1. Improve customer service
2. Facilitate ecommerce
3. Provide better, more personal engagement
4. Enhance customers’ experience
5. Address customers’ needs in real-time
6. Integrate utility into the brand experience
7. Have conversation at scale
8. Talk openly and honestly, even about sensitive topics
9. Create opportunities for services integration
10. Improve customer loyalty programs
A number of AIML files like Knowledge.aiml have been embedded into the bot already from the AliceBot.org site to give it a little bit of brains (try asking it “how far is alpha centauri”). The parser takes in chat input and then runs it through the various AIML file it loads. I worked with a developer in Belarus via UpWork to upgrade PascAlice to that it supports Firemonkey in Delphi Berlin on Android, IOS, OSX, and Windows. We brought in the source code from this demo for the chat interface and upgraded it a bit with new virtual keyboard handling code. We replaced LibXmlParser with the cross platform OXml library. I wanted to implement some text to speech functionality so on Windows, IOS, and OSX it utilizes an unofficial Google Translate API but also supports a Voice RSS API for text to speech(TTS). The API based TTS works by just passing a translation URL to the TMediaPlayer component. On Android it utilizes the built in text to speech functionality using a component built by Delphi developers Jim McKeeth and Jeff Overcash.
Two other pieces of functionality that have been added to the bot are the ability to say “google delphi” or “google firemonkey” and it will open in an in app TWebBrowser. The bot for example could have functionality that opens the web browser based on various chat responses. The second piece of functionality that was added is a dynamic form response very similar to how the Facebook Messenger Bot works where it embeds a piece of content in the chat response. The dynamic form is built via JSON and you can test out the functionality by pressing the Inline Form button to see it in action. It will create a TFrame on the fly and then populate it with buttons from the JSON that is passed to it. In this case there are two lines with a Yes and No button. If the Yes button is pressed it will send a navigation to the TWebBrowser as well. The dynamic form functionality is pretty powerful and only a small amount of functionality is built into the demo. The dynamic from can also be saved and loaded from JSON using the Duck Duck Delphi library. Keep in mind this is a quick prototype with limited testing.
You can also connect the chat bot to a variety of REST APIs like Motion.ai, Api.AI, PandoraBots, or even Amazon Echo using TRESTClient if needed. Alternately, you can run your own AIML server with Program O for PHP. Or you could pull out the AIML parser and deploy it in your own Delphi server implementation. Who’s going to be the first developer to turn the Delphi documentation into a chat bot? Let us know in the comments how you use it!
Update: Fixed a small cross platform issue.