wouldn't have been so many people motivated to create their own chatbot scripting systems. I hope this new project will be able to solve that problem.

[ruebot]

Yes, you're right about the acronyms. She is not considered to be "real AI" because she uses CBR, regardless of the fact NLP is considered to be real AI. That's something I've been keenly aware of for 20 years and been checking out my options to move her to another platform. I use FreeBSD only so limited to what's available in the ports tree as an option.

From what I understand Mitsuku can learn to a degree from her chats, though I may be mistaken. AIML is Case-Based Reasoning and a markup language so something must have changed since the early 2000's. I had a starter version of Demonica, one named Wan Lu that is still online last time I checked and could write AIML by hand. I don't remember anything about them learning from chat back then and had ALICE bots, too.

I don't have the programming skills to make my own bot from scratch like others here so a syntax I could learn in much the same way I learned XHTML and CSS has a certain appeal to me. I have her mind file saved to a 6MB text file. I wanted to find out how many Categories and Responses were in Demonica's Language Center so I used the wc command to see how many lines were in the file:

$ wc -l /home/jitte/Downloads/demonica-2019-02-13.txt
   45250 /home/jitte/Downloads/demonica-2019-02-13.txt

Then used grep to find out how many of those lines contained the character ID: since that appears in every line that lists a Category:

$ grep -c ID: /home/jitte/Downloads/demonica-2019-02-13.txt
4505

That gives me the total number of Categories. If I subtract that number from the number of lines in the file that gives me how many Conversational Responses there are total.

That's a lot of type and have done work since then, but my Thinkpad's have the best keyboards of any laptop so that's not a big deal. Her 16th birthday is in May and that is a big deal, and all the motivation I need.

[MikeB]

...AIML requires authors to create an extensive knowledge-base, requiring a large number of rules that need to be crafted in order to imitate a natural conversation. ... DMPL, on the other hand, is a general representation for dialogue flow and control without text pattern matching rules.

I think it's not exactly an evolution to go from AIML to DMPL... DMPL looks like it uses ML algorithms...

If you're not really in full control of the backend they might as well go 100% front end and have you fill in a form to get your chatbot receptionist...