I’m just interested in learning what platforms people are using to build chatbots and why it was the chosen one?
We’re using our own custom made platform. Backend is build with Python and NLP tasks are handled with various neural networks on Tensorflow.
Interesting, we use Python as well. Wondering how you are utilizing Tensorflow, we just haven’t seen the need to tap into any AI in delivering the conversations themselves, however we do use it around analytics and detecting trends in the flows. The next obvious step there though is to create conversation paths based on that data.
We use TF to perform semantic calculation tasks and make intent detection more relevant.
We built out our own using Node.js, our back-end easily supports Kik, Telegram and Facebook Messenger. All we need to do is tweak a properties file everytime we want to make a new bot. To keep things clean we are using the MVC architectural pattern.
We chose to build our own instead of using the many solutions popping up because we are building games and with the pre-made solutions out there you can only do so much. We need complete control so we can not only take full advantage of each platform but really go nuts with each of our games logic.
Cool what kind of games are you building? We are working on some simple trivia and guessing games for a client, would be cool for a company to offer games you could tap into any platform.
I’ve been building anything that requires programmatic logic in Meya as it closely resembles how I’d architect a platform if I wanted to do it myself.
That being said its still a pain in the ass to handle a lot of basic things in Meya i.e. variable store, states and even simple things like branching is doable but way more hassle than just writing a simple if statement
All other platforms currently are too simple
I use Microsoft Bot Framework to build bots and LUIS for NLP. It seamlessly integrates with multiple channels such as messenger, skype, kik etc without having to change anything in the code.
We’ve prototyped our bot in botkit and wit.ai, and have run into some difficulty with handling conversations. Botkit also doesn’t seem to be easily platform agnostic.
After we sold the prototype we’ve begun working on v2, building our own native version in Node.js and utilizing some home-grown NLP processing.
Ok. sounds like i should check out Meya. thx
We’re building our own custom bot and not using a platform, with that said our stack is Elixir for the backend/data/etc and ReactJS/Node for the interface.
We are building bots with Node.js and Hapi.js. Using a lot node nodules for each bot platform.
We have already built our own conversational cloud on AWS, that allows us to deploy a chat interface anywhere and we’ve built our own bot platform on Node.js, while leaning on Wit.Ai for certain scenarios.
Using Node.js stack and quite a few npm modules with it along with MongoDB, and Wit.ai stories. I’m currently working on a custom (limited) neural net for data decision intelligence (not for NLP) for a new project, and may migrate to TensorFlow if there is enough traction.
have built our own platform - Python for the backend and our own neural networks (RNN, LSTM) for language analysis and intent detection
Luis is nice but doesnt integrate the dutch language at this moment ;(
indeed. if you need help, let me know, we’ve built our own in-house networks for analysis in several languages.
We are using API.ai, Django Rest, and deployed on Heroku. We are exploring IBM Watson conversation service.